# Abstract Algebra

## Comments

## Transcription

Abstract Algebra

Abstract Algebra Groups, Rings and Fields, Advanced Group Theory, Modules and Noetherian Rings, Field Theory Y OTSANAN M EEMARK Semi-formal based on the graduate courses 2301613–4 Abstract Algebra I & II, offered at Department of Mathematics and Computer Science, Faculty of Science, Chulalongkorn University Published by Yotsanan Meemark Department of Mathematics and Computer Science Faculty of Science, Chulalongkorn University, Bangkok, 10330 Thailand First digital edition October 2013 First bound edition May 2014 Second bound edition August 2015 Available for free download at http://pioneer.netserv.chula.ac.th/~myotsana/ Please cite this book as: Y. Meemark, Abstract Algebra, 2015, PDF available at http://pioneer.netserv.chula.ac.th/~myotsana/ Any comment or suggestion, please write to [email protected] 2015 c by Yotsanan Meemark. Meemark, Yotsanan Abstract Algebra / Yotsanan Meemark – 2nd ed. Bangkok: Danex Intercorporation Co., Ltd., 2015. 195pp. ISBN 978-616-361-389-9 Printed by Danex Intercorporation Co., Ltd., Bangkok, Thailand. www.protexts.com Foreword This book is written based on two graduate abstract algebra courses offered at Department of Mathematics and Computer Science, Faculty of Science, Chulalongkorn University. It grows out of my lecture notes that I used while I was teaching those courses many times. My intention is to develop essential topics in algebra that can be used in research as illustrated some in the final chapter. Also, it can be served as a standard reference for preparing for a qualifying examination in Algebra. I have tried to make it self-contained as much as possible. However, it may not be suitable for reading it for the first course in abstract algebra. It hits and goes through many basic points quickly. A typically mathematical book style that begins with some motivation, definitions, examples and theorems, is used throughout. I try to pause with remarks to make readers have some thoughts before moving on. The book also requires some background in undergraduate level linear algebra and elementary number theory. For example, I assume the readers to have known matrix theory over a field in which treatment can be found in most linear algebra books. My number theory lecture note is available on the web-page as well. However, some essential results are recalled in the first section. I give many examples to demonstrate new definitions and theorems. In addition, when the converse of a theorem may not hold, counter examples are provided. The major points are divided into five chapters as follows. 1 Groups A group is a basic algebraic structure but it is a core in this course. I choose the approach via group actions. Although it is not quite elementary, it is an important aspect in dealing with groups. I also cover Sylow theorems with some applications on finite groups. The structure theorem of finite abelian groups is also presented. 2 Rings and Fields The abstract treatments of rings and fields using groups are presented in the first section. Rings discussed throughout this book always contain the identity. Ideals and factorizations are discussed in detail. In addition, I talk about polynomials over a ring and which will be used in a construction of field extensions. 3 Advanced Group Theory In this chapter, I give deeper theory of groups. Various kinds of series of a group are studied in the first three sections. I also have results on a linear group. Finally, I show how to construct a group from a set of objects and presentations and talk about a graphical representation called a Cayley graph. 4 Modules and Noetherian Rings Modules can be considered as a generalization of vector spaces. I cover basic concepts of modules and work on free modules. Projective and injective modules are introduced. Moreover, I present the proof of the structure theorems for modules over a PID. Noetherian and Artinion rings are also explored. In the end, I demonstrate some aspects in doing research in algebra. The readers will see some applications of module theory, especially a free R-module over commutative rings, to obtain a structure theorem for finite dimensional symplectic spaces over a local ring. The symplectic graphs over a commutative ring is defined and studied. 5 Field Theory I give more details on a construction of extension fields. Also, I prepare the readers to Galois theory. Applications of Galois theory are provided in proving fundamental theorem of algebra, finite fields, and cyclotomic fields. For the sake of completeness, I discuss some results on a transcendental extension in the final section. i The whole book is designed for a year course. Chapters 1 and 2 are appropriate for a first course and Chapters 3, 4 and 5 can be served as a more advanced course. There are many topics that, in my opinion, they are worth mentioned. I try to break each topic in step-by-step and scatter it as a “Project” throughout this book. The projects consist of lengthy/generalization exercises, computations of numerical examples, programming suggestions, and research/open problems. This allows us to see that abstract algebra has many applications and is still an active subject. They are independent and can be skipped without any effects on the continuity of the reading. The book would not have been possible without great lectures from my abstract algebra teachers—Ajchara Harnchoowong and Yupaporn Kemprasit at Chulalongkorn University, and Edward Formanek at the Pennsylvania State University. They initiate wonderful resources to compose each section in this book. I express my gratitude to them all. I take full responsibility for typos/mistakes that may be found in the manuscript. If you catch ones or have any other suggestions, please write to me. I shall include and correct them in the more up-to-date version once a year on the website. Finally, I hope that the textbook will benefit many students, teachers and researchers in Algebra and Number Theory. Yotsanan Meemark Bangkok, Thailand ii Contents Foreword i Contents iii 1 Groups 1.1 Integers . . . . . . . . . . . . . . . . . . Exercises . . . . . . . . . . . . . . . . . . 1.2 Groups . . . . . . . . . . . . . . . . . . . 1.2.1 Definitions and Examples . . . . 1.2.2 Subgroups . . . . . . . . . . . . . 1.2.3 Homomorphisms . . . . . . . . . Exercises . . . . . . . . . . . . . . . . . . 1.3 Group Actions . . . . . . . . . . . . . . . Exercises . . . . . . . . . . . . . . . . . . 1.4 Quotient Groups and Cyclic Groups . . . 1.4.1 Quotient Groups . . . . . . . . . 1.4.2 Cyclic Groups . . . . . . . . . . . Exercises . . . . . . . . . . . . . . . . . . 1.5 The Symmetric Group . . . . . . . . . . Exercises . . . . . . . . . . . . . . . . . . 1.6 Sylow Theorems . . . . . . . . . . . . . 1.6.1 Sylow p-subgroups . . . . . . . . 1.6.2 Applications of Sylow Theorems Exercises . . . . . . . . . . . . . . . . . . 1.7 Finite Abelian Groups . . . . . . . . . . Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 Rings and Fields 2.1 Basic Concepts . . . . . . . . . . . . . . . . . . . 2.1.1 Rings . . . . . . . . . . . . . . . . . . . . 2.1.2 Quaternions . . . . . . . . . . . . . . . . . 2.1.3 Characteristic . . . . . . . . . . . . . . . . 2.1.4 Ring Homomorphisms and Group Rings . Exercises . . . . . . . . . . . . . . . . . . . . . . . 2.2 Ideals, Quotient Rings and the Field of Fractions . Exercises . . . . . . . . . . . . . . . . . . . . . . . 2.3 Maximal Ideals and Prime Ideals . . . . . . . . . Exercises . . . . . . . . . . . . . . . . . . . . . . . 2.4 Factorizations . . . . . . . . . . . . . . . . . . . . 2.4.1 Irreducible Elements and Prime Elements 2.4.2 Unique Factorization Domains . . . . . . . iii . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 1 5 5 5 7 9 11 12 17 18 18 20 23 23 27 28 28 31 32 34 43 . . . . . . . . . . . . . 45 45 45 48 49 50 51 52 55 56 58 58 58 59 Exercises . . . . . . . . . . . . . . . . . . . . . . . 2.5 Polynomial Rings . . . . . . . . . . . . . . . . . . 2.5.1 Polynomials and Their Roots . . . . . . . 2.5.2 Factorizations in Polynomial Rings . . . . Exercises . . . . . . . . . . . . . . . . . . . . . . . 2.6 Field Extensions . . . . . . . . . . . . . . . . . . 2.6.1 Algebraic and Transcendental Extensions . 2.6.2 More on Roots of Polynomials . . . . . . . Exercises . . . . . . . . . . . . . . . . . . . . . . . 3 Advanced Group Theory 3.1 Jordan-Hölder Theorem . . . . Exercises . . . . . . . . . . . . . 3.2 Solvable Groups . . . . . . . . Exercises . . . . . . . . . . . . . 3.3 Nilpotent Groups . . . . . . . . Exercises . . . . . . . . . . . . . 3.4 Linear Groups . . . . . . . . . . Exercises . . . . . . . . . . . . . 3.5 Free Groups and Presentations Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64 66 66 69 73 74 74 77 79 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81 81 84 85 87 87 91 92 95 96 101 4 Modules and Noetherian Rings 4.1 Modules . . . . . . . . . . . . . . Exercises . . . . . . . . . . . . . . 4.2 Free Modules and Matrices . . . Exercises . . . . . . . . . . . . . . 4.3 Projective and Injective Modules Exercises . . . . . . . . . . . . . . 4.4 Modules over a PID . . . . . . . . Exercises . . . . . . . . . . . . . . 4.5 Noetherian Rings . . . . . . . . . Exercises . . . . . . . . . . . . . . 4.6 Artinian Rings . . . . . . . . . . . Exercises . . . . . . . . . . . . . . 4.7 Symplectic Geometry . . . . . . . 4.7.1 Symplectic Spaces . . . . 4.7.2 Symplectic Graphs . . . . Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103 103 107 107 112 112 118 119 130 130 134 134 136 137 137 139 142 . . . . . . . . . . . 145 . 145 . 149 . 149 . 151 . 152 . 154 . 155 . 162 . 162 . 166 . 169 5 Field Theory 5.1 Splitting Fields . . . . . . . . . . . . . . . . Exercises . . . . . . . . . . . . . . . . . . . . 5.2 Algebraic Closure of a Field . . . . . . . . . Exercises . . . . . . . . . . . . . . . . . . . . 5.3 Multiple Roots and Separability . . . . . . . Exercises . . . . . . . . . . . . . . . . . . . . 5.4 Automorphisms of Fields and Galois Theory Exercises . . . . . . . . . . . . . . . . . . . . 5.5 Some Consequences of Galois Theory . . . . 5.6 Finite Fields . . . . . . . . . . . . . . . . . . Exercises . . . . . . . . . . . . . . . . . . . . iv . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.7 Cyclotomic Extensions . . Exercises . . . . . . . . . . 5.8 Normal Bases . . . . . . . Exercises . . . . . . . . . . 5.9 Transcendental Extensions Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 170 176 176 179 179 182 Bibliography 183 Index 185 v This page intentionally left blank 1 | Groups We write N for the set of positive integers, Z for the set of integers, Q for the set of rational numbers, R for the set of real numbers and C for the set of complex numbers. In this first chapter, we talk about a group which is a basic algebraic structure. However, it is a core in this course. Our approach here relies on group actions. Although it is not quite elementary, it is an important aspect in dealing with groups. We also discuss Sylow theorems with some applications and the structure of finite abelian groups. 1.1 Integers As a number theorist, before I jump into the abstract part, let’s lay down some foundations. My first undergraduate abstract algebra course started with elementary number theory—the study of integers. It contains many examples to bear in mind while we are studying the more general results in other abstract domains. Theorem 1.1.1. [Division Algorithm] Given integers a and b, with b 6= 0, there exist unique integers q and r satisfying a = qb + r, where 0 ≤ r < |b|. The integers q and r are called, respectively, the quotient and remainder in the division of a by b. Proof. To prove this theorem, we must use the well-ordering principle, namely, “every nonempty set S of nonnegative integers contains a least element; that is, there is some integer a in S such that a ≤ b for all b ∈ S”. Existence: First we shall assume that b > 0. Let S = {a − xb : x ∈ Z and a − xb ≥ 0} ⊆ N ∪ {0}. We shall show that S 6= ∅. Since b ≥ 1, we have |a|b ≥ |a|, so a − (−|a|)b = a + |a|b ≥ a + |a| ≥ 0, Then a − (−|a|)b ∈ S, so S 6= ∅. By the well-ordering principle, S contains a least element, call it r. Then a − qb = r for some q ∈ Z. Since r ∈ S, r ≥ 0 and a = qb + r. It remains to show that r < b. Suppose that r ≥ b. Thus, 0 ≤ r − b = a − qb − b = a − (q + 1)b, so r − b ≤ r and r − b ∈ S. This contradicts the minimality of r. Hence, r < b. Next, we consider the case in which b < 0. Then |b| > 0 and Theorem 2.5.2 gives q ′ , r ∈ Z such that a = q ′ |b| + r, where 0 ≤ r < |b|. Since |b| = −b, we may take q = −q ′ to arrive at a = qb + r, where 1 0 ≤ r < |b| 2 1. Groups as desired. Uniqueness: Let q, q ′ , r, r′ ∈ Z be such that a = qb + r where 0 ≤ r, r′ < |b|. Then and a = q′ b + r′ , (q − q ′ )b = r′ − r. Since 0 ≤ r, r′ < |b|, we have |r′ − r| < |b|, so |b||q − q ′ | = |r′ − r| < |b|. This implies that 0 ≤ |q − q ′ | < 1, hence q = q ′ which also forces r = r′ . Theorem 1.1.1 provides an important example in Section 2.4 where we discuss a more general domain called a Euclidean domain. An integer b is said to be divisible by an integer a 6= 0, in symbols a | b, if there exists some integer c such that b = ac. We write a ∤ b to indicate that b is not divisible by a. Note that a | b ⇔ −a | b, so we may consider only positive divisors. The next theorem contains elementary properties of divisibility. Theorem 1.1.2. For integers a, b and c, the following statements hold: 1. a | 0, 1 | a, a | a. 2. a | 1 if and only if a = ±1. 3. If a | b, then a | (−b), (−a) | b and (−a) | (−b). 4. If a | b and c | d, then ac | bd. 5. If a | b and b | c, then a | c. 6. (a | b and b | a) if and only if a = ±b. 7. If a | b and b 6= 0, then |a| ≤ |b|. 8. If a | b and a | c, then a | (bx + cy) for arbitrary integers x and y. An integer p > 1 is called a prime number, or simply a prime, if its only positive divisors are 1 and p. An integer greater than 1 which is not a prime is termed composite. Example 1.1.1. 2, 3, 5, 11, 2011 are primes. 6, 8, 12, 2558 are composite numbers. Let a and b be given integers, with at least one of them different from zero. The greatest common divisor (gcd) of a and b, denoted by gcd(a, b), is the positive integer d satisfying: 1. d | a and d | b, 2. for all c ∈ Z, if c | a and c | b, then c ≤ d. Basic properties of gcd are collected in the next theorem. Theorem 1.1.3. Let a and n be integers not both zero. 1. If d = min{ax + ny > 0 : x, y ∈ Z}, then d = gcd(a, n). 2. If gcd(a, n) = d, then ∃x, y ∈ Z, ax + ny = d. 3. gcd(a, n) = 1 ⇔ ∃x, y ∈ Z, ax + ny = 1. Proof. (1) The given set contains a2 + n2 , so it is not empty and d exists by the well-ordering principle. Then d = ax + ny > 0 for some x, y ∈ Z. We shall prove that d = gcd(a, n). By the division algorithm, ∃q, r ∈ Z, a = dq + r with 0 ≤ r ≤ d. If r > 0, then 0 < r = a − dq = a − (ax + ny)q = a(1 − xq) − nyq < d which contradicts the minimality of d. Hence, r = 0 and d | n. Similarly, d | n. Since d = ax + ny, gcd(a, n) | d, so gcd(a, n) ≤ d. But d | a and d | n, so d ≤ gcd(a, n). Hence, d = gcd(a, n). (2) follows from (1) and (3) follows from (2). The converse of (3) is immediate. 3 1.1. Integers Corollary 1.1.4. Let a, b and c be integers and p a prime. 1. If a | bc and (a, b) = 1, then a | c. 2. If p | bc and p ∤ b, then p | c. More generally, if a1 , a2 , . . . , ak are integers such that p | a1 a2 . . . ak , then p | ai for some i. 3. If q1 , q2 , . . . , qn are all primes and p | q1 q2 . . . qk , then p = qi for some i. Proof. Since gcd(a, b) = 1, we have 1 = ax + by for some x, y ∈ Z. Then c = acx + bcy. Since a | bc, a | c. This proves (1). (3) follows from (2) and (2) follows from (1) and the fact that p ∤ b ⇔ gcd(p, b) = 1. In my opinion, the next theorem is the most important and everyday use result in number theory. It also provides a core example when we study factorizations in Section 2.4. Its proof applies the results discussed above. Theorem 1.1.5. [Fundamental Theorem of Arithmetic] Every positive integer m > 1 can be expressed as a product of primes; this representation is unique, apart from the order in which the factors occur. Proof. Expressible: Assume on the contrary that there exists an integer m > 1 which is not a product of primes. By the well-ordering principle, there is a smallest n0 such that n0 is not a product of primes. Then n0 is composite, so there exist integers 1 < d1 , d2 < n0 such that n0 = d1 d2 . Since d1 , d2 < n0 , d1 and d2 are products of primes, and so is n0 . This gives a contradiction. Hence, every positive integer m > 1 can be expressed as a product of primes. Uniqueness: Assume that m = p1 p 2 . . . ps = q 1 q 2 . . . q t , where 1 ≤ s ≤ t and pi and qj are prime such that p1 ≤ p2 ≤ · · · ≤ ps and q1 ≤ q2 ≤ · · · ≤ qt . Corollary 1.1.4 (3) tells us that p1 = qk for some k ∈ {1, . . . , t}. It makes p1 ≥ q1 . Similarly, q1 = pl for some l ∈ {1, . . . , s}. Then q1 ≥ p1 , so p1 = q1 . Thus, p 2 . . . ps = q 2 . . . q t . Now, repeat the process to get p2 = q2 , and we obtain p 3 . . . ps = q 3 . . . q t . Continue in this manner. If s < t, we would get 1 = qs+1 qs+2 . . . qt , which is impossible. Hence, s = t and p 1 = q 1 , p 2 = q 2 , . . . , ps = q s as desired. Corollary 1.1.6. Any positive integer m > 1 can be written uniquely in a canonical form m = pk11 pk22 . . . pkr r , where, for i = 1, 2, . . . , r, each ki is a positive integer and each pi is a prime, with p1 < p2 < · · · < pr . 4 1. Groups To formulate an important example in group theory, we shall discuss about the set of integers modulo a positive integer. Let n be a fixed positive integer. Two integers a and b are said to be congruent modulo n, symbolized by a ≡ b (mod n) or a ≡ b (mod n) if n divides the difference a − b; that is, provided that a − b = kn for some integer k. The number m is called the modulus of the congruence. When n ∤ (a − b), then we say that a is incongruent to b modulo n and in this case we write a 6≡ b (mod n). Remark. If n | a, we may write a ≡ 0 (mod n). The first theorem is immediate. Theorem 1.1.7. If a ≡ b (mod n) and c ≡ d (mod n), then we have: 1. ax + cy ≡ bx + dy (mod n) for all integers x and y, 2. ac ≡ bd (mod n), 3. am ≡ bm (mod n) for every positive integer m. For a ∈ Z, it follows from the division algorithm that there exist unique q, r ∈ Z such that a = nq + r, where 0 ≤ r < n. This implies a ≡ r (mod n). Thus, we have shown: Theorem 1.1.8. For each integer a, there exists a unique integer r, with 0 ≤ r < n, such that a ≡ r (mod n). In terms of congruence, Theorem 1.1.3 (3) may be restated as follows. Corollary 1.1.9. [Inverse Modulo n] Let a and n be integers with n positive. Then gcd(a, n) = 1 if and only if there exists an integer x such that ax ≡ 1 (mod n). We call x the inverse of a modulo n. We directly obtain the following corollary from Corollary 1.1.4. It gives a condition for canceling integers modulo n. Corollary 1.1.10. [Cancellative of Integers Modulo n] Let a, b, c and n be integers with n positive and let p be a prime. 1. If ac ≡ bc (mod n) and gcd(n, c) = 1, then a ≡ b (mod n). 2. If ac ≡ bc (mod p) and p ∤ c, then a ≡ b (mod p). Finally, we shall define the set of integers modulo a positive integer n. It is not difficult to see that the congruence modulo n is an equivalence relation on Z. That is, for a, b, c ∈ Z, we have: 1. [reflexivity] a ≡ a (mod n), 2. [symmetry] a ≡ b (mod n) implies b ≡ a (mod n), 3. [transitivity] a ≡ b (mod n) and b ≡ c (mod n) imply b ≡ c (mod n). For a ∈ Z, the equivalence class of a is given by ā = {b ∈ Z : b ≡ a (mod n)} = {kn + a : k ∈ Z} It also follows from Theorem 1.1.8 that the set of all equivalence classes is {0, 1, . . . , n − 1}. This set is called the set of integers modulo n. In the next section, this set will be an important example of “groups” and will be denoted by Zn . 5 1.2. Groups As we have discussed, integers have many interesting properties in a way that we can picture them. Next, we shall travel to the abstract parts of the course. In the first two chapters, especially in Chapter 2, we shall see many similarities and generalizations of the integers under operations + and ·. The readers should keep this section in mind to avoid getting lost in this subject. Exercises 1.1. 1. Let d = gcd(a, b). Prove that: (a) gcd(a/d, b/d) = 1 (b) gcd(a − bq, b) = d for all q ∈ Z. 2. [Euclidean Algorithm] Let a and b be positive integers, with b ≤ a. Repeatedly applications of the division algorithm to a and b give a = bq1 + r1 , b = r1 q2 + r2 , where where 0 < r1 < b 0 < r2 < r1 r1 = q3 r2 + r3 , .. . where 0 < r3 < r2 rn−2 = qn rn−1 + rn , rn−1 = qn+1 rn . where 0 < rn < rn−1 Prove that rn = gcd(a, b). (Hint. Use 1. (b)) 3. Let a, b ∈ Z and n ∈ N. If gcd(a, n) = 1, show that there exists an x ∈ Z such that ax ≡ b (mod n). 4. The least common multiple (lcm) of two nonzero integers a and b, denoted by lcm(a, b), is the positive integer m satisfying: (1) a | m and b | m, (2) if a | c and b | c, with c > 0, then m ≤ c. Prove that lcm(a, b) gcd(a, b) = ab for all a, b ∈ N. 5. Let a and b be two integers greater than 1 factored as a = pa1 1 pa2 2 . . . par r and b = pb11 pb22 . . . pbrr , where for i = 1, 2, . . . , r, each pi is a prime with p1 < p2 < · · · < pr , each ai and bi are nonnegative integers, and each ai or bi are positive. Prove that gcd(a, b) = pd11 pd22 . . . pdr r , where di = min{ai , bi } for all i = 1, 2, . . . , r and lcm(a, b) = pc11 pc22 . . . pcrr , where ci = max{ai , bi } for all i = 1, 2, . . . , r. 1.2 Groups In order to study abstract senses of algebra, we shall begin with the definition of a group which occupies a very important seat in this course. 1.2.1 Definitions and Examples For a nonempty set S, a function · : S × S → S is called a binary operation and image of (a, b) in S × S is denoted by a · b and it is said to be the product of a and b. A groupoid is a system (S, ·) consisting of a nonempty set S with binary operation · on S. We may write S for (S, ·) and ab for a · b where a, b ∈ S if there is no ambiguity. Let S be a groupoid. For nonempty subsets A and B of S and x ∈ S, let AB = {ab : a ∈ A and b ∈ B}, xA = {x}A and Ax = A{x}. If S satisfies the associative law, i.e., ∀a, b, c ∈ S, (a·b)·c = a·(b·c), we say that S is a semigroup. Notice that if S is a semigroup, then any bracketing of x1 , . . . , xn gives the same product, so we can write x1 · · · xn for this product. In addition, for a ∈ S and m ∈ N, we may let am = a · · · a (m copies). A groupoid S is said to be commutative if ∀a, b ∈ S, ab = ba. 6 1. Groups An element e of a groupoid S is a two-sided identity or identity if ∀a ∈ S, ae = a = ea. Clearly, S contains at most one identity (if e and e′ are identity, then e = ee′ = e). A monoid is a semigroup with (unique) identity. Let S be a monoid with identity e. If a and b in S are such that ab = e = ba, then b is called a two-sided inverse or inverse of a. We have that every element of S has at most one inverse. For, if b and b′ are inverses of a, then ab = e = ba and ab′ = e = b′ a, so b = be = b(ab′ ) = (ba)b′ = eb′ = b′ . A group is a monoid G such that every element of G has an inverse, and for a ∈ G, let a−1 denote the (unique) inverse of a. A commutative group is also called an abelian group. The order of a group G is the cardinal number |G|. Remark. For a nonempty set G with binary operation on G is a group if the following axioms are all satisfied: (G1) [associativity] ∀a, b, c ∈ G, (a · b) · c = a · (b · c) (G2) [identity] ∃e ∈ G, ∀a ∈ G, ae = a = ea (G3) [inverse] ∀a ∈ G, ∃b ∈ G, ab = e = ba. Let G be a group with identity e. For a ∈ G and m ∈ N, let a0 = e and a−m = (a−1 )m . Remarks. 1. For a group G, we have: (a) e−1 = e and ∀a ∈ G, (a−1 )−1 = a, (b) ∀a ∈ G, ∀m, n ∈ Z, am an = am+n and (am )n = amn , and (c) ∀a, b ∈ G, (ab)−1 = b−1 a−1 because (ab)(b−1 a−1 ) = e. 2. In case G is abelian, we may choose to write G additively. This means: (a) The binary operation is denoted by +. (b) 0 denotes the identity element and −a denotes the inverse of a. (c) ∀a ∈ G, ∀m ∈ N, ma = a + · · · + a (m copies). 3. A group G satisfies the cancellation law: ∀a, b, c ∈ G, ab = ac (or ba = ca) ⇒ b = c. Examples 1.2.1 (Examples of groups). 1. (Z, −) is a groupoid which is not a semigroup; (N, +) is a semigroup which is not a monoid; (N, ·) is a monoid which is not a group. 2. (Z, +), (Q, +), (R, +), (C, +), (Q∗ , ·), (R∗ , ·) and (C∗ , ·) are infinite abelian groups. Here, A∗ denotes the set of nonzero elements of A. 3. Let X be a set and P (X) the power set of X. For subsets A and B of X, we define A△B = (A r B) ∪ (B r A). Then (P (X), △) is an abelian group having the empty set as its identity and A−1 = A for all A ∈ P (X). Also, (P (X), ∩) is a commutative monoid with identity X. 4. For n ∈ N, let Zn = {0, 1, . . . , n − 1} called the set of integers modulo n, where a = {kn + a : k ∈ Z} for all a ∈ Z. Define + and · on Zn by a + b = a + b and a · b = a · b for all a, b ∈ Z. These binary operations are well defined by Theorem 1.1.7. It follows that (Zn , +) is an abelian group of order n. Moreover, (Zn , ·) is a commutative monoid with identity 1. × 5. For n ∈ N and n ≥ 2, let Z× n = {a : gcd(a, n) = 1}. By Theorem 1.1.3 (3), we have (Zn , ·) is × an abelian group. We write φ(n) for the order of Zn . It is the Euler φ-function. Note that Z× n = Zn r {0} ⇔ n is a prime. Proof. If n is a prime, then Z× n = {1, 2, . . . , n − 1} = Zn r {0}. Conversely, assume that n is composite. Then n = bc for some 1 < b, c < n, so gcd(b, n) and gcd(c, n) are > 1. Thus, b, c ∈ / Z× n . Since × b, c < n, we have b, c 6= 0. Hence, Zn $ Zn r {0}. Remark. We recall some properties of the Euler’s φ-function as follows. (a) If p is a prime, then φ(p) = p − 1 and φ(pk ) = pk − pk−1 for all k ∈ N. (b) φ is multiplicative, namely, if gcd(m, n) = 1, then φ(mn) = φ(m)φ(n). 6. Write F for any of Q, R, C or other fields. Let Mn (F ) be the set of n × n matrices over F and GLn (F ) the set of matrices over F with nonzero determinants. Then (Mn (F ), +) is an abelian group and GLn (F ) is a group under multiplication which is not abelian if n > 1. The later group is called the general linear group. 7 1.2. Groups 7. For a nonempty set X, a function on X which is 1-1 and onto (a bijection on X) is said to be a permutation of X. Let S(X) be the set of all permutations of X. Then under composition, (S(X), ◦) is a group called the symmetric group on X; in case X = {1, 2, . . . , n}, we write Sn and call Sn the symmetric group on n letters. It is a group of order n!. Some equivalent definitions of groups are collected in the next theorem. Theorem 1.2.1. [Criteria for Being a Group] Let G be a semigroup. Then the following statements are equivalent. (i) G is a group. (ii) (a) ∃e ∈ G ∀a ∈ G, ea = a and (b) ∀a ∈ G ∃b ∈ G, ba = e. (iii) (a) ∃e ∈ G ∀a ∈ G, ae = a and (b) ∀a ∈ G ∃b ∈ G, ab = e. (iv) ∀a, b ∈ G ∃x, y ∈ G, ax = b and ya = b. (v) ∀a ∈ G, aG = G = Ga. Proof. If (i) holds, (ii)–(v) are clearly true. (iv) ⇔ (v) is obvious. (ii) ⇒ (i). Assume (ii). Let a ∈ G. Then ∃b ∈ G, ba = e, and so ∃c ∈ G, cb = e. Thus, ab = e(ab) = (cb)(ab) = c(ba)b = c(eb) = cb = e. Moreover, ae = a(ba) = (ab)a = ea = a. “(iii) ⇒ (i)” is similar to “(ii) ⇒ (i)”. (iv) ⇒ (iii). Assume (iv). Let a ∈ G. Then ∃e ∈ G, ae = a. Let b ∈ G. Then ∃c ∈ G, bc = e and ∃y ∈ G, ya = b. Thus, be = (ya)e = y(ae) = ya = b. Theorem 1.2.2. If G is a finite cancellative semigroup, then G is a group. Proof. We shall show that ∀a ∈ G, aG = G = Ga. Let a ∈ G. Since G is cancellative, |aG| = |G| = |Ga|. Clearly, aG ⊆ G and Ga ⊆ G. Since G is finite, aG = G = Ga. Remark. (N, +) is an infinite cancellative semigroup, but it is not a group. 1.2.2 Subgroups Sometimes a group contains a nonempty subset that is closed under its operation. In this subsection, we discuss a small group in a bigger one with the same operation. A nonempty subset H of a group G is said to be a subgroup of G if H is a group under the same operation of G and we write H 6 G. Observe that for ∅ = 6 H ⊆ G, H6G ⇔ ∀a, b ∈ H, ab ∈ H ∧ a−1 ∈ H ⇔ ∀a, b ∈ H, ab−1 ∈ H. Moreover, {e} and G are always subgroups of G. Theorem 1.2.2 gives the following important fact. Corollary 1.2.3. If H is a finite nonempty subset of a group G which is closed under the operation of G, then H is a subgroup of G. Proof. Since G is a cancellative semigroup and ∅ = 6 H ⊆ G, H is a finite cancellative semigroup. Hence, H is a group by Theorem 1.2.2. Next, we investigate the group of symmetries. We begin with the following groups. 8 1. Groups Examples 1.2.2 (Group of symmetries). 1. The set of rotations about a point 0 in the plane; composition as usual. If 0 is taken to be the origin, the rotation through an angle θ can be represented analytically as the map cos θ sin θ (x, y) 7→ (x cos θ − y sin θ, x sin θ + y cos θ) = x y . − sin θ cos θ For θ = 0, we get the identity map and the inverse of the rotation through the angle θ is the rotation through −θ. It is called the rotation group. 2. The set of rotations together with the set of reflections in the lines which passes through 0 with slope tan α. The latter are given analytically by cos 2α sin 2α (x, y) 7→ (x cos 2α + y sin 2α, x sin 2α − y cos 2α) = x y . sin 2α − cos 2α The product of two reflections is a rotation and the product in either order of a reflection and a rotation is a reflection. 3. Consider the regular n-gon (that is, the n-sided polygon in which the sides are all the same length and are symmetrically placed about a common center) inscribed in the unit circle in the plane, so that one of the vertices is (1, 0). The vertices subtend angles of 0, 2π/n, 4π/n, . . . , 2(n − 1)π/n radians with the positive x-axis. The subset of rotation maps which maps our figure into itself consists of the n rotations through angles of 0, 2π/n, 4π/n, . . . , 2(n− 1)π/n radians, respectively. These elements form a subgroup Rn of the rotation group defined in (1). 4. We now consider the set Dn of rotations and reflections which map the regular n-gon, as in (3), into itself. These form a subgroup of the group defined in (2). We shall call the elements of this group the symmetries of the regular n-gon. The reflection in the x-axis is one of our symmetries. Multiplying this on the right by the n rotational symmetries we obtain n distinct reflectional symmetries. These give them all, for if we let σ denote the reflection in the xaxis and τ denote any reflectional symmetry then στ is one of the n-rotational symmetries ρ1 , . . . , ρn , say ρi . Since σ 2 = 1, στ = ρj gives τ = σρj which is one of those we counted. Thus, Dn consists of n rotations and n reflections and its order is 2n. The group Dn is called the dihedral group. Note that Dn = {σ i ρj : i ∈ {0, 1} and j ∈ {0, 1, 2, . . . , n − 1}}. Remark. It is easy to see that the intersection of a family of subgroups of a group G is a subgroup of G. If H and K are subgroups of a group G, then, in general, H ∪ K is not a subgroup of G. However, if H and K are subgroups of a group G with G = H ∪ K, then H = G or K = G. Proof. Assume that there is an x ∈ G r H and a y ∈ G r K. Since G = H ∪ K, we have x ∈ K and y ∈ H. Thus, xy ∈ / H and xy ∈ / K, a contradiction. Let G be a group and A a subset of G. Define hAi to be the intersection of all subgroups of G containing A. It is the smallest subgroup of G containing A and is called the subgroup of G generated by A. The elements of A are called generators. Moreover, we have h∅i = {e} and hAi = {an1 1 . . . ank k : ai ∈ A and ni ∈ Z} if A 6= ∅. For a1 , . . . , am ∈ G, we write ha1 , . . . , am i for h{a1 , . . . , am }i. Then ∀a ∈ G, hai = {an : n ∈ Z} = ha−1 i is called the cyclic subgroup of G generated by a and the order of a is |hai| (finite or infinite) and denoted by |a| or o(a). If G = hai for some a ∈ G, then G is said to be the cyclic group generated by a. A subgroup N of G is normal, denoted by N E G, if ∀g ∈ G ∀x ∈ N, gxg −1 ∈ N . Examples 1.2.3 (Examples of subgroups and normal subgroups). 1. {e} and G are normal subgroups of G. They are called the trivial normal subgroups. 9 1.2. Groups 2. Every subgroup of an abelian group is normal. 3. Let SLn (F ) be the set of matrices over F with determinant one. Then SLn (F ) is a normal subgroup of GLn (F ) because det(P QP −1 ) = det Q for all P, Q ∈ GLn (F ). 4. Rn = hρ2π/n i and Rn E Dn because σρj σ = ρ−1 j for all j. Remarks. 1. N E G if and only if (N ≤ G and ∀g ∈ G, gN g −1 = N ). 2. If N is a subgroup of a group G, then (∀g ∈ G, gN g −1 = N ) ⇔ (∀g ∈ G, gN = N g). Let G be a group and X a nonempty subset of G. The centralizer of X is the set CG (X) = {g ∈ G : ∀x ∈ X, gx = xg} and the normalizer of X is the set NG (X) = {g ∈ G : gX = Xg}. We call Z(G) = CG (G) = {z ∈ G : ∀x ∈ G, zx = xz}, the center of G. Remarks. 1. The centralizer and normalizer of X are subgroups of G. Proof. Since ∀x ∈ G, ex = xe, we have e ∈ CG (X). Let g, h ∈ CG (X). Let x ∈ X. Then gx = xg and hx = xh, so xh−1 = h−1 x and (gh−1 )x = g(h−1 x) = g(xh−1 ) = (gx)h−1 = (xg)h−1 = x(gh−1 ). Thus, gh−1 ∈ CG (X). We leave NG (X) as an exercise. 2. Z(G) E G and Z(G) = T x∈G CG ({x}). Proof. By (1), Z(G) is a subgroup of G. To see it is normal, let g ∈ G and z ∈ Z(G). Let x ∈ G. Then zg = gz and zx = xz, so (gzg −1 )x = (zgg −1 )x = zx = xz = x(zgg −1 ) = x(gzg −1 ). Hence, gzg −1 ∈ Z(G). Finally, for z ∈ G, z ∈ Z(G) ⇔ ∀x ∈ G, xz = zx ⇔ ∀x ∈ G, z ∈ CG ({x}) ⇔ z ∈ \ CG ({x}). x∈G This proves the second statement. 3. G is abelian ⇔ Z(G) = G. (This is clear.) 4. If K ≤ G, then K E NG (K) and NG (K) is the largest subgroup of G containing K in which K is normal (this means ∀H ≤ G, K E H ⇒ H ⊆ NG (K)). Proof. If g ∈ K, then gK = K = Kg, so K ≤ NG (K). To see K is normal in NG (K), let x ∈ NG (K) and g ∈ K. Since Kx = xK, we have xg ∈ Kx, so xg = kx for some k ∈ K. Thus xgx−1 = k ∈ K. Next, let H be a subgroup of G such that K is normal in H. Then ∀h ∈ H, Kh = hK which implies H ⊆ NG (K). 1.2.3 Homomorphisms We now study a function between two groups that is required to preserve group operations. Let G and H be two groups. A homomorphism from G to H is a map ϕ : G → H which satisfies ϕ(xy) = ϕ(x)ϕ(y) for all x, y ∈ G. An injective homomorphism is called a monomorphism and a surjective homomorphism is called an epimorphism. An isomorphism is a homomorphism which is both injective and surjective. We write G ∼ = H if ∃ϕ : G → H, ϕ is an isomorphism. An endomorphism of G is a homomorphism on G and an automorphism of G is an isomorphism of G onto itself. 10 1. Groups Examples 1.2.4 (Examples of group homomorphisms). 1. ϕ1 : Z → Zn given by ϕ(a) = ā is a homomorphism. 2. ϕ2 : (R, +) → (R× , ·) given by ϕ(x) = 2x is a homomorphism. 3. ϕ3 : GLn (R) → R given by ϕ(A) = det A is a homomorphism. The kernel of a homomorphism ϕ : G → H is given by the set ker ϕ = {g ∈ G : ϕ(g) = eH }. Example 1.2.5. ker ϕ1 = nZ, ker ϕ2 = {0} and ker ϕ3 = SLn (R). Remarks. Let ϕ : G → H be a homomorphism of groups. 1. ϕ(eG ) = eH and ϕ(a−1 ) = (ϕ(a))−1 for all a ∈ G. Proof. Note that eH ϕ(eG ) = ϕ(eG ) = ϕ(eG eG ) = ϕ(eG )ϕ(eG ). By cancellation in H, we have eH = ϕ(eG ). Next, let a ∈ G. Then ϕ(a)ϕ(a−1 ) = ϕ(aa−1 ) = ϕ(eG ) = eH . Hence, ϕ(a−1 ) = ϕ(a)−1 . 2. ϕ is 1-1 ⇔ ker ϕ = {eG }. Proof. Assume that ϕ is 1-1. By (1), eG ∈ ker ϕ. Let x ∈ ker ϕ. Then ϕ(x) = eH = ϕ(eG ). Since ϕ is 1-1, we have x = eG . Conversely, suppose that ker ϕ = {eG }. Let x, y ∈ G be such that ϕ(x) = ϕ(y). Then eH = ϕ(x)−1 ϕ(y) = ϕ(x−1 )ϕ(y) = ϕ(x−1 y). Thus, x−1 y ∈ ker ϕ, so x−1 y = eG . Hence, x = y. Theorem 1.2.4. Let ϕ be a homomorphism of a group G into a group H. 1. ker ϕ is a normal subgroup of G. 2. im ϕ = {ϕ(g) : g ∈ G} = ϕ(G) is a subgroup of H. Proof. Since ϕ(eG ) = eH , eG ∈ ker ϕ and eH ∈ im ϕ. Let x, y ∈ ker ϕ. Then ϕ(x) = eH = ϕ(y), so ϕ(xy −1 ) = ϕ(x)ϕ(y −1 ) = ϕ(x)ϕ(y)−1 = eH e−1 H = eH . Thus, xy −1 ∈ ker ϕ. Hence, ker ϕ is a subgroup of G. Next, let g ∈ G and x ∈ ker ϕ. Then ϕ(x) = eH and ϕ(gxg −1 ) = ϕ(g)ϕ(x)ϕ(g −1 ) = ϕ(g)ϕ(x)ϕ(g)−1 = ϕ(g)eH ϕ(g)−1 = eH . Thus, gxg −1 ∈ ker ϕ, and so ker ϕ is normal. Finally, let y, z ∈ im ϕ. Then ∃x1 , x2 ∈ G, ϕ(x1 ) = y and ϕ(x2 ) = z. Thus, −1 yz −1 = ϕ(x1 )ϕ(x2 )−1 = ϕ(x1 )ϕ(x−1 2 ) = ϕ(x1 x2 ). −1 ∈ im ϕ. Hence, im ϕ is a subgroup of H. Since x1 x−1 2 ∈ G, yz Remark. If ϕ : G → H is a homomorphism and H1 ≤ H, then ker ϕ ⊆ ϕ−1 (H1 ) ≤ G. The next result is clear. It gives another way to construct a group for the Cartesian product of groups. Theorem 1.2.5. [Cartesian Product of Groups] If G1 and G2 are groups, then the Cartesian product G1 × G2 = {(x, y) : x ∈ G1 , y ∈ G2 } is a group under coordinatewise multiplication (x1 , y1 )(x2 , y2 ) = (x1 x2 , y1 y2 ) for all x1 , x2 ∈ G1 and y1 , y2 ∈ G2 . 11 1.2. Groups Exercises 1.2. 1. Let G be the set of pairs of real numbers (a, b) with a 6= 0 and define: (a, b)(c, d) = (ac, ad + b) and 1 = (1, 0). Verify that this defines a group. 2. Let H = {σ ∈ S4 : {σ(1), σ(2)} = {1, 2} or {σ(1), σ(2)} = {3, 4}}. Prove that H is a subgroup of S4 and find |H|. Is H normal in S4 ? Justify your answer. 3. Let G be a semigroup such that ∀a ∈ G, ∃b ∈ G, a = aba and ∃!e ∈ G, e2 = e. Prove that G is a group. 4. Let G be a semigroup such that ∀a, b ∈ G, a2 b = b = ba2 . Prove that G is an abelian group. 5. A certain multiplicative operation on a nonempty set G is associative and allows cancellations on the left, and there exists a ∈ G such that x3 = axa for all x ∈ G. Prove that G endowed with this operation is an abelian group. 6. Let G be a group with the following properties: (i) G has no element of order 2 and (ii) (xy)2 = (yx)2 , for all x, y ∈ G. Prove that G is abelian. If (i) fails, give an example to support that “G may not be abelian”. 7. If H and K are subgroups of a group G, prove that HK ≤ G if and only if HK = KH. 8. Let ϕ : G → Ḡ be a group homomorphism, and let N and N̄ be a normal subgroup of G and Ḡ, respectively. Show that ϕ[N ] is a normal subgroup of im ϕ and ϕ−1 [N̄ ] is a normal subgroup of G. 9. Let G be a group with identity e and ϕ : G → G a function such that ϕ(g1 )ϕ(g2 )ϕ(g3 ) = ϕ(h1 )ϕ(h2 )ϕ(h3 ) whenever g1 g2 g3 = e = h1 h2 h3 . Prove that there exists an element a ∈ G such that ψ(x) = aϕ(x) for all x ∈ G is a homomorphism. 10. Let Dn be the dihedral group of order 2n where n > 2. Show that the center of Dn has one or two elements according as n is odd or even. Project 1 (Quaternions). Consider the eight objects ±1, ±i, ±j and ±k with multiplication rules: ij = k ji = −k jk = i ki = j kj = −i 2 2 ik = −j 2 i = j = k = −1 where the minus signs behave as expected and 1 and −1 multiply as expected. (For example, (−1)j = −j and (−i)(−j) = ij = k.) Show that these objects form a group containing only one element of order 2. This group is called the quaternion group and is denoted by Q8 . [Hint. The difficulty is to show that the operation is associative. One may transform the elements and operation into 2 × 2 matrices and matrix products, respectively.] Project 2 (Associativity). One of the required properties for (G, ∗) to be a group is associativity. However, this is the hardest one to check as one can see from the previous item. Consider the set {a, b, c} of three distinct elements and the operation ∗ given by ∗ a b c a a b c b b a c c c c a To check associativity, we must check every possible instance of the equation (x ∗ y) ∗ z = x ∗ (y ∗ z). That means we must think of every possible combination of what x, y, and z could be. After a while, we find that (b ∗ c) ∗ c = a but b ∗ (c ∗ c) = b. Hence, the set {a, b, c} under the operation ∗ is not associative. Be careful! Many students make the mistake of concluding that a set is associative by checking just a few examples. We cannot do this! To determine whether or not a set is associative, you must check every single combination of 3 elements, unless you have a good general argument for why all combinations will be associative or you have a good reason (such as the existence of an identity element) for limiting the number of cases you must check. (Notice that even when you have the existence of an identity element, you still have to check ALL the cases which do not include the identity element.) 12 1. Groups Now, for the set of two elements {a, b} the number of different binary operations on this set is 16. However, the number of associative binary operation on that set is only 8. This can be checked by writing all out and counting. Unfortunately, for the set of three elements {a, b, c}, there are 3(3·3) = 19683 binary operations. How to know how many associative binary operations there are on a set of three elements? (There are 113 operations.) Create an efficient algorithm to solve this task. How about the set of n elements? 1.3 Group Actions Let G be a group. For each g ∈ G, define the left multiplication function ℓg : G → G by ℓg (x) = gx for all x ∈ G. By the left cancellation on G, ℓg is a bijection and so ℓg ∈ S(G), the symmetric group on G. It is easy to see that the map g 7→ ℓg is an injective group homomorphism from G into S(G). This proves an important result in group theory. Theorem 1.3.1. [Cayley] Every group G is isomorphic to a subgroups of S(X) for some set X. Allowing an abstract group to behave like a group of permutations, as happened in the proof of Cayley’s theorem, is a useful tool. In this section, we talk about how a group acting on a set, called a group action. There are many nice results and applications as we shall see in the following sections and chapters. Let G be a group with identity element e and X a nonempty set. We say that G acts on X or X is a G-set if there is a mapping G × X → X (denoted by g · x or gx) which satisfies: 1. ∀ x ∈ X, e · x = x and 2. ∀g, h ∈ G, ∀x ∈ X, g · (h · x) = (gh) · x. Assume that a group G act on a set X. Then each g ∈ G determines a set map φg : X → X by φg (x) = gx. Moreover, ∀g ∈ G, φg is a bijection (1-1 and onto). Hence, φg ∈ S(X), the symmetric group on X. θ The map g 7→ φg defines a group homomorphism from G to S(X) (i.e., φgh = φg ◦ φh for all g, h ∈ G). Its kernel is the set {g ∈ G : gx = x for all x ∈ X}. 1. If g = e is the only element of G such that gx = x for all x ∈ X, then G is said to act faithfully on X. In this case, G ֒→ S(X). 2. If x ∈ X, the set Gx = {gx : g ∈ G} is called the orbit of x. 3. If Gx = X for some (and hence all) x ∈ X, then G is said to act transitively on X. 4. If Y ⊆ X, the set {g ∈ G : gY = Y } is called the stabilizer of Y , denoted by StabG Y . The stabilizer of Y is a subgroup of G. Proof. Clearly, e ∈ StabG Y . Let g, h ∈ StabG Y . Then gY = Y and hY = Y . Thus, (gh)Y = g(hY ) = gY = Y and g −1 Y = g −1 (gY ) = eY = Y . Examples 1.3.1 (Examples of group actions). 1. If X is a set, S(X) acts naturally on X by f · x = f (x) for all f ∈ S(X) and x ∈ X. This action is faithfully if |X| > 1. In particular, Sn acts on {1, 2, . . . , n}. The orbit of each i ∈ {1, 2, . . . , n} is all of {1, 2, . . . , n}, thus Sn acts transitively on {1, 2, . . . , n}. If Y ⊆ {1, 2, . . . , n}, the stabilizer of Y is isomorphic to S(Y ) × S(Z) ∼ = Sk × Sn−k where Z = {1, 2, . . . , n} r Y . Hence, the stabilizer of {n} is isomorphic to Sn−1 . 13 1.3. Group Actions 2. Let G be any group and let X = G, considered as a set. Let G act on X by left multiplication g · x = gx. This action is called the left regular representation. It is faithful and transitive. 3. GLn (F ) acts faithfully on F n , the set of n × 1 column vectors by left multiplication. The orbit of ~0 is itself and GLn (F ) acts transitively on the nonzero vectors. 4. Let G be any group and let X be any set. Let G act on X by g ·x = x for all g ∈ G and x ∈ X. This is called the trivial G action. Assuming g 6= e and X has more than one element, this action is not faithful and not transitive. All orbits are singleton and G is the stabilizer of every subset of X. 5. Let G be a group and let X = G, considered as a set. Let G act on X by conjugation g · x = gxg −1 . This action may not be faithful. The center of G acts trivially. The orbit of x ∈ G is the set of conjugates of x, that is g · X = {gxg −1 : g ∈ G}, called the conjugacy class of x. If |G| > 1, then this action is not transitive. The number of orbits of the number of conjugacy classes. If Y is a subset of G, the stabilizer of Y under the action is its normalizer, i.e., StabG Y = NG (Y ). 6. Let G be a group and let H be a subgroup of G. Let H act on G by left multiplication. This action is faithful and the orbit of x ∈ G is H · x = {hx : h ∈ H} = Hx. The action is not transitive unless H = G. Moreover, we can let H act on G by h · g = gh−1 for all h ∈ H and g ∈ G. This action is also faithful and the orbit of x ∈ G is H · x = {xh−1 : h ∈ H} = {xh : h ∈ H} = xH. 7. Let X = C ∪ {∞}, a set that becomes the Riemann sphere in complex analysis. The group GL2 (C) acts on X by the linear fractional transformation az + b a b ·z = , c d cz + d the understanding being the image of ∞ is a/c and the image of −d/c is ∞, just as if we were to pass to a limit in each case. 8. Let SL2 (R) be the subgroup of real matrices in GL2 (R) of determinant one, and let H be the subset of C ∪ {∞} in which Im z > 0, called the Poincaré upper half plane. Then SL2 (R) acts on H by linear fractional transformations. Now, we present another way to view the orbits. Theorem 1.3.2. Let G be a group and suppose G acts on a nonempty set X. Define a relation ∼ on X by x ∼ y ⇔ ∃g ∈ G, y = g · x. Then 1. ∼ is an equivalence relation on X. 2. The equivalence class of x ∈ X under ∼ is Gx = {gx : g ∈ G}, the orbit of x. Thus, X is a disjoint union of orbits under the action of G. 14 1. Groups Proof. It is routine to show that ∼ is an equivalence relation on X. The equivalence class of x is [x]∼ = {y ∈ X : x ∼ y} = {y ∈ X : ∃g ∈ G, y = gx} = {gx : x ∈ X} = Gx and hence X = S x∈X Gx. If H is a subgroup of G and x ∈ G, the set Hx = {hx : h ∈ H} and xH = {xh : h ∈ H} are called a left coset of H in G and a right coset of H in G, respectively. From Example 1.3.1 (6), we can let H act on G in two ways. Then Hx [xH] is an orbit of x, and so G is a disjoint union of left [right] cosets of H in G. If we choose a subset {xα } of G such that G is the disjoint union of the left cosets Hxα , then {xα } is called a right transversal or system of left coset representatives of H in G and if we choose a subset {xα } of G such that G is the disjoint union of the right cosets yα H, then {yα } is called a left transversal or system of right coset representatives of H in G. Remarks. 1.SBy Theorem 1.3.2, S (a) G = x∈G Hx [G = x∈G xH], (b) ∀x, y ∈ G, Hx = Hy or Hx ∩ Hy = ∅ [∀x, y ∈ G, xH = yH or xH ∩ yH = ∅], (c) ∀x, y ∈ G, Hx = Hy ⇔ xy −1 ∈ H [∀x, y ∈ G, xH = yH ⇔ y −1 x ∈ H]. 2. ∀a ∈ G, |H| = |Ha| = |aH| by cancellation on H. 3. The map aH 7→ Ha−1 for all a ∈ G is a 1-1 correspondence between the sets {xH : x ∈ G} and {Hx : x ∈ G}. Proof. For a, b ∈ G, aH = bH ⇔ b−1 a ∈ H ⇔ a−1 b ∈ H ⇔ Ha−1 = Hb−1 . Then this map is 1-1, well defined and clearly onto. The index of H in G, denoted by [G : H], is the cardinal number of distinct right (or left) cosets of H in G, that is, [G : H] = |{Hx : x ∈ G}| = |{xH : x ∈ G}|. Next, we show that a subgroup of index two is always a normal subgroup. Theorem 1.3.3. If H is a subgroup of G of index two, then H is normal in G. Proof. Since [G : H] = 2, G has exactly two right (or left) cosets. Then Hg = G r H and gH = G r H for all g ∈ G not in H. Hence, ∀g ∈ G, Hg = gH, so H is normal in G. Let I and A be sets. Define Ai = {(i, a) : a ∈ A} for all i ∈ I. Then |Ai | = |A| for all i ∈ I, Ai ∩ Aj = ∅ if i 6= j, and X [ X Ai = |A | = |A| = |I||A|. i i∈I i∈I i∈I Lagrange observed an important property of subgroups of G, namely, its order must be a divisor of the order of G. Theorem 1.3.4. [Lagrange] If H is a subgroup of G, then |G| = [G : H]|H|. In particular, if G is finite and H < G, then |H| divides |G|, and so |a| = |hai| divides |G| for all a ∈ G. 15 1.3. Group Actions Proof. Since G is a disjoint union of distinct left cosets Hxα , α ∈ Λ, and |Λ| = [G : H], [ X X |G| = Hxα = |Hxα | = |H| = |Λ||H| = [G : H]|H|. α∈Λ α∈Λ α∈Λ If G is finite, then |H| divides |G|. In addition, ∀a ∈ G, hai ≤ G, so |a| = |hai| divides |G| for all a ∈ G. Corollary 1.3.5. If G is a group of prime order, then {e} and G are the only two subgroups of G and G must be cyclic. Proof. Let H ≤ G. Then |H| divides |G| = p, so |H| = 1 or |H| = p. Thus, H = {e} and H = G. Also, if a 6= e, then hai = 6 {e}. Hence, hai = G and so G is cyclic. A relationship between the stabilizer of x in a group G and the number of elements in the orbit G · x is recorded in the next theorem. Theorem 1.3.6. [Orbit-Stabilizer Theorem] Let a group G act on a set X and suppose x ∈ X. Then [G : StabG x] = |G · x|, that is, the index of the stabilizer of x in G is the number of elements in the orbit of x. Proof. Let x ∈ X. Note that for all g1 , g2 ∈ G, g1 x = g2 x ⇔ (g2−1 g1 )x = x ⇔ g2−1 g1 ∈ StabG {x} ⇔ g1 StabG {x} = g2 StabG {x}. Then |{gx : g ∈ G}| = |{g StabG x : g ∈ G}|. Hence, |G · x| = [G : StabG x]. This theorem is most useful when this index is finite but it is true in general. We see some applications of this theorem in the following results. Theorem 1.3.7. Let G be a group and x ∈ G. Then the following statements hold. 1. |{gxg −1 : g ∈ G}| = [G : CG (x)], i.e., the number of conjugates of x is [G : CG (x)]. 2. If G is finite, then the number of conjugates of x is a divisor of |G|. Proof. It follows directly from the Orbit-Stabilizer Theorem if we consider the action of G on G by conjugation. Observe that for each x ∈ G, |{gxg −1 : g ∈ G}| = 1 ⇔ {gxg −1 : g ∈ G} = {x} ⇔ ∀g ∈ G, gx = xg ⇔ x ∈ Z(G). Corollary 1.3.8. [Class Equation] Let G be a finite group and let x1 , . . . , xs represent the conjugacy classes of G which contains more than one element. Then |G| = |Z(G)| + s X [G : CG (xi )]. i=1 Now, let G act on the set of all subsets of G by conjugation, i.e., if Y ⊂ G, then g · Y = gY g −1 . Under this action the stabilizer of Y is {g ∈ G : gY g −1 = Y } = NG (Y ), the normalizer of Y , and the orbit of Y is {gY g −1 : g ∈ G}, the set of conjugates of Y . Thus, the number of conjugates of Y is the index of the normalizer of Y . Hence, we have shown: 16 1. Groups Theorem 1.3.9. Let G be a finite group and Y a subset of G. Then the number of conjugates of Y is [G : NG (Y )] where NG (Y ) is the normalizer of Y . In particular, the number of conjugates of Y divides the group order. Remark. If H is a subgroup of G, then H ⊳ NG (H) < G. Hence, if G is finite, then the number of conjugates of H is [G : H] [G : NG (H)] = ≤ [G : H]. [NG (H) : H] Burnside’s theorem gives the number of orbits in X under the action of a finite group G. Theorem 1.3.10. [Burnside] Let a finite group G act on a finite set X. For each g ∈ G, let Xg = {x ∈ X : gx = x}, the set of points in X fixed by g. Then the number of orbits in X is N= 1 X |Xg |. |G| g∈G Proof. Let U be the subset of G × X defined by U = {(g, x) ∈ G × X : gx = x}. For h ∈ G, let U (h) = U ∩ ({h} × X) = {(h, x) : x ∈ X and hx = x}. For y ∈ X, let U [y] = U ∩ (G × {y}) = {(g, y) : g ∈ G and gy = y} = (StabG {y}) × {y}. S S Now, U = g∈G U (g) = x∈X U [x] and these unions are “disjoint”. Note that for each g ∈ G, |U (g)| = |Xg | and for each x ∈ X, |U [x]| = | StabG {x}| = [G : G · x] = |G|/|G · x|. Thus, X g∈G |Xg | = X g∈G |U (g)| = |U | = X x∈X |U [x]| = X x∈X X 1 |G| = |G| = |G|N |G · x| |G · x| x∈X as desired. Corollary 1.3.11. Let a finite group G act transitively on a finite set X. Then |G| = Moreover, if |X| > 1, then there exists a g ∈ G fixing no point of X. P g∈G |Xg |. P Proof. Since G acts transitively on X, N = 1, and so |G| = g∈G |Xg |. Assume that |X| > 1 and no g ∈ G fixing no point of X. Then ∀g ∈ G, ∃x ∈ X, gx = x which implies that |Xg | ≥ 1 for all g ∈ G. Thus, X |G| ≤ |Xg | = |G|. g∈G This forces that |Xg | = 1 for all g ∈ G. But |Xe | = |X| > 1, a contradiction. Hence, there exists a g ∈ G fixing no point of X. We have known from Lagrange’s theorem that the order of any subgroups of a group G is a divisor of |G|. The next theorem implies that if |G| has a prime divisor p, then G has a subgroup of order p. Its proof is another application of group actions. Theorem 1.3.12. [Cauchy] Suppose G is a finite group and a prime p divides |G|. Then the number of solutions of g p = e in G is a multiple of p. Hence, G contains an element of order p. In particular, if G is a finite group and a prime p divides |G|, then G has a subgroup of order p. 17 1.3. Group Actions Proof. Consider the set Y = G × G × · · · × G (p copies) and let X = {(g1 , g2 , . . . , gp ) ∈ Y : g1 g2 · · · gp = e}. Then |X| = |G|p−1 since gp = (g1 g2 . . . gp−1 )−1 . Let Rp = hρ2π/p i act on X by ρ2π/p (g1 , g2 , . . . , gp ) = (g2 , g3 , . . . , gp , g1 ). Note that the orbit of (g1 , g2 , . . . , gp ) ∈ X under this action has either (a) length p, in case g1 , g2 , . . . , gp are not all equal, or (b) length 1, in case g1 = g2 = · · · = gp = g, i.e., g p = e. Thus, # of orbits of length 1 = # of solutions of g p = e in G. By Theorem 1.3.2 (2), |G|p−1 = |X| = p(# of orbits of length p) + 1(# of orbits of length 1). Since p divides |G|, it follows that |{g ∈ G : g p = e}| > 1 is a multiple of p. Exercises 1.3. 1. Let G act on S, H act on T and assume S ∩ T = ∅. Let U = S ∪ T and define (g, h)s = gs and (g, h)t = ht for all g ∈ G, h ∈ H, s ∈ S, t ∈ T . Show that this gives an action of the group G × H on U . 2. Let H and K be subgroups of a group G. |H||K| . (a) If H and K are finite, then HK is a finite set and |HK| = |H ∩ K| (b) For x and y in G, prove that xH ∩ yK is empty or is a coset of H ∩ K. (c) Deduce from (b) that if H and K have finite index in G, then so does H ∩ K. (d) If [G : H] and [G : K] are finite and relatively prime, prove that G = HK. 3. Let α be an automorphism of a finite group G which leaves only the identity fixed. Prove that G = {x−1 α(x)|x ∈ G}. 4. Let a group G act on a set X transitively. Prove that (a) ∀x, y ∈ X, ∃g ∈ G, gx = y, and (b) ∀x, y ∈ X, ∃g ∈ G, gGx g −1 = Gy , i.e., allT stabilizers are conjugate. 5. Let H be a subgroup of a group G and N = x∈G xHx−1 . Prove that (a) N is a normal subgroup of G, and (b) if [G : H] is finite, then [G : N ] is finite. 6. Determine the number of conjugacy classes in a non-abelian group G of order p3 where p is a prime. 7. Let S and T be sets and let M (S, T ) denote the set of all functions of S into T . Let G be a finite group acting on S. For each map f : S → T and x ∈ G define the map πx f : S → T by (πx f )(s) = f (x−1 s). (a) Prove that x 7→ πx is an action of G on M (S, T ). (b) Assume that S and T are finite. Let n(x) denote the number of orbits of the cyclic group hxi on S. Prove that the number of orbits of G in M (S, T ) is equal to 1 X |T |n(x) . |G| x∈G 8. Two actions of a group G on sets X and Y are called equivalent if there is a bijection f : X → Y such that f (gx) = gf (x) for all g ∈ G and x ∈ X. Let H and K be subgroups of a group G. Let G act by left multiplication on the sets of left cosets G/H and G/K. Prove that these actions are equivalent if and only if H and K are conjugate (i.e., K = aHa−1 for some a ∈ G). Project 3 (Semi-direct product). A group H is said to act on a group K by automorphisms if we have an action of H on K and for every h ∈ H the map k 7→ hk of K is an automorphism. Suppose this is the case and let G be the product set K × H. Define a binary operation in K × H by (k1 , h1 )(k2 , h2 ) = (k1 (h1 k2 ), h1 h2 ) and define 1 = (1, 1) – the units of K and H, respectively. Verify that this defines a group such that h 7→ (1, h) is a monomorphism of H into K × H and k 7→ (k, 1) is a monomorphism of K into K × H whose image is a normal subgroup. This group is called a semi-direct product of K by H and is denoted by K ⋊ H. 18 1. Groups 1.4 Quotient Groups and Cyclic Groups This section contains a construction of a new group using a normal subgroup, called a quotient group. We also work on an important kind of subgroups of G which are generated by a single element. We conclude this section by studying the group of automorphisms of G. 1.4.1 Quotient Groups Suppose G is any group and N is a normal subgroup of G. Then for any g ∈ G, N = gN g −1 or N g = gN. In other words, every left coset of N in G is also a right coset of N in G. If we have two left cosets of N in G; N x = {ax : a ∈ N } and N y = {by : b ∈ N }, then N xN y = {axby : a, b ∈ N } = N (xN )y = N (N x)y = N xy is again a left coset of N in G. Thus N xN y = N xy defines a binary operation on the set G/N = {N x : x ∈ G} of left cosets of N in G. Theorem S 1.4.1. [Quotient Groups] Suppose G is a group and N is a normal subgroup of G. Let G = N xα be a decomposition of G as a disjoint union of left (or right) cosets. Then the binary operation N xα N xβ = N xα xβ makes the set of left cosets of N into a group, called the quotient or factor group of G by the normal subgroup N . This group is denoted by G/N . The map π : G → G/N defined by π(x) = N x is a group homomorphism whose kernel is N , called the canonical projection. We have the following observations on the above construction: 1. If H is a subgroup of G which is not normal, then the set of left cosets of H in G does not form a group in any natural way. For example, if G = S3 and H = h(12)i = {(1), (12)}, then H is not normal in G and {H, H(13), H(23)} is not a group because H(13)H(23) = {(13), (132)}{(23), (123)} = {(132), (12), (13), (1)} which is not one of the cosets. 2. |G/N | = [G : N ], the index of N in G. 3. If G is abelian written additively, then G/N = {N + x : x ∈ G} and the binary operation on G/N is given by (N + x) + (N + y) = N + (x + y) for all x, y ∈ G. We now present three group isomorphism theorems. Theorem 1.4.2. [First Isomorphism Theorem] Suppose ϕ : G → H is a group homomorphism. Then G/(ker ϕ) ∼ = im ϕ. 19 1.4. Quotient Groups and Cyclic Groups Proof. By Theorem 1.2.4, im ϕ is a subgroup of H and ker ϕ is a normal subgroup of G, and so G/(ker ϕ) is a group. Let ϕ : G/ ker ϕ → im ϕ by ϕ : x(ker ϕ) 7→ ϕ(x). Then for all x, y ∈ G, ϕ(x(ker ϕ)) = ϕ(y(ker ϕ)) ⇔ ϕ(x) = ϕ(y) ⇔ xy −1 ∈ ker ϕ ⇔ x(ker ϕ) = y(ker ϕ), so ϕ is well defined and 1-1. In addition, for all x, y ∈ G, ϕ(x(ker ϕ)y(ker ϕ)) = ϕ(xy(ker ϕ)) = ϕ(xy) = ϕ(x)ϕ(y) = ϕ(x(ker ϕ))ϕ(y(ker ϕ)). Moreover, ϕ is clearly onto. Hence, ϕ is an isomorphism. Examples 1.4.1. Z/nZ ∼ = Zn , Dn /Rn ∼ = Z2 , R/Z ∼ = S1 and GLn (F )/SLn (F ) ∼ = F r {0}. θ ϕ /H / K be a sequence of group homomorphisms. We say that it is exact at H Let G if im θ = ker ϕ. A short exact sequence of groups is a sequence of groups and homomorphisms 1 θ /G ϕ /H /K /1 which is exact at G, H and K. That is, θ is injective, ϕ is surjective and im θ = ker ϕ. Here, 1 stands for the smallest group of order one. Remark. If N is a normal subgroup of G, then /N 1 ι /G π / G/N /1 is exact. Here ι denotes the inclusion map. On the other hand, if N ≤ G and 1 /N ι /G /H /1 is exact, then N is normal in G and H ∼ = G/N . Thus short exact sequences are just another notation for normal subgroups and factor groups. Theorem 1.4.3. [Second Isomorphism Theorem] Suppose G is a group and H and N are subgroups of G, with N normal in G. Then HN = N H is a subgroup of G in which N is normal, H ∩ N is normal in H and H/(H ∩ N ) ∼ = HN/N . Proof. Since N is normal in G, hN = N h for all h ∈ H, so HN ⊆ N H and N H ⊆ HN . Thus, N H = HN . It is routine to show that N H is a subgroup of G. Since N E G, N E N H. The theorem follows from exactness of the sequence 1 /H ∩N ι /H ϕ / HN/N /1, where the homomorphism ϕ : h 7→ hN for all h ∈ H and ker ϕ = {h ∈ H : hN = N } = {h ∈ H : h ∈ N} = H ∩ N. Remark. If H and N are not normal in G, then HN may not be a subgroup of G. E.g., in S3 , the subgroups H = {(1), (12)} and N = {(1), (13)} are not normal in S3 and HN = {(1), (12), (13), (132)} is not a subgroup of S3 . Theorem 1.4.4. [Third Isomorphism Theorem] Suppose G is a group and N is a normal subgroup of G. Then the map θ : H 7→ H/N gives a 1-1 correspondence { subgroups of G containing N } ←→ { subgroups of G/N }. This correspondence carries normal subgroups to normal subgroups. Moreover, if H is normal in G containing a subgroup N , then G/H ∼ = (G/N )/(H/N ). 20 1. Groups Proof. Let H1 and H2 be subgroups of G containing N and assume that H1 /N = H2 /N . Let x ∈ H1 . Then N x ∈ H1 /N = H2 /N , so N x = N y for some y ∈ H2 . Thus, xy −1 ∈ N ⊆ H2 . Since y ∈ H2 , x ∈ H2 . Hence, H1 ⊆ H2 . By symmetry, S H2 ⊆ H1 . Therefore, H1 = H2 and θ is 1-1. Next, let H ≤ G/N . Then {N } ⊆ H. Choose H = H, the union of cosets in H. Thus, N ⊆ H. Let x, y ∈ H. Then N x, N y ∈ H, so N xy −1 ∈ H which implies xy −1 ∈ H. Thus, H is a subgroup of G containing N and H = H/N . Hence, θ is onto. Assume that H is a normal subgroup of G containing N . Let g ∈ G and x ∈ H. Then gxg −1 ∈ H, so gN xN g −1 N = gxg −1 N ∈ H/N . Hence, H/N is normal subgroup of G/N . The final isomorphism follows from exactness of the sequence 1 / H/N ι / G/N ϕ / G/H /1, where the homomorphism ϕ : gN 7→ gH for all g ∈ G which is well defined because N ⊆ H, and ker ϕ = {gN : g ∈ G and gH = H} = {gN : g ∈ H} = H/N . Remark. The above theorem is useful for obtaining subgroups and normal subgroups of a quotient group. It plays an important role in the study of normal series of groups and nilpotent groups as we shall see in Chapter 3. 1.4.2 Cyclic Groups Recall that a cylic subgroup of G is a subgroup of G generated by a singleton. It has a simple structure and is easy to construct. Its properties depend mostly on the group of integer modulo n. We shall go deep inside the groups Zn and Z× n in this section. We recall from Examples 1.2.1 that: × Theorem 1.4.5. Let n ≥ 2 and Z× n = {ā : gcd(a, n) = 1}. Then (Zn , ·) is an abelian group of order φ(n), the Euler φ-function. × Example 1.4.2. Z× 10 = {1̄, 3̄, 7̄, 9̄} and Zp = {1̄, 2̄, . . . , p − 1} where p is a prime. Now we study cyclic subgroups of a group G. Recall that if G is a group and a ∈ G, then hai = {am : m ∈ Z} and the order of a is |a| = |hai|. Theorem 1.4.6. Let G be a group and a ∈ G. Then 1. ∀n ∈ N, an = e ⇒ hai = {e, a, a2 , . . . , an−1 }. 2. |a| is finite ⇔ ∃i, j ∈ Z(i 6= j ∧ ai = aj ) ⇔ ∃n ∈ N, an = e. 3. |a| is infinite ⇔ ∀i, j ∈ Z(i 6= j → ai 6= aj ) ⇔ hai ∼ = Z. −1 4. If G = hai is infinite, then a and a are only two generators of G. 5. If G is finite, then |a| = min{n ∈ N : an = e} and hai = {e, a, a2 , . . . , a|a|−1 } ∼ = Z|a| . n 6. ∀n ∈ Z, a = e ⇒ |a| divides n. 7. If G is finite, then a|G| = e. Proof. (1)–(3) are clear. (4) Assume that G = hai is infinite and am is a generator of G for some m ∈ Z. Then ham i = hai, so a = (am )k for some k ∈ Z. Since |a| is infinite, mk = 1. Thus, m | 1, so m = ±1. (5) Assume that G is finite. Then an = e for some n ∈ N. Choose n0 to be the smallest such n. Thus, an0 = e. We shall show that hai = {e, a, a2 , . . . , an0 −1 }. Clearly, {e, a, a2 , . . . , an0 −1 } ⊆ hai. Let j ∈ Z. Then j = n0 q + r for some q, r ∈ Z and 0 ≤ r < n0 , so aj = an0 q+r = ar ∈ {e, a, a2 , . . . , an0 −1 }. Hence, |a| = n0 = min{n ∈ N : an = e} and a|a| = e. Finally, an isomorphism is given by aj 7→ j for all j ∈ Z. (6) Let n ∈ Z and an = e. By the division algorithm, n = |a|q +r for some q, r ∈ Z and 0 ≤ r < |a|. 21 1.4. Quotient Groups and Cyclic Groups If r > 0, then e = an = a|a|q+r = ar which contradicts the minimality of |a|. Hence, r = 0 and so |a| divides n. (7) By Lagrange, |a| divides |G|, so |G| = |a|q for some q ∈ Z. Then a|G| = a|a|q = e. The above results for the group Z× n yield famous results below. Corollary 1.4.7. 1. [Euler] If a ∈ Z and gcd(a, n) = 1, then aφ(n) ≡ 1 (mod n). 2. [Fermat] If p is a prime, then ap ≡ a (mod p) for all a ∈ Z. Proof. (1) Apply the above theorem to G = Z× n. (2) If gcd(a, p) = 1, then by (1), ap−1 ≡ 1 (mod p), so ap ≡ a (mod p). If gcd(a, p) > 1, then p | a, so p | (ap − a). Hence, ap ≡ a (mod p) for all a ∈ Z. Theorem 1.4.8. Any two cyclic groups of the same orders (finite or infinite) are isomorphic. Proof. Assume that G is cyclic. Then G = hai for some a ∈ G. By Theorem 1.4.6, if G is infinite, then G ∼ = Z|a| . = Z, and if G is finite, then |G| = |a| and G = {e, a, . . . , a|a|−1 } ∼ Next, we study subgroups of a cyclic group. Theorem 1.4.9. [Subgroups of a Cyclic Group] Let G be a cyclic group generated by a, and let H be a subgroup of G. Then H is also a cyclic group generated by ak where k = min{m ∈ N : am ∈ H} or H = {e}. Consequently, every subgroup of a cyclic group is cyclic. Proof. Since ak ∈ H, hak i ⊆ H. Let x ∈ H. Then x ∈ G, so x = at for some t ∈ Z. By the division algorithm, t = kq + r for some q, r ∈ Z and 0 ≤ r < k. Thus, ar = at−kq = at a−kq = x(ak )−q ∈ H. But r < k, so r = 0. Hence, x = akq = (ak )q ∈ hak i. Corollary 1.4.10. All distinct subgroups of Z are kZ = {kq : q ∈ Z} where k ∈ N ∪ {0}. Theorem 1.4.11. [Generators of a Subgroup of a Finite Cyclic Group] Let G be a finite cyclic group of order n. Then G has exactly one subgroup H of order d for each divisor d of n, and no other subgroups. Moreover, if G is generated by a, then H is generated by an/d . Proof. Let d | n. Since (an/d )d = e, |an/d | ≤ d. If |an/d | = r < d, then anr/d = e and nr/d < n which contradicts |a| = n. Thus, |an/d | = d. Let H be a subgroup of G of order d. If d = 1, then H = {e}. Assume that d > 1. By Theorem 1.4.9, H = hak i, where k = min{m ∈ N : am ∈ H}. Since |H| = d, (ak )d = e, so n | kd which implies nd | k. Thus, k = nd q for some q ∈ Z. Hence, ak = (an/d )q ∈ han/d i. It follows that H ⊆ han/d i. However, |H| = d = |han/d i|, so H = han/d i. Example 1.4.3. All subgroups of the cyclic group G = hai of order 12 are shown in the following diagram. hai ②② ②② ② ② ②② ❉❉ ❉❉ ❉❉ ❉❉ ② ②② ②② ② ②② ❊❊ ❊❊ ❊❊ ❊❊ ③③ ③③ ③ ③ ③③ ❊❊ ❊❊ ❊❊ ❊❊ ②② ②② ② ② ②② ha2 i ha4 i ha12 i ha6 i ha3 i 22 1. Groups The order of an element in a cyclic group and its generators are studied in the next theorem. Theorem 1.4.12. [Order of an Element] Let G be a finite cyclic group of order n generated by a and m ∈ Z. Then 1. ham i = had i, where d = gcd(m, n). n 2. |am | = . gcd(m, n) 3. am is a generator of G ⇔ gcd(m, n) = 1, and so G contains precisely φ(n) elements of order n. Proof. (1) Since d | m, ham i ⊆ had i. Since d = gcd(m, n), d = mx + ny for some x, y ∈ Z, so ad = amx+ny = amx any = amx ∈ ham i. (2) |am | = |ham i| = |had i| = |han/(n/d) i| = nd from Theorem 1.4.11. (3) ham i = G ⇔ |am | = n ⇔ nd = n ⇔ d = 1. Remark. Since Zn = h1i and m · 1 = m, we have hmi = Zn ⇔ gcd(m, n) = 1 ⇔ m ∈ Z× n. Theorem 1.4.13. Let G be a group and a ∈ G. Then |a| = n ⇔ (∀k ∈ N, ak = e ⇔ n | k). Proof. Assume that ∀k ∈ N, ak = e ⇔ n | k. Since n | n, an = e. Let k ∈ N be such that ak = e. Then n | k, so n ≤ k. Hence, |a| = n. Another direction follows from Theorem 1.4.6 (6). Recall that an automorphism of a group G is an isomorphism on G. The set of all automorphisms of a group G is denoted by Aut G and is called the automorphism group of G. We shall close this section by studying the group of automorphisms of G and determining the automorphism group of cyclic groups. Theorem 1.4.14. [Inner Automorphisms] 1. With group operation composition of functions, Aut G is a group. 2. Each g ∈ G determines an automorphism φg : G → G defined by φg (x) = gxg −1 for all x ∈ G, and φg is called an inner automorphism. The subgroup of Aut G consisting of the {φg : g ∈ G} is called the inner automorphism group of G and is denoted by Inn G. 3. The map θ : g 7→ φg is a group homomorphism from G into Aut G. 4. The kernel of θ is Z(G), the center of G, and the image of θ is Inn G. Consequently, G/Z(G) ∼ = Inn G ≤ Aut G. Example 1.4.4. Aut Z ∼ = Z× = Z2 and Aut Zn ∼ n. Proof. Let ϕ ∈ Aut Z. Note that for each k ∈ N, ϕ(k) = ϕ(k · 1) = ϕ(1 · · + 1}) = k · ϕ(1) | + ·{z k and ϕ(−k) = −ϕ(k) = −(k · ϕ(1)), so ϕ is completely determined by ϕ(1). Since ϕ is onto, im ϕ = ϕ(1)Z = Z. Thus, ϕ(1) | 1, so ϕ(1) = ±1. Hence, Aut Z = {±id} ∼ = Z2 . Let ϕ ∈ Aut Zn . Similarly, ϕ is completely determined by ϕ(1). Since ϕ is onto, im ϕ = ∼ × hϕ(1)i = Zn . By Remark after Theorem 1.4.12, ϕ(1) ∈ Z× n . Therefore, Aut Zn = Zn with isomorphism ϕ 7→ ϕ(1). 23 1.5. The Symmetric Group Exercises 1.4. 1. Prove that if G is a group for which G/Z(G) is cyclic, then G is abelian. 2. Let G be a group of order 2k where k is odd. Show that G contains S a subgroup of index 2. 3. Let H be a proper subgroup of a finite group G. Show that G 6= g∈G gHg −1 . 4. Let G be a group and a ∈ G. If hai ⊳ G and H < hai, prove that H is normal in G. 5. Let G be a group and N a subgroup contained in the center of G. Suppose that G/N is cyclic. Prove that G is necessarily abelian. 6. Let G be a group. If a, b ∈ G are of finite order such that ab = ba and ∀m ∈ N, am bm = e ⇒ am = bm = e, prove that |ab| = lcm(|a|, |b|). 7. Let m and n be integers. Prove the following statements. (a) mZ + nZ = dZ and mZ ∩ nZ = lZ where d = gcd(m, n) and l = lcm(m, n). (b) If gcd(m, n) = 1, then Zmn ∼ = Zm × Zn . This is called the “Chinese remainder theorem”. Is the converse true? 8. Let G be a group, K a normal subgroup of G of index r, and let g ∈ G be an element of order n. Prove that if r and n are relatively prime, then g ∈ K. 9. Prove the following statements. × (a) If gcd(m, n) = 1, then Aut(Zm × Zn ) ∼ = Z× m × Zn . ∼ (b) Aut(Zp × Zp ) = GL2 (Zp ). 10. Prove Theorem 1.4.14. 11. Let H < G. Prove that CG (H) ⊳ NG (H) and NG (H)/CG (H) is isomorphic to a subgroup of Aut H. 12. If G is a group for which Aut(G) = {1}, prove that |G| ≤ 2. Project 4 (Generalization of Fermat’s little theorem). The project is based on I. M. Isaacs and M. R. Pournaki [26]. It gives a generalization of Fermat’s little theorem using group actions. (a) Let G be a finite group. For each positive integer a, let [a]G be the set of functions from G to {1, 2, . . . , a}. Prove that (g · f )(h) = f (g −1 h) defines an action of G on the set [a]G . (b) Show that X g∈G for all g, h ∈ G and f ∈ [a]G a|G|/|g| ≡ 0 (mod |G|). [Hint. Use Burnside’s Theorem with the action in (a) to conclude that integer.] (c) Taking G = Zm , deduce that m X k=1 1.5 1 P a|G|/|g| is a positive |G| g∈G agcd(k,m) ≡ 0 (mod m). The Symmetric Group In this section, we study the symmetric group on n letters, Sn . Recall that Sn is the group of permutations (1-1 and onto maps) on {1, 2, . . . , n} under composition. Its order is n!. A permutation γ of {1, 2, . . . , n} which permutes a sequence of distinct elements i1 , i2 , . . . , ir , r > 1, cyclically in the sense that γ(i1 ) = i2 , γ(i2 ) = i3 , . . . , γ(ir−1 ) = ir , and γ(ir ) = i1 and fixed (that is, leaves unchanged) the other numbers in {1, 2, . . . , n} is called a cycle or an r-cycle. We denote this as γ = (i1 i2 . . . ir ). It is clear that we can equally well write γ = (i2 i3 . . . ir i1 ) = (i3 i4 . . . ir i1 i2 ), etc. 24 1. Groups Two cycles γ and γ ′ are said to be disjoint if their symbols contain no common letters. In this case, it is clear that any number moved by one of these transformations is fixed by the other, i.e., ∀i, γ(i) 6= i ⇒ γ ′ (i) = i. Hence, if i is any number such that γ(i) 6= i, then γγ ′ (i) = γ(i), and since also γ 2 (i) 6= γ ′ (i), γ ′ γ(i). Similarly, if γ ′ (i) 6= i, then γ ′ γ(i) = γ ′ (i) = γγ ′ (i). Also if γ(i) = i = γ ′ (i), then γγ ′ (i) = γ ′ γ(i). Thus γγ ′ = γ ′ γ, that is, we have proved (1) of the following theorem. Theorem 1.5.1. [Order of a Cycle] 1. Any two disjoint cycles commute. 2. If γ = (i1 i2 . . . ir ) is an r-cycle, then the order of γ is r. 3. If α = (i1 i2 . . . ir1 )(j1 j2 . . . jr2 ) . . . (k1 k2 . . . krs ) is a product of disjoint cycles, then the order of α is the least common multiple of r1 , r2 , . . . , rs . Proof. For (2), clearly, γ r = (1). Let 1 ≤ s < r. Then γ s (i1 ) = is+1 = 6 i1 , so γ s 6= (1). (3) follows from (2) and the fact that |ab| = lcm(|a|, |b|) for all a, b ∈ G such that ab = ba and ∀m ∈ N, am bm = e ⇒ am = bm = e (see Exercises 1.4). It is convenient to extend the definition of cycles and the cycle notation to 1-cycles where we adopt the convention that for any i, (i) is the identity mapping. With this convention, we can see that: Theorem 1.5.2. [Decomposition of a Permutation] Every permutation is a product of disjoint cycles. Moreover, the product is unique up to rearranging its cycles and cyclically permuting the numbers within each cycle. Proof. Let σ ∈ Sn . If σ = (1), we are done. Assume that σ 6= 1. Let G = hσi act on {1, 2, . . . , n naturally as in Examples 1.3.1 (1). Let B1 , B2 , . . . , Br be distinct orbits of {1, 2, . . . , n} under this action. For each j ∈ {1, 2, . . . , r}, we define the cycle µi by ( σ(x), if x ∈ Bi ; µi (x) = x, if x ∈ {1, 2, . . . , n} r Bi . Since Bi , i = 1, 2, . . . , r, are disjoint, µi are disjoint cycles, and clearly, σ = µ1 µ2 . . . µr . Remark. The above two theorems tell us how to find the order of an element in Sn . Next, we shall discuss the cycle structure and the conjugacy class of a permutation. Lemma 1.5.3. If α ∈ Sn is a permutation, then α(i1 i2 . . . ir )α−1 = (α(i1 )α(i2 ) . . . α(ir )). Proof. For x ∈ {1, 2, . . . , n}, α(i1 i2 . . . ir )(x) = = ( ( α(im+1 α(x), mod r ), if x = im mod r ; if x ∈ / {i1 , i2 , . . . , ir } α(im+1 α(x), mod r ), if α(x) = α(im mod r ); if α(x) ∈ / {α(i1 ), α(i2 ), . . . , α(ir )} = (α(i1 )α(i2 ) . . . α(ir ))(α(x)). Hence, α(i1 i2 . . . ir ) = (α(i1 )α(i2 ) . . . α(ir ))α. 25 1.5. The Symmetric Group If σ ∈ Sn is the product of disjoint cycles of lengths r1 , r2 , . . . , rs with r1 ≤ r2 ≤ . . . ≤ rs (including its 1-cycles) then the integers r1 , r2 , . . . , rs are called the cycle structure of σ. A partition of a positive integer n is any nondecreasing sequence of positive integers whose sum is n. For example, 5 has seven partitions, namely, 1 + 1 + 1 + 1 + 1, 1 + 1 + 1 + 2, 1 + 2 + 2, 1 + 1 + 3, 1 + 4, 2 + 3 and 5. The following result shows that the number of conjugacy classes of Sn and the number of partitions of n are coincide. Theorem 1.5.4. Two elements of Sn are conjugate if and only if they have the same cycle structure. The number of conjugacy classes of Sn equals the number of partitions of n. Proof. Assume that σ and τ are conjugate. Then τ = ασα−1 for some τ ∈ Sn . Write σ = (i1 i2 . . . ir1 )(j1 j2 . . . jr2 ) . . . (k1 k2 . . . krs ) as a product of disjoint cycles. Thus, τ = ασα−1 = α(i1 i2 . . . ir1 )α−1 α(j1 j2 . . . jr2 )α−1 α(k1 k2 . . . krs )α−1 = (α(i1 )α(i2 ) . . . α(ir1 ))(α(j1 )α(j2 ) . . . α(jr2 ))(α(k1 )α(k2 ) . . . α(krs )). Hence, σ and τ have the same cycle structure. Conversely, suppose that σ and τ have the same cycle structure written as a product of s disjoint cycles (including 1-cycles) as σ = (a1 a2 . . . ar1 )(ar1 +1 ar1 +2 . . . ar1 +r2 ) . . . (ar1 +r2 +···+rs−1 +1 . . . an−1 an ) and τ = (b1 b2 . . . br1 )(br1 +1 br1 +2 . . . br1 +r2 ) . . . (br1 +r2 +···+rs−1 +1 . . . bn−1 bn ). Define α ∈ Sn by α(ai ) = bi for all i ∈ {1, 2, . . . , n}. Then ασα−1 = τ . Example 1.5.1. The number of conjugacy classes of S5 is 7 and |{α(12)(345)α−1 : α ∈ S5 }| = 52 (3 − 1)! = 20. Example 1.5.2. The Klein group V4 = {(1), (12)(34), (13)(24), (14)(23)} is normal in S4 because V4 contains of all products of disjoint 2-cycles and so ∀α ∈ S4 , αV4 α−1 = V4 by Theorem 1.5.4. Moreover, since the group {(1), (12)(34)} is of index two in V4 , it is normal in V4 by Theorem 1.3.3. However, {(1), (12)(34)} is not normal in S4 . Thus, normality of subgroups is not transitive. Corollary 1.5.5. [Center of Sn ] If n ≥ 3, then the center of Sn is trivial, i.e., Z(Sn ) = {(1)}. Proof. We wish to prove that ∀α ∈ Sn [∀β ∈ Sn , βαβ −1 = α ⇒ α = (1)]. By Theorem 1.5.4, ∀α ∈ Sn [α 6= (1) ⇒ |{βαβ −1 : β ∈ Sn }| > 1]. By Corollary 1.3.8,∀α ∈ Sn [|{βαβ −1 : β ∈ Sn }| > 1⇒α∈ / Z(Sn )]. Hence, let α ∈ Z(Sn ). Then |{βαβ −1 : α ∈ Sn }| = 1, so α = (1). To define an important subgroup of Sn , namely the alternating group, we shall need some results on 2-cycles. A cycle of the form (ab), where a 6= b, is called a transposition. It is easy to verify that (i1 i2 . . . ir ) = (i1 ir ) . . . (i1 i3 )(i1 i2 ), a product of r − 1 transpositions. It follows that any α ∈ Sn is a product of transpositions. Also, a transposition (ab) has order two in Sn . Theorem 1.5.6. 1. (1) is always a product of even number of transpositions. 2. If α ∈ Sn is written as a product of transpositions, then either the number of transpositions in any product is always odd or always even. 26 1. Groups Proof. (1) Assume that we have two transposition decompositions (1) = (x1 y1 )(x2 y2 ) . . . (xk yk ) = (1x1 )(1y1 )(1x1 )(1x2 )(1y2 )(1x2 ) . . . (1xk )(1yk )(1xk ) with xi < yi for all i ∈ {1, 2, . . . , k}. Consider any (1u), u > 1, in the right hand side. Since the opposite side is (1), (1u) must occur twice (or even number of times) in the right hand side. Note that (1 7→ u and u 7→ 1) will give u 7→ u. Thus each transposition in the right hand side occurs even numbers of times, which implies that the right hand side should have even number of transpositions. Hence, k is even. (2) Assume α = (x1 y1 )(x2 y2 ) . . . (xk yk ) = (w1 z1 )(w2 z2 ) . . . (wl zl ) for some xi 6= yi , wj 6= zj and k, l ∈ N. Since |(wi zi )| = 2 for all i, (x1 y1 )(x2 y2 ) . . . (xk yk )(wl zl )−1 (wl−1 zl−1 )−1 . . . (w1 z1 )−1 = (1) (x1 y1 )(x2 y2 ) . . . (xk yk )(wl zl )(wl−1 zl−1 ) . . . (w1 z1 ) = (1), so k + l is even. Hence, k and l have the same parity. The previous theorem leads to the definition of parity of a permutation. We call the permutation α even or odd according as α factors as a product of an even or an odd number of transpositions. Remarks. Let α, β ∈ Sn . 1. αβ is even ⇔ α and β have the same parity. 2. Since αα−1 = (1) which is even, α and α−1 have the same parity. Theorem 1.5.7. Let n > 1. The set An of all even permutations forms a normal subgroup of Sn of index two. It is called the alternating group of degree n and |An | = n!/2. Proof. By Theorem 1.5.6, (1) is even. It is clear that the product of even permutations is even. Since a transposition has order two, the inverse of an even permutation is even. Hence, An is a subgroup of Sn . Since n > 1, let (ab) be a transposition in Sn . Clearly, (ab) is an odd permutation. We will show that Sn = An ∪ (ab)An . Let α ∈ Sn . If α is even, then α ∈ An . On the other hand, assume that α is odd. Then (ab)α is even, so (ab)α ∈ An , i.e., α ∈ (ab)An . Thus, [Sn : An ] = 2. In addition, since α and α−1 have the same parity, αAn α−1 ⊆ An . Hence, An is normal in Sn . The above proof also shows that if n > 1, then Sn = An ∪ (ab)An and the number of even permutations and odd permutations are the same. Corollary 1.5.8. Let a group G act on a finite set X, and assume that some element h ∈ G induces an odd permutation on X. Then there exists a normal subgroup N of G with [G : N ] = 2 and h∈ / N. Proof. Consider the diagram G θ / S(X) π / S(X)/A(X) , where A(X) is the alternating group of even permutations on X, θ : g → φg and π is the canonical map. Since φh is an odd permutation, π ◦ θ is onto. Choose N = ker π ◦ θ. Then N ⊳ G and G/N ∼ = S(X)/A(X). Thus, [G : N ] = [S(X) : A(X)] = 2. Since (π ◦ θ)(h) = φh A(X) 6= A(X), we have h ∈ / N. 27 1.5. The Symmetric Group Finally, we talk about the simplicity of An . A group is simple if it has no nontrivial normal subgroup. That is, all normal subgroups of G are {e} and G. For example, Zp is simple for all primes p. Corollary 1.5.9. Let |G| = 2m, where m is odd. Then G has a normal subgroup of order m. In particular, if m > 1, then G is not simple. Proof. Since |G| is even, let g be an element of order two in G. Let G act on G by left multiplication θ and consider G → S(G). Since the action is faithful, θ is 1-1, so |θ(g)| = 2. Thus, θ(g) = φg is an odd permutation. By the previous corollary, there exists a normal subgroup N of G such that [G : N ] = 2. Hence, G is not simple. Example 1.5.3. Since the Klein group V4 = {(1), (12)(34), (13)(24), (14)(23)} is normal in A4 , it follows that A4 is not simple. In general, we have the next theorem. Theorem 1.5.10. An is simple for all n 6= 4. Proof. Clearly, A2 and A3 are simple. For n ≥ 5, we give a step-by-step guideline in Project 5. Corollary 1.5.11. If n 6= 4, then the only normal subgroups of Sn are {(1)}, An and Sn . Exercises 1.5. 1. Prove that A4 is the only subgroup of S4 of order 12. 2. Prove that A4 has no subgroup of order six. 3. Determine all normal subgroups of S4 . (Hint. Use conjugacy classes.) 4. (a) Find the largest positive integer n such that S10 has a permutation of order n. (b) The exponent of a finite group G is the smallest positive integer n such that g n = 1 for all g ∈ G. Find the exponent of S30 , the symmetric group on 30 letters. 5. Show that if H is any subgroup of Sn , n ≥ 2, then either all permutations in H are even or exactly half are even. 6. Let G be a group of order 360 having a maximal subgroup isomorphic to A5 . Prove that G ∼ = A6 . Project 5 (Simplicity of An ). Prove that An is simple for n ≥ 5, following the steps and hints given. (a) Show An contains every 3-cycle if n ≥ 3. (b) Show An is generated by the 3-cycles for n ≥ 3. [Hint. Note that (a, c)(a, b) = (a, b, c) and (a, b)(c, d) = (a, c, b)(a, c, d).] (c) Let r and s be fixed elements of {1, 2, . . . , n} for n ≥ 3. Show that An is generated by the n “special” 3-cycles of the form (r, s, i) for 1 ≤ i ≤ n. [Hint. Show every 3-cycle is the product of “special” 3-cycles by computing (r, s, i)2 , (r, s, j)(r, s, i)2 , (r, s, j)2 (r, s, i) and (r, s, i)2 (r, s, k)(r, s, j)2 (r, s, i). Observe that these products give all possible types of 3-cycles.] (d) Let N be a normal subgroup of An for n ≥ 3. Show that if N contains a 3-cycle, then N = An . [Hint. Show that (r, s, i) ∈ N implies that (r, s, j) ∈ N for j = 1, 2, . . . , n by computing ((r, s)(i, j))(r, s, i)2 ((r, s)(i, j))−1 .] (e) Let N be a nontrivial normal subgroup of An for n ≥ 5. Show that one of the following cases must hold, and conclude in each case that N = An . Case 1. N contains a 3-cycle. Case 2. N contains a product of disjoint cycles, at least one of which has length greater than 3. [Hint. Suppose N contains the disjoint product σ = µ(a1 , a2 , . . . , ar ). Show that σ −1 (a1 , a2 , a3 )σ(a1 , a2 , a3 )−1 28 1. Groups is in N , and compute it.] Case 3. N contains a disjoint product of the form σ = µ(a4 , a5 , a6 )(a1 , a2 , a3 ). [Hint. Show that σ −1 (a1 , a2 , a4 )σ(a1 , a2 , a4 )−1 is in N , and compute it.] Case 4. N contains a disjoint product of the form σ = µ(a1 , a2 , a3 ) where µ is a product of disjoint 2-cycles. [Hint. Show σ 2 ∈ N and compute it.] Case 5. N contains a disjoint product σ of the form σ = µ(a3 , a4 )(a1 , a2 ), where µ is a product of an even number of disjoint 2−cycles. [Hint. Show that σ −1 (a1 , a2 , a3 )σ(a1 , a2 , a3 )−1 is in N , and compute it to deduce that α = (a2 , a4 )(a1 , a3 ) is in N . Using n ≥ 5 for the first time, find i ∈ {1, 2, . . . , n}, where i 6= a1 , a2 , a3 , a4 . Let β = (a1 , a3 , i). Show that β −1 αβα ∈ N , and compute it.] Project 6 (Wilson). Let p be a prime. Taking G = Sp in the proof of Cauchy’s theorem (Theorem 1.3.12), we see that the set {σ ∈ Sp : σ p = (1)} is of cardinality a multiple of p. Count the number of elements in this set and deduce that (p − 1)! ≡ −1 (mod p). This provides another proof of Wilson’s theorem. 1.6 Sylow Theorems We know the order of a subgroup of a finite group G must divide |G|. If |G| is cyclic (even only abelian), then there exist subgroups of every order dividing |G|. A natural question is: If k divides |G| is there always a subgroup of G of order k? A little experimenting shows that this is not so. For example, the alternating group A4 , whose order is 12, contain no subgroup of order six. Moreover, An for n ≥ 5 is simple, that is, contains no normal subgroup 6= 1, An . Since any subgroup of index two is normal, it follows that An , n ≥ 5, contains no subgroup of order n!/4. 1.6.1 Sylow p-subgroups The main positive result of the type we are discussing was discovered by Sylow /see lov/. Its proof provides us another application of action of a group on a set. Unless specified otherwise, p denotes a prime. A group G is said to be a p-group if |a| is a power of p for all a ∈ G. Example 1.6.1. The group Zpn is a p-group. If X is a set, then (P (X), △) is a 2-group. Since we mainly study finite groups, the following corollary will be useful. Lagrange theorem and Cauchy theorem imply each direction, respectively. Corollary 1.6.1. Let G be a finite group. Then G is a p-group ⇔ |G| is a power of p. Remark. Let P and Q be subgroups of G. If P is a p-group and Q is a q-group, where p and q are distinct primes, then P ∩ Q = {e}. Theorem 1.6.2. Let G be a finite p-group and |G| > 1. Then the following statements hold. 1. |Z(G)| > 1. 2. If |G| = p2 , then G is abelian. Proof. By Corollary 1.6.1, |G| = pl for some l ∈ N. Recall from Corollary 1.3.8 that |G| = |Z(G)| + s X i=1 |{gxi g −1 : g ∈ G}| = |Z(G)| + s X i=1 [G : CG (xi )], 29 1.6. Sylow Theorems where x1 , . . . , xs represent the conjugacy classes of G which contains more than one element. Since [G : CG (xi )] = |G|/|CG (xi )| > 1 for all i and |G| = pl , p divides |{gxi g −1 : g ∈ G}| for all i ∈ {1, 2, . . . , s}. Hence, p | |Z(G)|, so |Z(G)| > 1. This proves (1). For the second part, assume that |G| = p2 . We know that Z(G) is a normal subgroup of G and |Z(G)| > 1. By Lagrange Theorem, |Z(G)| = p or |Z(G)| = p2 . If |Z(G)| = p2 , we have Z(G) = G and so G is abelian. Suppose that |Z(G)| = p. Then G/Z(G) is of order p and so a cyclic group. This implies that G is abelian. Because all subgroups of a p-group have p-power index, the length of an orbit under an action by a p-group is a multiple of p unless the point is a fixed point, when its orbit has length one. This leads to an important congruence modulo p when a p-group is acting. Lemma 1.6.3. [Fixed Point Congruence] Let G be a finite p-group. If G acts on a finite set X and X0 = {x ∈ X : gx = x for all g ∈ G}, then |X0 | ≡ |X| mod p. Here, X0 is called the set of fixed points. Proof. We observe first that X0 = {x ∈ X : |G · x| = 1}. Let x1 , . . . , xs represent the orbits of X which contains more than one element. Then |X| = |X0 | + s X i=1 |G · xi |. By Orbit-Stabilizer Theorem, for each i ∈ {1, 2, . . . , s}, 1 < |G · xi | = [G : StabG xi ] which is divisible by p. Hence, |X0 | ≡ |X| mod p as desired. Lemma 1.6.4. Let G be a finite group and H, P ≤ G. If H is a p-group, then 1. |{xP : x ∈ G and H ⊆ xP x−1 }| ≡ [G : P ] mod p, 2. [NG (H) : H] ≡ [G : H] mod p, and 3. if p | [G : H], then p | [NG (H) : H] and NG (H) 6= H. Proof. Let X = {xP : x ∈ G} and let H act on X by h · xP = hxP for all x ∈ G and h ∈ H. Clearly, |X| = [G : P ] and X0 = {xP : x ∈ G and hxP = xP for all h ∈ H} = {xP : x ∈ G and x−1 hx ∈ P for all h ∈ H} = {xP : x ∈ G and x−1 Hx ⊆ P } = {xP : x ∈ G and H ⊆ xP x−1 }, so |{xP : x ∈ G and H ⊆ xP x−1 }| ≡ [G : P ] mod p by Lemma 1.6.3. Furthermore, if P = H, we have X0 = {xH : x ∈ G and H ⊆ xHx−1 }. Since ∀x ∈ G, |xHx−1 | = |H| and H is finite, we have |X0 | = |{xH : x ∈ G and x−1 Hx = H}| = |{xH : x ∈ NG (H)}| = [NG (H) : H], so [NG (H) : H] ≡ [G : H] mod p . The final result clearly follows from (2). We now discuss three theorems due to Sylow. The first theorem shows the existance of a maximal p-subgroup of a finite group G. Theorem 1.6.5. [First Sylow Theorem] Let G be a group of order pn m where n ≥ 1 and p does not divide m. Then the following statements hold. 1. G contains a subgroup of order pi for all 1 ≤ i ≤ n. 2. For each i, where 1 ≤ i < n, every subgroup H of G of order pi is a normal subgroup of a subgroup of order pi+1 . 30 1. Groups Proof. Since p divides |G|, by Cauchy theorem, G has a subgroup H1 of order p. Assume that k ∈ {1, 2, . . . , n − 1} and G has a subgroup Hk of order pk . Then the index [G : Hk ] = pn−k m and n − k ≥ 1. By Lemma 1.6.4, p divides [NG (Hk ) : Hk ] = |NG (Hk )/Hk |. Again, by Cauchy theorem, NG (Hk )/Hk has a subgroup H of order p. By the Third Isomorphism Theorem, H = Hk+1 /Hk for some subgroup Hk+1 of NG (Hk ) containing Hk . Moreover, Hk ⊳ Hk+1 and |Hk+1 | = |H||Hk | = p pk = pk+1 . Hence, there are subgroups H1 , H2 , . . . , Hn of G such that |Hi | = pi for i = 1, 2, . . . , n and H1 ⊳ H2 ⊳ · · · ⊳ Hn . A maximal p-subgroup of a group G is called a Sylow p-subgroup of G. By Zorn’s lemma, we have the following statements. 1. A Sylow p-subgroup of a group G always exists and it may be trivial. 2. Every p-subgroup of a group G is contained in a Sylow p-subgroup of G. By Corollary 1.6.1 and Theorem 1.6.5, if G is a finite group and p is a prime such that p divides |G|, then G has a Sylow p-subgroup of order pn k|G|. (Here, pn k|G| means n is the highest power of p dividing |G|.) That is, [G : P ] is not divisible by p. Moreover, we have: Corollary 1.6.6. Let G be a group of order pn m where n ≥ 1 and p does not divide m. 1. G has a Sylow p-subgroup of order pn . 2. For H < G, H is a Sylow p-subgroup of G ⇔ |H| = pn . 3. Every conjugate of a Sylow p-subgroup of G is a Sylow p-subgroup of G. 4. If P is the only one Sylow p-subgroup of G, then P is normal in G. Proof. (1) and (2) follow from the definition and the above discussion. Since a conjugate of a subgroup of G is of the same order as the subgroup, (2) implies (3). Finally, (4) follows from (3). The second and third Sylow theorems determine all Sylow p-subgroups and possible numbers of Sylow p-subgroups, respectively. Also, they give the converse of the above results. Theorem 1.6.7. [Second Sylow Theorem] Let G be a finite group. 1. If P is a Sylow p-subgroup of G and H is a p-subgroup of G, then H ⊆ xP x−1 for some x ∈ G. 2. Any two Sylow p-subgroups of G are conjugate. Proof. By Lemma 1.6.4 |{xP : x ∈ G and H ⊆ xP x−1 }| ≡ [G : P ] mod p. Since P is a Sylow p-subgroup of G, p ∤ [G : P ], so {xP : x ∈ G and H ⊆ xP x−1 } = 6 ∅. Thus, there exists an x ∈ G −1 such that H ⊆ xP x . Next, we let P1 and P2 be Sylow p-subgroups of G. Then there exists an x ∈ G such that P1 ⊆ xP2 x−1 . But |P1 | = |P2 | = |xP2 x−1 | and G is finite, P1 = xP2 x−1 . Corollary 1.6.8. Let G be a finite group and let P be a Sylow p-subgroup of G. 1. {xP x−1 : x ∈ G} is the set of all Sylow p-subgroups of G. 2. The number of Sylow p-subgroups of G is [G : NG (P )] and it divides [G : P ] and |G|. 3. P is normal in G ⇔ P is the only one Sylow p-subgroup of G. For a finite group G and a prime p divides |G|, we write np (G) for the number of Sylow p-subgroups of G. Theorem 1.6.9. [Third Sylow Theorem] If G is a finite group and a prime p divides |G|, then np (G) ≡ 1 mod p. 31 1.6. Sylow Theorems Proof. Let P be a Sylow p-subgroup of G. Then the set X = {xP x−1 : x ∈ G} consists of all Sylow p-subgroups of G. Let P act on X by conjugation, namely, (g, xP x−1 ) 7→ gxP x−1 g −1 for all g ∈ P and x ∈ G. Since gP g −1 = P for all g ∈ P , P ∈ X0 . Let Q ∈ X0 . Then gQg −1 = Q for all g ∈ P , so P ⊆ NG (Q). Since P and Q are Sylow p-subgroups of NG (Q) and Q is normal in NG (Q), P = Q by the uniqueness of normal Sylow p-subgroup. This proves X0 = {P }. By Lemma 1.6.3, we have np (G) = |X| ≡ |X0 | = 1 mod p as desired. 1.6.2 Applications of Sylow Theorems Here, we present some applications of Sylow theorems on a finite group. The proofs use basic properties of subgroups, quotient groups, cyclic groups and symmetric groups studied previously. We shall see many techniques in group theory in this subsection. Theorem 1.6.10. Let G be a finite group. If P is a Sylow p-subgroup of G, then NG (NG (P )) = NG (P ). Proof. Since P ⊳ NG (P ), P is the only Sylow p-subgroup of NG (P ). Let x ∈ NG (NG (P )). Then xNG (P )x−1 = NG (P ). Since P ⊆ NG (P ), xP x−1 ⊆ NG (P ). Thus, xP x−1 = P since xP x−1 is a Sylow p-subgroup of G. Hence, x ∈ NG (P ). Theorem 1.6.11. [Group of Order pq] Let G be a group of order pq where p and q are primes and p < q. Then G is a cyclic group, or G has q Sylow p-subgroups and p | (q − 1). Proof. Since the number of Sylow p-subgroups divides |G| = pq, it is 1, p, q or pq. But this number is ≡ 1 mod p, so it is 1 or q. If G has q Sylow p-subgroups, then we are done. Assume that G has only one Sylow p-subgroup, say P . Then P is normal in G. Consider the number of Sylow q-subgroups of G. It is again 1, p, q or pq, and ≡ 1 mod q, so the only possibility is 1 since p < q. Thus, G also has a unique Sylow q-subgroup, say Q, and so Q is normal in G. Since the orders of P and Q are prime, both P and Q are cyclic. Let a and b be generators of P and Q, respectively. Note that aba−1 b−1 ∈ P ∩ Q = {e}. Thus, ab = ba, so |ab| = pq = |G|. Hence, G = habi. Remark. Theorem 1.6.11 demonstrates the power of the Sylow theorems in classifying the finite groups whose orders have small numbers of prime factor. Results along this lines of this theorem exist for groups of order p2 q, p2 q 2 , p3 and p4 , where p < q are primes. Example 1.6.2. There can be no simple groups of order 200 and of order 280. Proof. Let H be a group of order 200. Let P be a Sylow 5-subgroup of H. Then n5 (H) divides [H : P ] = 8 and n5 (H) ≡ 1 (mod 5), so n5 (H) = 1. Hence, P is normal in H. Next, let G be a group of order 280. By Corollary 1.6.8 and Theorem 1.6.9, we have n2 (G) = 1, 5, 7 or 35, n5 (G) = 1 or 56 and n7 (G) = 1 or 8. If n5 (G) = 1 or n7 (G) = 1, we are done. Assume that n5 (G) = 56 and n7 (G) = 8. Then we have 56 · 4 = 224 elements of order 5, and 8 · 6 = 48 elements of order 7. Hence, G has a unique Sylow 2-subgroup. Example 1.6.3. Let G be a group of order 30. Then 1. Either a Sylow 3-subgroup or a Sylow 5-subgroup is normal in G. 2. G has a normal subgroup of order 15. 3. Both a Sylow 3-subgroup and a Sylow 5-subgroup are normal in G. 32 1. Groups Proof. Assume that neither a Sylow 3-subgroup nor a Sylow 5-subgroup are normal in G. By Corollary 1.6.8, n3 (G) and n5 (G) are more than one and are factors of |G|. By Third Sylow Theorem, n3 (G) ≥ 10 and n5 (G) ≥ 6, so G contains at least 20 elements of order three and at least 24 elements of order five. This exceeds the number of elements of G, a contradiction. Thus, we have (1). Now, let P3 and P5 be a Sylow 3-subgroup and a Sylow 5-subgroup of G, respectively. By (1), we see that P3 or P5 is normal in G, so P3 P5 is a subgroup of G. Since P3 ∩ P5 = {e}, |P3 P5 | = 15, so the index [G : P3 P5 ] is two. Hence, P3 P5 is normal in G. This proves (2). Finally, we assume that P3 is normal while P5 is not. Thus, G has two elements of order three at least 24 elements of order five. By Theorem 1.6.11, P3 P5 is cyclic, so G has φ(15) = 8 elements of order 15. Hence, G contains more than 30 elements, a contradiction. On the other hand, we assume that P5 is normal while P3 is not. Thus, G has four elements of order five at least 20 elements of order three. Again, G also contains 8 elements of order 15. This leads to a contradiction, so P3 and P5 are normal in G as desired. Example 1.6.4. Every group G of order 12 that is not isomorphic with A4 contains an element of order 6. Proof. If A is a Sylow 3-subgroup, then A = hai and |a| = 3. Let G act on {A, x2 A, x3 A, x4 A} by (g, xA) 7→ gxA. This action induces a homomorphism θ : G → S4 whose kernel K is a subgroup of A. Then K = {e} or K = A. If K = {e}, then G is isomorphic to a subgroup of S4 of order 12, so G ∼ = A4 which is excluded by hypothesis. Thus, A = K is normal in G which implies that A is a unique Sylow 3-subgroup of G. Hence, a and a2 are only two elements of order 3 in G. Since [G : CG (a)] is the number of conjugates of a which is 1 or 2, |CG (a)| = 12 or 6, so there is a b ∈ CG (a) of order two. Since ab = ba, |ab| = 6. Example 1.6.5. Recall that V4 = {(1), (12)(34), (13)(24), (14)(23)} is a normal subgroup of A4 . Since |A4 | = 12 = 22 · 3, V4 is the unique Sylow 2-subgroup of A4 . Moreover V4 has three subgroups of order two, namely h(12)(34)i, h(13)(24)i and h(14)(23)i. Next, we analyze the Sylow 3-subgroups of A4 . They are cyclic subgroups of order three generated by a 3-cycle. Note that there are eight 3-cycles in A4 , so we have four subgroups of order three, which are h(123)i, h(124)i, h(134)i and h(234)i. By Exercises 1.5, A4 has no subgroup of order six. Hence, the diagram below shows all subgroups of A4 . ❯ ❑❑❯❳❯❳❯❳❯❳❯❳❳❳❳ ❤❤❤ A4 ❳ ❑❑ ❯❯❯ ❳❳❳❳❳ ❤❤❤❤ ❤ ❤ ❤ ❑❑ ❯ ❤❤❤ ❤ ❑❑ ❯❯❯❯❯❳❯❳❳❳❳❳❳❳❳❳ ❤ ❤ ❑❑ ❯❯❯❯ ❳❳❳❳❳ ❤❤❤ ❤ ❤ ❤ ❳❳❳❳ ❯ ❤ ❤❤❤❤ h(123)i h(134)i V h(124)i h(234)i 4 ◆◆ ◆◆◆ ♠ t ♣♣♣ ✝ ♠ t ♣ ♠ ◆ ♣ ✝ t ◆◆◆ ♣ ♠♠♠ tt ✝✝ ◆◆◆ ♣♣♣ ♠♠♠ tt ✝ ♣ ♠ t ♠ ♣ ✝ t ♠ ♣ ✝ ♠♠ tt tt ♠♠♠♠♠ ✝✝ h(14)(23)i h(12)(34)i ❨❨❨❨ h(13)(24)i ❱ t ✝ t ♠ ❨❨❨❨❨❨ ▲▲▲ ✝ ♠♠ tt ❨❨❨❨❨❨ ❱❱❱❱❱❱❱❱ ✝✝ tttt ♠♠♠♠♠ ❨❨❨❨❨❨ ❱❱❱❱ ▲▲▲▲ ✝ ♠ ❨❨❨❨❨❨ ❱❱❱❱ ▲▲ ✝ tt ♠♠ ❨❨❨❨❨❨❱❱❱❱ ▲ ✝✝t♠t♠♠♠ ❨❨❨❨❱ {(1)} We shall see more applications of Sylow theorems in Section 3.3. It turns out that any finite nilpotent group is the direct product of its Sylow p-subgroups. Exercises 1.6. 1. If G is a finite p-group where p is a prime, N is normal in G and N 6= {e}, prove that N ∩ Z(G) 6= {e}. 2. Prove that if |G| = pn with p > n, p is a prime, and H is a subgroup of G of order p, then H ⊳ G. 3. Let p be the smallest prime dividing the order of a finite group G. Show that any subgroup H of G of index p is normal. 33 1.6. Sylow Theorems 4. Let G be a group of order pn where p is a prime and n ∈ N. Prove that there exist normal subgroups N1 , . . . , Nn of G such that N1 < N2 < · · · < Nn with |Ni | = pi for all i ∈ {1, 2, . . . , n}. 5. Let G be a group, M ⊳ G and N ⊳ G. Prove the following statements. (a) If M ∩ N = {e}, then xy = yx for all x ∈ M and y ∈ N . (b) If M and N are finite cyclic subgroups of G and gcd(|M |, |N |) = 1, then M N is a cyclic subgroup of G of order |M ||N |. 6. Let P be a Sylow p-subgroup of a finite group G and N a normal subgroup of G. Show that: (a) P ∩ N is a Sylow p-subgroup of N , (b) P N/N is a Sylow p-subgroup of G/N . 7. Show that there are no simple groups of order 148 or of order 56. 8. How many elements of order 7 are there in a simple group of order 168? 9. Let G be a group of order 153. Prove that G is abelian. 10. Let G be a group of order 231. Show that n11 (G) = 1 and the Sylow 11-subgroup of G is contained in Z(G). 11. Show that there is no non-abelian finite simple group of order less than 60. (Hint. We may focus on groups of the following orders: 24, 30, 40, 48, 54 and 56.) 12. Let G be a group of order 385. Show that a Sylow 11-subgroup of G is normal and a Sylow 7-subgroup of G is contained in Z(G). 13. Let p be a prime and P a Sylow p-subgroup of a finite group G. Suppose that, for all g ∈ G, if P 6= gP g −1 , then P ∩ gP g −1 = {e}. Show that np (G) ≡ 1 mod |P |. 14. Let G be a group of order 2013. Prove that G has a proper normal subgroup N such that G/N is cyclic. 15. (a) Let G be a finite group and N a normal subgroup of G. If N contains a Sylow p-subgroup of G, prove that the number of Sylow p-subgroups of N is the same as that of G (i.e., np (N ) = np (G)). (b) Show that if G is a group of order 130, then G has a normal subgroup of order 5. 16. Let G be a finite group acting transitively on a finite set X. Let x ∈ X and Gx the stabilizer of x. Let P be a Sylow p-subgroup of Gx . Show that the subgroup NG (P ) = {z ∈ G : zP z −1 = P } of G acts transitively on Y = {y ∈ X : hy = y for all h ∈ P }. Project 7 (Simple groups of small order). We have learned from the above exercise that the smallest nonabelian simple group is of order 60 by using Sylow’s theorems to eliminate the groups of smaller order. Write a computer program that uses Sylow’s theorems to eliminate all orders between 1 and say 1,000 (or more) for which group that cannot be simple. For any order that could have a simple group G, list np (G) for all primes p dividing the order. Project 8 (Lucas’ congruence). Let p be a prime and let n ≥ m be non-negative integers. Write n = pn′ +a0 and m = pm′ + b0 where 0 ≤ a0 , b0 ≤ p − 1. Decompose {1, 2, . . . , n} into a union of p blocks on n′ consecutive integers, from 1 to pn′ , followed by a final block of length a0 . That is, let Ai = {in′ + 1, in′ + 2, . . . , (i + 1)n′ } for 0 ≤ i ≤ p − 1, so {1, 2, . . . , n} = A0 ∪ A1 ∪ · · · ∪ Ap−1 ∪ {pn′ + 1, pn′ + 2, . . . , pn′ + a0 }. For 1 ≤ t ≤ n′ , let σt be the p-cycle σt = (t, n′ + t, 2n′ + t, . . . , (p − 1)n′ + t). This cycle cyclically permutes the numbers in A0 , A1 , . . . , Ap−1 that are ≡ t (mod n′ ). The σt ’s for different t are disjoint, so they commute. Set σ = σ1 σ2 . . . σn . Then σ has order p as a permutation of {1, 2, . .. , n} n . Let (fixing all numbers above pn′ ). Let X be the set of m-element subsets of {1, 2, . . . , n}. Then |X| = m the group hσi act on X. n′ Deduce, by Lemma 1.6.3, that (a) Show that the number of fixed points of this action is ab00 m ′ . n′ a0 n m ≡ b0 m′ (mod p). 34 1. Groups (b) Prove Lucas’ congruence: if n = a0 + a1 p + a2 p2 + · · · + ak pk and m = b0 + b1 p + b2 p2 + · · · + bk pk with 0 ≤ ai , bi ≤ p − 1, then n a0 a1 ak ≡ ··· (mod p). b0 b1 bk m 1.7 Finite Abelian Groups The study of finite non-abelian groups is complicated as we have learned that the Sylow theorems give us some important information about them. This section gives us complete information about all finite abelian groups. We start with formal definitions of the direct product of groups. Let A and B be groups. The direct product of A and B is defined as: 1. a set A × B = {(a, b) : a ∈ A and b ∈ B} is the Cartesian product of A and B, 2. multiplication is coordinatewise, namely, (a, b)(c, d) Q = (ac, bd). More generally, if {Ai : i ∈ I} is a family of groups, then i∈I Ai is a group with coordinatewise multiplication. It is called the direct product of the groups Ai . The subgroup n o Y Yw Ai = (ai ) ∈ Ai : ai = e for all but finitely many i i∈I i∈I Q it is normal of Qi∈I Ai is called the external weak direct product of the groups AP i . Note thatQ A for A . In case A are all additive abelian groups, we may write in i i∈I Ai and i∈I i Pw i∈I i Qw i∈I Ai for i∈I Ai . Let G be a group. It is easy to show that: 1. If N1 , . . . , Nm are normal subgroups of G, then N1 N2 · · · Nm = hN1 ∪ N2 ∪ · · · ∪ Nm i. 2. If {Ni : i ∈ I} is a family of normal subgroups of G, then D[ E [ N i1 · · · N im . Ni = i∈I i1 ,...,im ∈I, m∈N Theorem 1.7.1. Let {Ni : i ∈ I} be a familyD of normal subgroups of G such that E S S and 2. ∀k ∈ I, Nk ∩ 1. G = i∈Ir{k} Ni = {e}. i∈I Ni Q w ∼ Then G = Ni . i∈I The group G satisfies conditions Theorem 1.7.1 is called the internal weak direct product Qof w of {Ni : i ∈ I} and we write G = i∈I Ni . If L G is additive abelian, then G is called the internal direct sum of {Ni : i ∈ I} and we write G = i∈I Ni . Corollary 1.7.2. Let N1 , N2 , . . . , Nm be normal subgroups of G. If G = N1 . . . Nm and Nk ∩ (N1 . . . Nk−1 Nk+1 . . . Nm ) = {e} for all k ∈ {1, . . . , m}, then G ∼ = N1 × · · · × Nm . Proof of Theorem 1.7.1. From (2), for each i, j ∈ I with i 6= j, we have Ni ∩ Nj = {e}, this implies that xy = yx for Q all x ∈ Ni and y ∈ Nj because Ni and Nj are normal in G. Define ϕ : w i∈I Ni → G by Y ϕ({ai }) = ai i∈I which is a finite product since ai = e for all but a finite number of i ∈ I and it is a well defined homomorphism by the previous observation. To show that this map is onto, let x ∈ G. Since 35 1.7. Finite Abelian Groups G is generated by S i∈I Ni , G = S i1 ,...,im ∈I, Ni1 · · · Nim , so there are distinct k1 , . . . , kl ∈ I m∈N Q x = ak1 . . . akl . Let {xi }Qin w i∈I Ni be defined by xi = and a1 ∈ Nk1 , . . . , al ∈ Nkl such that ai if . . . a x = a = x as required. i ∈ {k1 , . . . , kl } and xi = e for other i. Then ϕ({x }) = kl i∈I i Qw i Qk1 Finally, we show that ϕ is injective, let {a } ∈ a = e. Then for each N be such that i i i i∈I i∈I S Q N i = {e}. This implies that a = e for all k ∈ I, and is in N ∩ h = k ∈ I, a−1 k k i∈Ir{k} i i∈Ir{k} k hence ϕ is an isomorphism. The proof of injectivitity above also implies the following theorems. Theorem 1.7.3. Let N1 , N2 , . . . , Nm be normal subgroups of G. Then the following statements are equivalent. (i) G is the internal weak direct product of N1 , . . . , Nm . (ii) ∀x ∈ G, ∃!a1 ∈ N1 , . . . , am ∈ Nm , x = a1 . . . am . Theorem 1.7.4. Let {Ni : i ∈ I} be a family of normal subgroups of a group G. Then the following statements are equivalent. (i) G is the internal weak direct product of {Ni : i ∈ I}. (ii) ∀x ∈ G r {e}, ∃!i1 , . . . , im ∈ I, ∃!ai1 ∈ Ni1 r {e}, . . . , aim ∈ Nim r {e}, x = ai1 . . . aim . Corollary 1.7.5. [Internal Direct Product] Let G be a group. Suppose that A and B are normal subgroups of G such that 1. A ∩ B = {e} 2. AB = G and 3. ∀a ∈ A, b ∈ B, ab = ba. Then G ∼ = A × B. In this case, we say that G is the internal direct product of A and B. An application of the first isomorphism theorem with the natural map gives the next results. Theorem i ∈ I} be a family of groups, andQfor i ∈ I, let Ni be normal Qw in Gi . Q Q Q 1.7.6. Let {Gi : Q ∼ (G /N ). Similarly, N G / and Ni is normalQ in i∈I G Then i∈IQ = i i i∈I i∈I Ni is i∈I i i∈I Qi w Qwi w w ∼ normal in i∈I Gi and i∈I Gi / i∈I Ni = i∈I (Gi /Ni ). Corollary 1.7.7. Let G1 , . . . , Gm be abelian groups, and for 1 ≤ j ≤ m, let Hj be a subgroup of Gj . Then (G1 ⊕ · · · ⊕ Gn )/(H1 ⊕ · · · ⊕ Hn ) ∼ = (G1 /H1 ) ⊕ · · · ⊕ (Gn /Hn ). Next, we study the structure of a finite abelian group. Results on elements of finite order are presented in the next theorem and we recall the Chinese Remainder Theorem in a group theoretic language. Their proof are routine and left as exercises. Theorem 1.7.8. Let A be an abelian group and n ∈ N. Then the following statements hold. 1. The mapping ϕn : A → A defined by ϕn (a) = an is a group homomorphism. 2. An = {an : a ∈ A} = im ϕn is a subgroup of A. 3. A(n) = {a ∈ A : an = e} = ker ϕn is a [ subgroup of A. A(n) is a subgroup of A. It is called the torsion 4. τ (A) = {a ∈ A : ∃k ∈ N, ak = e} = n∈N subgroup of A. Theorem 1.7.9. [Chinese Remainder Theorem] Suppose m1 , . . . , mk are pairwise relatively prime (i.e., if i 6= j, then gcd(mi , mj ) = 1), and let n1 , . . . , nk be any integers. Then there exists a unique integer v modulo m = m1 . . . mk such that v ≡ ni for all i = 1, . . . , k. mod mi 36 1. Groups Remark. The Chinese remainder theorem may be restated as: if m1 , . . . , mk are pairwise relatively primes and m = m1 . . . mk , then Zm ∼ = Z m1 ⊕ · · · ⊕ Z mk . Corollary 1.7.10. If m = pn1 1 . . . pnk k where n1 , . . . , nk ∈ N and p1 , . . . , pk are distinct primes, then Zm ∼ = Zpn1 1 ⊕ · · · ⊕ Zpnk . k We shall use the Chinese remainder theorem to prove our first lemma. Lemma 1.7.11. Let g ∈ G be an element such that g m = 1 where m = m1 . . . mk and mi are pairwise relatively prime. Then there exist unique elements g1 , . . . , gk of G satisfying the following conditions: (a) gimi = e for all i ∈ {1, . . . , k} (b) g1 , . . . , gk commute pairwise and (c) g = g1 . . . gk . Proof. First we show existence, then uniqueness. The gi are in fact powers of g. Existence: By the Chinese Remainder Theorem, choose v1 , . . . , vk satisfying vi ≡ 1 mod mi and vi ≡ 0 mod m/mi . For each i, let vi = λi (m/mi ) for some λi ∈ Z and set gi = g vi . Then we have (i) gimi = g vi mi = g λi (m/mi )mi = g λi m and (ii) g1 , . . . , gk are powers of g and hence commute pairwise. (iii) Note that v1 + · · · + vk − 1 ≡ 0 mod mi for i = 1, 2, . . . , k, that is, mi |(v1 + · · · + vk − 1). Since m1 , . . . , mk are pairwise relatively prime, v1 + · · · + vk ≡ 1 mod m1 . . . mk , so g1 . . . gk = g v1 . . . g vk = g v1 +···+vk = g. Uniqueness: Suppose g = g1 . . . gk where g = g1 . . . gk and g1 , . . . , gk satisfy (i), (ii) and (iii). Then for each i, g vi = g1vi . . . gkvi = gi , that is, gi = g vi is the only possibility. Example 1.7.1. Consider g ∈ G with g 60 = e. Then m = 3 · 4 · 5, so v1 = 45, v2 = 40 and v3 = 36. Thus, g = g 45 g 40 g 36 = g1 g2 g3 . In case g has order m = pa11 . . . pakk where mi = pai i and p1 , . . . , pk are distinct primes, gi is called the pi -primary part of g. We have the first step of our decomposition. Theorem 1.7.12. Let A be a finite abelian group of order m = m1 . . . mk where m1 , . . . , mk are pairwise relatively prime. For each i ∈ {1, 2, . . . , k}, let Ai = {g ∈ A : g mi = e}. Then A1 × · · · × Ak ∼ = A. Moreover, |Ai | = mi for all i. Proof. Define φ : A1 × · · · × Ak → A by φ(g1 , . . . , gk ) = g1 . . . gk . Clearly, φ is a group homomorphism. By Lemma 1.7.11, φ is 1-1 and onto. Finally, m1 . . . mk = |A| = |A1 × · · · × Ak | = |A1 | . . . |Ak |. Let u1r m1 = pu1111 . . . p1r11 u2r m2 = pu2121 . . . p2r22 .. . ukr mk = puk1k1 . . . pkrkk . 37 1.7. Finite Abelian Groups Here, the pij are distinct primes and uij ≥ 1. Since every element of Ai satisfies g mi = e, |Ai | involves only those primes occuring in mi by Cauchy theorem. This forces |Ai | = mi for all i. If m = pa11 . . . pakk , then Ai in Theorem 1.7.12 is just the Sylow pi -subgroup of A. It now suffices to studied each factor Ai which is a Sylow pi -subgroup of A. To investigate them, we shall need the following definition. The positive integer n is an exponent for a group G if for each g ∈ G, g n = 1. In this case, G is said to have finite exponent and the least such n is called the exponent of G. For example, 12 is an exponent of Z6 but 6 is the exponent of Z6 . We denote the exponent of a group G (if exists) by exp G. Note that Z has no exponent and exp G divides |G|. The exponent of a finite abelian p-group gives information on its structure as follows. Theorem 1.7.13. Let A be an abelian group with |A| = pu where p is a prime. Suppose A has the exponent p, (that is, ap = e for all a ∈ A). Then A∼ = Zp × · · · × Zp = (Zp )u . | {z } u copies Proof. Note that if a ∈ A r {e}, hai ∼ = Zp . For any subset {a1 , . . . , ak } ⊆ A r {e}, we can define a group homomorphism θ : ha1 i × · · · × hak i → A by θ(ai11 , . . . , aikk ) = ai11 . . . aikk . We shall say that a1 , . . . , ak are “linearly independent” if θ is 1-1. This is equivalent to saying: If θ(ai11 , . . . , aikk ) = ai11 . . . aikk = e, then i1 ≡ . . . ≡ ik ≡ 0 mod p. Now there exists subsets of A r {e} for which θ is 1-1, e.g., the empty set, a singleton set. Choose a subset {a1 , . . . , ak } for which θ is 1-1 and k is as large as possible. We claim that in this case θ ha1 i × · · · × hak i → A is onto, and hence is an isomorphism. To see that θ is onto, let b ∈ A. If b = e, clearly b ∈ im θ. If b 6= e, consider θ̄ ha1 i × · · · × hak i × hbi → A. By the maximal choice of {a1 , . . . , ak }, θ̄ is not 1-1. Thus, e = θ̄(ai11 , . . . , aikk , bj ) = ai11 . . . aikk bj and j 6≡ 0 mod p since θ is not 1-1. Hence, there is a λ such that jλ ≡ 1 mod p, so λik jλ λik 1 1 e = aλi = aλi 1 . . . ak b 1 . . . ak b which implies b = a1−λi1 . . . ak−λik ∈ im θ. Therefore, θ is onto as claimed, and we have an isomorphism ha1 i × · · · × hak i θ ∼ = /A. Thus, pk = |ha1 i| . . . |hak i| = |ha1 i×· · ·×hak i| = |A| = pu , so k = u and the theorem is proved. Remark. If we write A in Theorem 1.7.13 additively, we see that it is just a vector space over the field Zp . Since A is finite, it is a finite dimensional vector space over Zp . All we were doing in Theorem 1.7.13 is finding a basis for A as a vector space over Zp . 38 1. Groups Theorem 1.7.14. [Burnside Basis Theorem for Abelian p-groups] Suppose A is an abelian group of exponent pk where p is a prime. Let Ap = {ap : a ∈ A}. If H is a subgroup of A and HAp = A, then H = A. Equivalently, if the cosets Ap a1 , . . . , Ap ak of A/Ap generate A/Ap , then a1 , . . . , ak generate A. 2 Proof. Observe that HAp = A implies H p Ap = Ap , so 2 2 3 3 A = HAp = H(H p Ap ) = HAp . 2 3 Also, HAp = A implies H p Ap = Ap , so A = HAp = H(H p Ap ) = HAp . r k k Continue inductively, we have A = HAp for all r. But Ap = {e}, so A = HAp = H. This completes the proof. Theorem 1.7.15. Let A be a finite abelian group of exponent p where p is a prime, and let H be a subgroup of A. Then there exists a subgroup K of A such that H ∩ K = {e} and HK = A. In other words, A is the internal direct product of H and K. Proof. Let K be a subgroup of A satisfying H ∩K = {e} and among all subgroups K of A satisfying H ∩ K = {e}, K is as large as possible. We claim that HK = A which proves the theorem. For, suppose conversely that a ∈ A and a ∈ / HK. Then H ∩ hK, ai = 6 {e} by the maximal choice of K, so there is a nontrivial element e 6= h = kai ∈ H ∩ hK, ai where h ∈ H ∈ and k ∈ K. If p|i, ai = e and h = k ∈ H ∩ K = {e}, a contradiction. If p ∤ i, there is a λ with aiλ = a (iλ ≡ 1 mod p) and then a = aiλ = (hk −1 )λ ∈ HK, a contradiction. Hence, HK = A as required. Remark. As with Theorem 1.7.13, the above theorem can be regarded as a statement about vector spaces over Zp as follows: If V is a finite dimensional vector space over Zp and U is a subspace, then there is a subspace W such that V = U ⊕ W . We are now ready to prove the structure theorem for a finite abelian p-group. Theorem 1.7.16. Let A be a finite abelian p-group. Then A is (isomorphic to) a direct product of cyclic groups. Proof. We use induction on |A|. If |A| = 1, the result is clear. Now suppose |A| = pu > 1. We assume inductively that any p-group where order is less than pu is a direct product of cyclic groups. Consider the group Ap = {ap : a ∈ A}. Claim A 6= Ap . For, suppose A = Ap . Then 2 u A = Ap = Ap = · · · = Ap = {e} since |A| = pu . As |A| > 1, we must have A 6= Ap . Thus, A > Ap , so |A| > |Ap | and by the inductive hypothesis, Ap is the (internal) direct product of cyclic subgroup. But every element of Ap has the form ap . Hence, so do the generators of these cyclic factors, so there exist a1 , . . . , ak ∈ A such that Ap = hap1 i × · · · × hapk i. 39 1.7. Finite Abelian Groups θ pik pi1 0 k More precisely, the map (a1pi1 , . . . , api k ) 7−→ a1 . . . ak is an isomorphism. Now let H = ha1 , . . . , ak i be the subgroup of A generated by a1 , . . . , ak . We claim that it is the (internal) direct product of the groups ha1 i, . . . , hak i, or that θ ha1 i × · · · × hak i → H (ai11 , . . . , aikk ) 7→ ai11 . . . aikk is an isomorphism. Since H = ha1 , . . . , ak i, θ is onto. To see that θ is 1-1, suppose e = θ(ai11 , . . . , aikk ) = ai11 . . . aikk . (1.7.1) k Then e = (ai11 . . . aikk )p = a1pi1 . . . api k , so pik 1 api 1 = · · · = ak = e since the map θ0 above is 1-1. Consider the integers i1 , . . . , ik . If p ∤ it for some of these integers, t λt t = (aλt t it )p = apt , where λt it ≡ 1 mod |at |. Thus, = e implies apt = e. For, (api then api t ) t p | i1 , . . . , ik , so i1 = pj1 , . . . , ik = pjk . But then (1.7.1) becomes pjk pjk pj1 1 e = θ(apj 1 , . . . , ak ) = a1 . . . ak , pjk 1 so apj 1 = · · · = ak = e since θ0 is 1-1. Hence, θ is 1-1. ✇ ✇✇ ✇✇ ✇ ✇✇ ✇✇ H❋ ❋❋ ✈✈ ✈ ❋❋ ✈✈ ❋❋ ✈ ✈ ❋❋ ✈ ✈✈ Ap = H p A ■■ ■■ ■■ ■■ ■■ ✈✈ ✈✈ ✈ ✈✈ ✈✈ Ap H ∩ Ap ❆❆ ❆❆ ❆❆ ❆❆ K ❍❍ ❍❍ ❍❍ ❍❍ {e} ⑥⑥ ⑥⑥ ⑥ ⑥ ⑥⑥ Next, let Ap = {a ∈ A : ap = e}. Then Ap is a finite group of exponent p and contains H ∩ Ap as a subgroup. Therefore, H ∩ Ap has a component in Ap by Theorem 1.7.15. More precisely, there is a subgroup K of Ap such that (a) (H ∩ Ap ) ∩ K = {e} and (b) (H ∩ Ap )K = Ap . Note that since K is a group of exponent p, K is a direct product of copies of Zp by Theorem 1.7.13. Finally, we claim that (I) H ∩ K = {e} and (II) HK = A. They implies that A is a direct product of H and K which are both direct products of cyclic groups. (I) Suppose H ∩ K 6= {e}. Thus, there is some x ∈ H ∩ K with x 6= e and xp = e. But then x ∈ Ap , so (H ∩ Ap ) ∩ K = (H ∩ K) ∩ Ap 6= {e}, contradicting (i) above. pik ik p i1 1 p (II) Suppose a ∈ A. Then ap ∈ Ap = hap1 i × · · · × hapk i, so ap = api 1 . . . ak = (a1 . . . ak ) = b i where b = ai11 . . . akk ∈ H = ha1 , . . . , ak i. Thus, b−1 a ∈ Ap = (H ∩ Ap )K ⊆ HK by (ii) above. Hence, a = b(b−1 a) ∈ HK and A = HK as required. This completes the proof. In addition, the above decomposition is unique. Hence, we are able to count the number of non-isomorphic abelian p-groups. 40 1. Groups Theorem 1.7.17. Suppose A = Zp × . . . Zp × Zp 2 × · · · × Zp 2 × · · · × Zp m × · · · × Zp m | {z } | {z } | {z } u1 copies is isomorphic to um copies u2 copies B = Zp × . . . Zp × Zp 2 × · · · × Zp 2 × · · · × Zp m × · · · × Zp m | {z } | {z } | {z } v1 copies vm copies v2 copies where ui , vi ≥ 1. Then ui = vi for all i = 1, . . . , m. In other words, the orders and multiplicities of the factors in a decomposition of a finite abelian p-group uniquely determine the group up to isomorphism. Proof. Since A ∼ = B, it follows that for any positive integer n, # of solutions of xn = e in A = # of solutions of xn = e in B. Consider the following table. n p p2 p3 .. . # of solutions of xn = e in A pu1 +u2 +u3 +···+um u p 1 +2u2 +2u3 +···+2um pu1 +2u2 +3u3 +···+3um .. . # of solutions of xn = e in B pv1 +v2 +v3 +···+vm v p 1 +2v2 +2v3 +···+2vm pv1 +2v2 +3v3 +···+3vm .. . pm−1 pm pu1 +2u2 +3u3 +···+(m−1)um−1 +(m−1)um pu1 +2u2 +3u3 +···+(m−1)um−1 +mum pv1 +v2 +···+(m−1)vm−1 +(m−1)vm pv1 +v2 +···+(m−1)vm−1 +mvm Then we have u1 + u2 + u3 + · · · + um = v 1 + v 2 + v 3 + · · · + v m u1 + 2u2 + 2u3 + · · · + 2um = v1 + 2v2 + 2v3 + · · · + 2vm u1 + 2u2 + 3u3 + · · · + 3um = v1 + 2v2 + 3v3 + · · · + 3vm .. . u1 + 2u2 + 3u3 + · · · + (m − 1)um−1 +(m − 1)um = v1 + 2v2 + 3v3 + · · · + (m − 1)vm−1 + (m − 1)vm u1 + 2u2 + 3u3 + · · · + (m − 1)um−1 +mum = v1 + 2v2 + 3v3 + · · · + (m − 1)vm−1 + mvm . It is easy to see that the above equations force u1 = v1 , u2 = v2 , . . . , um = vm as required. Theorem 1.7.18. Let p be any prime and n a positive integer. Then {r1 ≤ r2 ≤ · · · ≤ rk } ←→ Zpr1 × Zpr2 × · · · × Zprk defines a 1-1 correspondence between partitions of n and isomorphism classes of abelian groups of order pn . In particular, the number of isomorphism classes of abelian groups of order pn is the number of partitions of n. Examples 1.7.2. 1. Abelian groups of order p3 . 41 1.7. Finite Abelian Groups partitions of 3 {1, 1, 1} {1, 2} {3} corresponding abelian groups Zp × Zp × Zp Zp × Zp 2 Zp 3 2. Abelian groups of order p5 . partitions of 5 {1, 1, 1, 1, 1} {1, 1, 1, 2} {1, 2, 2} {1, 1, 3} {2, 3} {1, 4} {5} Let corresponding abelian groups Zp × Zp × Zp × Zp × Zp Zp × Zp × Zp × Zp 2 Zp × Zp 2 × Zp 2 Zp × Zp × Zp 3 Zp 2 × Zp 3 Zp × Zp 4 Zp 5 Suppose A is a finite abelian group of order pa11 . . . pakk where p1 , . . . , pk are distinct primes. ai Ai = {g ∈ A : g pi = e}. a By Theorem 1.7.12, A ∼ = A1 × · · · × Ak where |Ai | = pi i . Since each Ai is a direct product of cyclic group by Theorem 1.7.17, this yields the following theorem. Theorem 1.7.19. A finite abelian group is (isomorphic to) a direct product of cyclic groups. Corollary 1.7.20. If m is a square free integer, then every abelian group of order m is cyclic. Proof. Assume that an abelian group A is of order m = p1 . . . pr where pi are distinct primes. Then A∼ = Zm = Zp 1 × · · · × Zp r ∼ by Theorem 1.7.12 and the Chinese remainder theorem, respectively. Let A ∼ = A1 × · · · × Ak as above. It is clear that Ai is the unique largest pi -subgroup of A. Moreover, if B is finite abelian, and B ∼ = B1 × · · · × Bk where Bi is a pi -group, then A∼ = Bk ). = B1 ∧ . . . ∧ Ak ∼ = B ⇔ (A1 ∼ u Recall that if m and n are relatively prime then Zm ×Zn ∼ = Zmn . It follows that if n = pu1 1 · · · pk k where p1 , . . . , pk are distinct primes, then Z p u1 × · · · × Z p uk ∼ = Zn . 1 (1.7.2) k This gives rise to a second way of writing a finite abelian group A as a direct product of cyclic groups. Namely, let p1 , . . . , pk be the primes dividing |A|, and let A = A1 × · · · × Ak , where Ai is the pi -primary part of A. Express each Ai as a direct product of cyclic factors and assume that t is the largest number of factors occurring in any Ai . Write A1 = Zpv11 × Zpv12 × · · · × Zpv1t 1 1 1 A2 = Zpv21 × Zpv22 × · · · × Zpv2t 2 2 2 .. . Ak = Zpvk1 × Zpvk2 × · · · × Zpvkt k k k 42 1. Groups where 0 ≤ vi1 ≤ vi2 ≤ · · · ≤ vit (1.7.3) and we have allowed (for notational convenience) some vij to be zero. Let n1 = pv111 pv221 · · · pvkk1 n2 = pv112 pv222 · · · pvkk2 .. . ni = pv11i pv22i · · · pvkki .. . nt = pv11t pv22t · · · pvkkt . Condition (1.7.3) guarantees that n1 | n2 , n2 | n3 , . . . , nt−1 | nt . Then, using (1.7.2) gives A = A1 × · · · × Ak = (Zpv11 × · · · × Zpv1t ) × (Zpv12 × · · · × Zpvk2 ) × · · · × (Zpv1t × · · · × Zpvkt ) 1 1 1 k 1 k ∼ = Zn1 × Zn2 × · · · × Znt . The integers nj are completely determined by the decomposition of A into a direct of cyclic pi -groups. Conversely, given that A = Zn1 × Zn2 × · · · × Znt . where n1 | n2 , n2 | n3 , . . . , nt−1 | nt , the decomposition of A into a direct product of cyclic pi groups is completely determined. Thus, we have: Theorem 1.7.21. [Structure Theorem for Finite Abelian Groups] Let A be a finite abelian group. Then there exist integers n1 , . . . , nt > 1 such that n1 | n2 , n2 | n3 , . . . , nt−1 | nt and A∼ = Zn1 × Zn2 × · · · × Znt , where these integers are uniquely defined by A. More precisely, if m1 , . . . , ms are positive integers greater than 1 such that m1 | m2 , m2 | m3 , . . . , ms−1 | ms , and A∼ = Z m1 × Z m2 × · · · × Z ms , = Zn1 × Zn2 × · · · × Znt ∼ then t = s, and n1 = m1 , . . . , nt = mt . Example 1.7.3. Find all non-isomorphic abelian groups of order: 1. 6 2. 12 3. 27 4. 500. Solution. By Example 1.7.2 and Theorem 1.7.21. We have the following answers. 6 = 2 · 3: Z2 × Z3 ∼ = Z6 , 2 12 = 2 · 3: Z2 × Z2 × Z3 ∼ = Z2 × Z6 and Z22 × Z3 ∼ = Z12 , 3 27 = 3 : Z3 × Z3 × Z3 , Z3 × Z32 and Z33 , 500 = 22 · 53 : Z2 × Z2 × Z5 × Z5 × Z5 ∼ = Z2 × Z10 × Z10 , Z2 × Z2 × Z5 × Z52 ∼ = Z10 × Z50 , ∼ ∼ ∼ Z2 × Z2 × Z53 = Z2 × Z250 , Z22 × Z5 × Z5 × Z5 = Z5 × Z5 × Z20 , Z22 × Z5 × Z52 = Z5 × Z100 and Z22 × Z53 ∼ = Z500 . By Cauchy’s theorem, if p is a divisor of the order of a group G, then G has a subgroup of order p. We also see that A4 is a group of order 12 but it has no subgroup of order six. Hence, it may not hold that G will have a subgroup of order m when m is a divisor of |G|. However, if G is an abelian group, we have our final results. 43 1.7. Finite Abelian Groups Corollary 1.7.22. Let A be a finite abelian group. If m divides the order of A, then A has a subgroup of order m. Proof. Write A = Zn1 × Zn2 × · · · × Znt as in the above theorem. Then |A| = n1 n2 . . . nt . Since m divides |A|, m = l1 l2 . . . lt with li | ni for all i ∈ {1, 2, . . . , t}. Then (ni /li )Zni is a subgroup of Zni of order li for all i. Thus, (n1 /l1 )Zn1 × (n2 /l2 )Zn2 × · · · × (nt /lt )Znt is a subgroup of A of order l1 l2 . . . lt = m as desired. Corollary 1.7.23. Let A be a finite abelian group. Then there exists g ∈ A such that the order of g is the exponent of A. Proof. By Theorem 1.7.21, there exist positive integers n1 , n2 , . . . , nt ≥ 1 such that n1 | n2 | · · · | nt and A∼ = Zn1 × Zn2 × · · · × Znt . Thus, exp A = nt and (0, 0, . . . , 0, 1) in the rightmost group has order nt . Remark. The above corollary is false if A is non-abelian. For example, the exponent of S3 is 6, however S3 contains no elements of order 6. Exercises 1.7. 1. Suppose G1 and G2 are finite groups of relatively prime orders. Show that every subgroup of G1 ×G2 is of the form H1 ×H2 for some subgroups H1 and H2 of G1 and G2 , respectively. 2. Let G1 and G2 be simple groups. Show that every nontrivial normal subgroup of G = G1 × G2 is isomorphic to either G1 or G2 . 3. Proof Theorem 1.7.8. 4. Find the order of torsion subgroup of Z4 × Z × Z and of Z12 × Z12 × Z. 5. Find the torsion subgroup of the multiplicative group R∗ . 6. Let G be an abelian group of order 72. (a) Can you say how many subgroups of order eight G has? Why? (b) Can you say how many subgroups of order four G has? Why? 7. Show that a finite abelian group is not cyclic if and only if it contains a subgroup isomorphic to Zp × Zp for some prime p. 8. Find the exponent of the following groups: (a) Z6 × Z9 (b) Z5 × S4 . 9. Find all non-isomorphic abelian groups of order (a) 35 (b) 48 (c) 360. 10. List (up to isomorphism) all abelian groups of order 108 and express each in the following two ways: (a) As a direct sum of cyclic groups of prime power order. (b) As a direct sum of cyclic groups of order d1 , d2 , . . . , dk where di | di+1 for i = 1, 2, . . . , k − 1. 11. List all groups of order 99 and of order 1225 up to isomorphism. (Hint. Show that they must be abelian.) 12. Let G be a finite group. Prove the following statements. (a) The exponent of G divides its order |G|. (b) If H is a subgroup of G, then the exponent of H divides the exponent of G. (c) If G is cyclic, then exp G = |G|. 13. Let G = S3 × Z4 . (a) Find exp G. (b) Determine all Sylow 2-subgroups and Sylow 3-subgroups of G. 14. Let G be a group of order 2156 = 22 · 72 · 11. (a) If G is abelian, list all G up to isomorphism. (b) Prove that G cannot be simple. 15. List all finite groups G which have the property: ∀g, h ∈ G, g is a power of h or h is a power of g. Project 9 (Characters of a group). Let S1 = {z ∈ C : |z| = 1} be the unit circle in the complex plane. It is a subgroup of C× . A character of a finite group G is a homomorphism from G to S1 . The character sending G to 1 is called the trivial character. Let G be a finite group. Prove the following statements. 44 1. Groups (a) All characters of G form an abelian group under pointwise multiplication, called the dual group of b G and denoted by G. b are isomorphic. (b) Prove that if G is a finite cyclic group, then G and G b (Hint. Use (b) and Theorem (c) If G is a finite abelian group, then G is isomorphic to its dual group G. 1.7.19.) This concludes the basic theory of groups. More advanced group theory will be studied in Chapter 3. 2 | Rings and Fields Rings and fields are the most common algebraic structures for students. They have learned addition together with multiplication since elementary schools. The abstract treatments using groups are presented in the first section. Ideals and factorizations are discussed in details. Finally, we talk about polynomials over a ring and which will be used in a construction of field extensions. 2.1 2.1.1 Basic Concepts Rings A ring is a triple (R, +, ·) where (R1) (R, +) is an abelian group, (R2) (R, ·) is a semigroup, and (R3) + and · satisfy the distributive laws, namely, ∀a, b, c ∈ R, a · (b + c) = (a · b) + (a · c) and ∀a, b, c ∈ R, (b + c) · a = (b · a) + (c · a). The binary operations + and · are called the addition and the multiplication of the ring (R, +, ·), respectively, 0, the identity of (R, +) is called the zero of R, for a, b ∈ R, ab (juxtaposition) may denote a · b and (R, +, ·) may be denoted by R. For a ∈ R, −a is called the additive inverse of a in R. A ring R is said to be commutative if ∀a, b ∈ R, ab = ba. If 1 is the identity of (R, ·), then 1 is called the identity or unity of the ring R. If R contains an identity, i.e., (R, ·) is a multiplicative monoid, then R is called a ring with identity. Unless the contrary is explicitly stated “ring” will mean “ring with identity”. A subset S of a ring R is a subring if S is a subgroup of the additive group and also a submonoid of the multiplicative monoid of R. Clearly the intersection of any set of subrings of R is a subring. Hence, if A is a subset of R, we may define the subring generated by A to be the intersection of all subrings of R which contain A. Examples 2.1.1 (Examples of rings). 1. Z, Q, R and C are commutative rings under usual addition and multiplication. 2. For n ∈ N, (Zn , +, ·) is a commutative ring. 3. Recall that for a nonempty set X and P (X) the power set of X, we define A△B = (A r B) ∪ (B r A) for all subsets A and B of X. Then (P (X), △, ∩) is a commutative ring with identity X. 4. If A is an abelian group, then End(A), the set of all homomorphisms on A, is a ring with the addition and multiplication are given by (f + g)(a) = f (a) + g(a) and (f · g)(a) = f (g(a)) for all f, g ∈ End(A) and a ∈ A. Its identity is the identity map. 45 46 2. Rings and Fields 5. If X is a topological space, then C(X), the set of all continuous functions from X to R, is a commutative ring with pairwise operations. That is, (f + g)(x) = f (x) + g(x) and (f · g)(x) = f (x)g(x) for all f, g ∈ C(X) and x ∈ X. √ √ 6. Let d be a square free integer. The set Z[ d] = {a + b d : a, b ∈ Z} is a subring of C. A number of elementary properties of rings are consequences of the fact that a ring is an abelian group relative to addition and a monoid relative to multiplication. For example, we have −(a + b) = −a − b := (−a) + (−b) and if na is defined for n ∈ Z as before, then the rules for multiples (or powers) in an abelian group, n(a + b) = na + nb, (n + m)a = na + ma, and (nm)a = n(ma) hold. There are also a number of simple consequences of the distributive laws which we now note. In the first place, induction on m and n give the generalization ! n m m X n X X X ai a i bj . bj = i=1 j=1 i=1 j=1 We note next that a0 = 0a = 0 for all a ∈ R; for we have a0 = a(0 + 0) = a0 + a0. Addition of −a0 gives a0 = 0. Similarly, 0a = 0. We have the equation 0 = 0b = (a + (−a))b = ab + (−a)b, which shows that (−a)b = −ab. Similarly, a(−b) = −ab; consequently (−a)(−b) = −a(−b) = −(−ab) = ab. If a and b commute, that is, ab = ba, then am bn = bn am . Also, by induction we can prove the binomial theorem n n−1 n n−2 2 n n n (a + b) = a + a b+ a b + ··· + abn−1 + bn 1 2 n−1 for all n ∈ N. Remark. In the case 1 = 0 in a ring R, we have that ab = (a · 1)b = a(1 · b) = a(0 · b) = a · 0 = 0 for all a, b ∈ R. A ring with this property is called a zero ring. If R and S are rings, so is R × S, with coordinatewise operations: (r, s) + (r′ , s′ ) = (r + r′ , s + s′ ) and (r, s)(r′ , s′ ) = (rr′ , ss′ ). Q Note that (1R , 1S ) is the identity of R × S. More generally, if Rα is a family of rings, then α Rα is a ring with coordinatewise operations. Let R be a ring. An element x ∈ R is said to be invertible or a unit if there is a y ∈ R such that xy = yx = 1. In this case, y is called the inverse of x. 47 2.1. Basic Concepts Remark. If x is invertible, its inverse is unique. The invertible elements of R form a group under multiplication, called the group of units of R and denoted by U (R) or R× . A ring D is a division ring or skew field if every nonzero element of D is invertible. A commutative division ring is called a field. Examples 2.1.2. 1. Q, R and C are fields. Z is not a field. 2. By Example 1.2.1 (5), we have ∀n ∈ N, Zn is a field ⇔ n is a prime. Example 2.1.3. Let n ∈ N, R a ring and Mn (R) the set of all n × n matrices over R. Then (Mn (R), +, ·) is a ring under the usual addition and multiplication of matrices with unity In , the identity matrix. If n > 1, then Mn (R) is not commutative. The group of invertible elements of Mn (R) is called the general linear group and denoted by GLn (R). For the case R is commutative, we can derive the determinant criterion for a matrix A to be invertible. We have the following results. Theorem 2.1.1. Let R be a commutative ring and A ∈ Mn (R). Then A(adj A) = (det A)In = (adj A)A. In particular, A is invertible if and only if its determinant is invertible in R. A noteworthy special case of the theorem is the next corollary. Corollary 2.1.2. If F is a field, A ∈ Mn (F ) is invertible if and only if det A 6= 0. Some rings do not have the property that the product of two nonzero elements is always nonzero. If so, it leads to the cancellation property in the rings. Let R be a ring and 0 6= a ∈ R. a is called a left [right] zero divisor if ∃b ∈ R r {0}, ab = 0 [ba = 0], and it is called a zero divisor if it is both a left and a right zero divisor. R is entire if it possesses no zero divisors. A commutative entire ring is called an integral domain. Examples 2.1.4. 1. Z is an integral domain which is not a field. 2. Every field is an integral domain. 3. C([0, 1]) is not an integral domain. Remark. Let R be a ring. Then we have R is entire ⇔ ∀a, b ∈ R [ab = 0 ⇒ (a = 0 or b = 0)] ⇔ (R r {0}, ·) is a cancellative semigroup. Theorem 1.2.2 implies an important result on finiteness of an integral domain. Theorem 2.1.3. Every finite integral domain is a field. Proof. Let D be a finite integral domain. Then (D r {0}, ·) is a finite cancellative semigroup. By Theorem 1.2.2, (D r {0}, ·) is a group. Hence, if a ∈ D and a 6= 0, then a has an inverse under ·. Since D is commutative, D is a field. 48 2.1.2 2. Rings and Fields Quaternions In 1843, W. R. Hamilton constructed the first example of a division ring in which the commutative law of multiplication does not hold. This was an extension of the field of complex numbers, whose elements were quadruples of real numbers (α, β, γ, δ) for which the usual addition and a multiplication were defined so that 1 = (1, 0, 0, 0) is the unit and i = (0, 1, 0, 0), j = (0, 0, 1, 0), and k = (0, 0, 0, 1) satisfy i2 = j 2 = k 2 = −1 = ijk. Hamilton called his quadruples, quaternions. Previously, he had defined complex numbers as pairs of real numbers (α, β) with the product (α, β)(γ, δ) = (αγ − βδ, αδ + βγ). Hamilton’s discovery of quaternions led to a good deal of experimentation with other such “hypercomplex” number systems and eventually to a structure theory whose goal was to classify such systems. A good deal of important algebra thus evolved from the discovery of quaternions. We shall not follow Hamilton’s way of introducing quaternions. Instead we shall define this system as a certain subring of the ring M2 (C) of 2 × 2 matrices with complex number entries. This will have the advantage of reducing the calculations to a single simple verification. We consider the subset H of the ring M2 (C) of complex 2 × 2 matrices that have the form √ √ a b α0 + α1 √−1 α2 + α3 √−1 x= = , αi real. (2.1.1) −α2 + α3 −1 α0 − α1 −1 −b̄ ā We claim that H is a subring of M2 (C). Since a1 − a2 = ā1 − ā2 for complex numbers, it is clear that H is closed under subtraction; hence H is a subgroup of the additive group of M2 (C). We obtain the unit matrix by taking a = 1, b = 0 in (2.1.1). Hence, 1 ∈ H. Since a b c d ac − bd¯ ad + bc̄ = −b̄ ā −d¯ c̄ −b̄c − ād¯ −b̄d + āc̄ and a1 a2 = ā1 ā2 , the right-hand side has the form u v −v̄ ū ¯ v = ad + bc̄. Therefore, H is closed under multiplication and so H is a subring where u = ac − bd, of M2 (C). We now show that H is a division ring. We note first that √ √ α0 + α1 √−1 α2 + α3 √−1 ∆ := det = α02 + α12 + α22 + α32 . −α2 + α3 −1 α0 − α1 −1 Since the αi are real numbers, this is real, and is 0 only if every αi = 0, that is, if the matrix is 0. Hence, every non-zero element of H has an inverse in M2 (C). Moreover, we have, by the definition of the adjoint, that a b ā −b adj = . −b̄ ā b̄ a ¯ = a, this is obtained from the x in (2.1.1) by replacing a by ā and b by −b, and so it is Since ā contained in H. Thus, if the matrix x is 6= 0 then its inverse is −1 ā∆ −b∆−1 b̄∆−1 a∆−1 and this is contained in H. Hence, H is a division ring. The ring in its center the field R of real numbers identified with the set of diagonal H contains α 0 matrices , α ∈ R. H also contains the matrices 0 α √ √ −1 0 −1 0 1 0 √ √ ,j = . ,k = i= 0 − −1 −1 0 −1 0 49 2.1. Basic Concepts We verify that x = α0 + α1 i + α2 j + α3 k (2.1.2) and if α0 + α1 i + α2 j + α3 k = β0 + β1 i + β2 j + β3 k, βi ∈ R, then √ √ √ √ α0 + α1 √−1 α2 + α3 √−1 β0 + β1 √−1 β2 + β3 √−1 = −α2 + α3 −1 α0 − α1 −1 −β2 + β3 −1 β0 − β1 −1 so αl = βl , 0 ≤ l ≤ 3. Thus, any x in H can be written in one and only one way in the from (2.1.2). The product of two elements in H (α0 + α1 i + α2 j + α3 k)(β0 + β1 i + β2 j + β3 k) is determined by the product and sum in R, the distributive laws and the multiplication table i2 = j 2 = k 2 = −1 ij = −ji = k, jk = −kj = i, ki = −ik = j. Incidentally, because these show that H is not commutative we have constructed an infinite division ring that is not a field. The ring H is called the division ring of real quaternions. Recall that Q8 = {±1, ±i, ±j, ±k} is called the quaternion group (see Porject 1). Algebra is not as rich in division ring as it is in fields. For example, there are no finite division rings. This is the content of a famous theorem of Wedderburn. Its proof can be found in Section 5.7. Theorem 2.1.4. [Wedderburn, 1909] A finite division ring is a field. 2.1.3 Characteristic Let R be a ring. If there is a smallest positive integer n such that na = 0 for all a ∈ R, then R is said to have characteristic n. If no such n exists, R is said to have characteristic zero. We denote the characteristic of R by char R. The characteristic of a ring gives some information on its additive group structure. Remark. It is easy to see that char R = n if and only if n is the smallest positive integer such that n1R = 0. Example 2.1.5. The rings Z, Q, R, C and H are of characteristic zero, char Zn = n and char(Zm × Zn ) = lcm(m, n). Theorem 2.1.5. [Characteristic of an Integral Domain] If R is an integral domain, then R is of characteristic zero or a prime p. In particular, every field is of characteristic zero or a prime p. Proof. Let R be an integral domain of characteristic n > 0. Assume that n = ab for some a, b ∈ N. It follows that 0 = n1R = (ab)1R = (a1R )(b1R ). Since R has no zero divisor, a1R = 0 or b1R = 0. Then a = n or b = n. Hence, n is a prime. Theorem 2.1.6. Let R be a ring of characteristic a prime p and a, b ∈ R. If a and b commute, then (a + b)p = ap + bp and k k (a + b)p = ap + bp k for all k ∈ N. 50 2. Rings and Fields Proof. Note that if 1 ≤ r ≤ p − 1, then the binomial coefficient pr is a multiple of p, so it is 0 in R. Hence, p p p−1 p p abp−1 + bp = ap + bp . a b + ··· + (a + b) = a + p−1 1 A simple induction on k gives the second equation. The inductive step is k (a + b)p = ((a + b)p k−1 )p = (ap k−1 − bp k−1 k ) p = a p + bp k and the proof is complete. 2.1.4 Ring Homomorphisms and Group Rings Like in groups, a ring homomorphism is a function between two rings that preserves both addition and multiplication. Let R and S be rings. A map ϕ : R → S is called a homomorphism if ϕ(a + b) = ϕ(a) + ϕ(b) and ϕ(ab) = ϕ(a)ϕ(b) for all a, b ∈ R. The definitions of monomorphisms, epimorphisms, endomorphisms, isomorphisms and automorphisms of rings are given as in groups. Remarks. Let ϕ : R → S be a ring homomorphism. 1. ϕ : (R, +) → (S, +) is an additive group homomorphism. 2. ϕ(1R ) may not be the identity of S. 3. If ϕ(1R ) = 0, then ϕ(x) = 0 for all x ∈ R. 4. If R has an identity and ϕ is onto, then ϕ(1R ) is the identity of S. Proof. Let s ∈ S. Since ϕ is onto, ∃x ∈ R, ϕ(x) = s. Then sϕ(1R ) = ϕ(x)ϕ(1R ) = ϕ(x1R ) = ϕ(x) = s. Similarly, ϕ(1R )s = s. Hence, ϕ(1R ) is the identity of S. Examples 2.1.6. 1. If ϕ : Z → Z is a ring homomorphism, then ϕ is the zero or the identity map. 2. If ϕ : Q → Q is a ring homomorphism, then ϕ is the zero or the identity map. 3. If ϕ : R → R is a ring homomorphism, then ϕ is the zero or the identity map. 4. If p is a prime and ϕ : Zp → Zp is a ring homomorphism, then ϕ is the zero or the identity map. Proof. (1) Observe that ϕ(1) = ϕ(1 · 1) = ϕ(1)ϕ(1), so ϕ(1) = 0 or ϕ(1) = 1. Moreover, ϕ(n) = nϕ(1) for all n ∈ Z. Thus, ϕ(n) = 0 for all n ∈ Z or ϕ(n) = n for all n ∈ Z as desired. (2) Similar to Z, ϕ(1) = 0 or ϕ(1) = 1 and ϕ(n) = nϕ(1) for all n ∈ Z. For m ∈ Z and n ∈ N, we have ϕ(m(1/n)) = mϕ(1/n) and ϕ(1) = ϕ(n(1/n)) = nϕ(n). If ϕ(1) = 0, then ϕ(1/n) = 0 for all n ∈ N, so ϕ is the zero map. On the other hand, if ϕ(1) = 1, then ϕ(1/n) = 1/n for all n ∈ N which implies ϕ(m/n) = m/n for all m ∈ Z and n ∈ N. (3) Assume that ϕ(x) is not the zero map. We can show that ϕ(x) = x for all x ∈ Q. Moreover, for √ √ x ∈ R+ , ϕ(x) = ϕ(( x)2 ) = (ϕ( x))2 > 0. This implies ∀a, b ∈ R, a < b ⇒ ϕ(a) < ϕ(b). Now, let x ∈ R. Suppose that ϕ(x) 6= x. Then ϕ(x) < x or x < ϕ(x). By the density theorem, ∃q1 , q2 ∈ Q such that ϕ(x) < q1 < x or x < q2 < ϕ(x). Thus, ϕ(x) < q1 < ϕ(x) or ϕ(x) < q2 < ϕ(x) yields a contradiction. Hence, ϕ(x) = x for all x ∈ R. (4) is proved in the next section. Let G = {gi : i ∈ I} be any multiplicative group, and let R be any commutative ring. Let RG be the set of all formal sums X ai gi i∈I 51 2.1. Basic Concepts for ai ∈ R and gi ∈ G, where all but finite number of the ai are 0. Define the sum of two elements of RG by X i∈I ai gi + X i∈I bi gi = X (ai + bi )gi . i∈I Observe that (ai + bi ) = 0 except for a finite number of indices i, so the above sum is again in RG. P It is immediate that (RG, +) is an abelian group with additive identity i∈I 0gi . Multiplication of two elements of RG is defined by the use of the multiplications in G and R as follows: X X X X ai gi bi gi = a j bk g i . i∈I i∈I i∈I gj gk =gi P Naively, we formally distribute the sum i∈I ai gi over the sum i∈I bi gi and rename a term a j g j bk g k P by aj bk gk where gj gk = gi in G. Since ai and bi are 0 for all but a finite number of i, the sum gj gk =gi aj bk contains only a finite number of nonzero summands aj bk ∈ R and may P thus be viewed as an element of R. Again at most a finite number of such sums gj gk =gi aj bk are nonzero. Thus, multiplication is closed on RG. We can proceed to show that P Theorem 2.1.7. [Group Ring] If G is a multiplicative group and R is a commutative ring, then (RG, +, ·) is a ring with unity 1R e. P If we rename the element i∈I ai gi of RG, where ai = 0 for i 6= j and aj = 1, by gj , we see that (RG, ·) can be considered to contain G naturally. Thus, if G is not abelian, RG is not commutative. Clearly, char RG = char R, for any group G. The ring RG defined above is the group ring of G over R. If F is a field, then F G is the group algebra of G over F . Exercises 2.1. 1. Define an addition and a multiplication on Z by a⊕b=a+b−1 and a ⊙ b = ab − (a + b) + 2 for all a, b ∈ Z. Prove that (Z, ⊕, ⊙) is an integral domain. √ 2. Let S be the set of complex numbers of the form m + n −3 where either m, n ∈ Z or both m and n are halves of odd integers. Show that S is a subring of C. 3. Show that if 1 − ab is invertible in a ring, then so is 1 − ba. 4. Let a and b be elements of a ring such that a, b and ab − 1 are units. Show that a − b−1 and (a − b−1 )−1 − a−1 are units and the following identity holds: ((a − b−1 )−1 − a−1 )−1 = aba − a. 5. A ring R is called a Boolean ring is x2 = x for all x ∈ R. Prove that every Boolean ring is commutative. 6. (a) Show that ϕ : Z12 → Z30 given by ϕ([a]12 ) = [10a]30 is a ring homomorphism. (b) Show that ϕ : Z12 → Z30 given by ϕ([a]12 ) = [5a]30 is a additive group homomorphism. Is it a ring homomorphism? 7. Consider (S, +, ·), where S is a set and + and · are binary operations on S which satisfy the distributive laws such that (S, +) and (S r {0}, ·) are groups. Show that (S, +, ·) is a division ring. 8. Let R be a ring. Define C(R) = {x ∈ R : ∀y ∈ R, xy = yx}, called the center of R. (a) Prove that C(R) is a commutative subring of R. (b) Determine the centers of H and Mn (F ) where F is a field. (c) If R is a division ring, show that C(R) is a field. 9. If p is a prime and ϕ : Zp → Zp is a ring homomorphism, show that ϕ is the zero or the identity map. 10. Show that if F is a field, A ∈ Mn (F ) is a zero divisor in this ring if and only if A is not invertible. Does this hold for arbitrary commutative ring R? Explain. 52 2. Rings and Fields 11. Let m and n be non-zero integers and let R be the subset of M2 (C) consisting of the matrices of the form √ √ c + d√m a + b √m n(c + d m) a − b m where a, b, c, d ∈ Q. Show that R is a subring of M2 (C) and that R is a division ring if and only if the only rational numbers x, y, z, t satisfying the equation x2 − my 2 − nz 2 + mnt2 = 0 are x = y = z = t = 0. Give a choice of m, n that R is a division ring and a choice of m, n that R is not a division ring. 12. Let R be a ring which may not contain the unity 1. Define two binary operations on R × Z by (r, k) + (s, m) = (r + s, k + m) and (r, k) · (s, m) = (rs + ks + mr, km). Prove that (R × Z, +, ·) is a ring with unity (0, 1) and of characteristic zero. Ditto the set R × Zn and prove that it is a ring of characteristic n. 13. A ring R is simple if R and {0} are the only ideals in R. Show that the characteristic of a simple ring is either 0 or a prime p. 14. If R is a finite integral domain, prove that |R| is a prime power. 2.2 Ideals, Quotient Rings and the Field of Fractions Ideals play an important role in ring theory. They are used to construct quotient rings like normal subgroups. Let R be a ring. A subset I of R is called a left [right] ideal of R if 1. I is a subgroup of (R, +) and 2. ∀r ∈ R∀a ∈ I, ra ∈ I [ar ∈ I]. It is called a two-sided ideal or an ideal of R if I is both a left and a right ideal. E.g., {0} and R are two-sided ideals of R. Theorem 2.2.1. Let ϕ : R → S be a homomorphism of rings. Then the kernel of ϕ given by ker ϕ = {x ∈ R : ϕ(x) = 0S } is an ideal of R. Proof. It is immediate that ker ϕ is a subgroup of (R, +). If a ∈ R and x ∈ ker ϕ, then ϕ(ax) = ϕ(a)ϕ(x) = ϕ(a)0 = 0 and ϕ(xa) = ϕ(x)ϕ(a) = 0ϕ(a) = 0. Hence, ax and xa are in ker ϕ. Remark. Similar to a group homomorphism, for a ring homomorphism, we have ϕ is one-to-one if and only if ker ϕ = {0}. For subsets X and Y of R, let XY denote the set of all finite sums in the form n X i=1 x i yi , where xi ∈ X, yi ∈ Y and n ∈ N. For a ∈ R, we have Ra = {ra : r ∈ R} and aR = {ar : r ∈ R}. Examples 2.2.1. 1. All distinct ideals of Zn are dZn , where d = 0 or (d ∈ N and d | n). 2. All distinct ideals of Z are mZ, where m ∈ N ∪ {0}. Remarks. Let R be a ring. 1. If I is an ideal of R, then IR = I = RI. 2. If a left [right, two-sided] ideal I of R contains a unit, then I = R. 3. If R is a division ring, then {0} and R are the only left [right, two-sided] ideals of R. 2.2. Ideals, Quotient Rings and the Field of Fractions 53 4. An arbitrary intersection of left [right, two-sided] ideals of R is a left [right, two-sided] ideal of R. 5. If S is a subring of R and I is an ideal of R, then S + I is a subring of R, I is an ideal of S + I and S ∩ I is an ideal of S. 6. If I and J are left [right, two-sided] ideals of R, then I + J and IJ are left [right, two-sided] ideals of R. a b Example 2.2.2. Let R be a ring and S = : a, b, c ∈ R a subring of M2 (R). 0 c a 0 0 0 Then : a ∈ R is a left ideal of S and : a ∈ R is a right ideal of S. 0 0 0 a They are not ideals of S. Let R be a ring. If X ⊂ R, then the ideal of R generated by X is the intersection of all ideals containing X and it is denoted by (X), so (X) is the smallest ideal of R containing X. For a1 , . . . , an ∈ R, let (a1 , . . . , an ) denote ({a1 , . . . , an }). An ideal I of R is called a principal ideal if I = (a) for some a ∈ R. Observe that for a ∈ R, (m ) X (a) = ri asi : ri , si ∈ R and m ∈ N . i=1 If R is a commutative ring, then ∀a ∈ R, (a) = aR = Ra. A ring R is a principal ideal ring if every ideal of R is principal. A principal ideal domain (PID) is a principal ideal ring which is an integral domain. Hence, an integral domain R is a PID if {Ra : a ∈ R} is the set of all ideals of R. Examples 2.2.3. 1. Zn is a principal ideal ring. 2. Z is a PID. 3. Every field has only two ideals, namely (0) and (1) = F , so it is a PID. Remark. Let F be a field, R a ring and ϕ : F → R a ring homomorphism. Then ker ϕ is either {0} or F which implies ϕ is 1-1 or is the zero map, respectively. Hence, every nonzero ring homomorphism of fields must be 1-1. In particular, one can readily verify that the only ring endomorphisms of Zp are the zero map and the identity map. This finishes the proof of Example 2.1.6. Theorem 2.2.2. Let R be a commutative ring whose only ideals are {0} and R itself. Then R is a field. Proof. Let a ∈ R r {0}. Then (a) = R, so 1 ∈ (a). Since R is commutative, there is a b ∈ R such that ab = 1 = ba. Suppose I is an ideal of R. Then I is a subgroup of R, considered as an abelian group, and so we can form the abelian group R/I. The elements of R/I are cosets r + I = {r + a : a ∈ I}. The addition in R/I is given by (r + I) + (s + I) = (r + s) + I. Now let us define a multiplication on R/I, namely (r + I)(s + I) = rs + I. Note that if r + I = r′ + I and s + I = s′ + I, then r − r′ and s − s′ are in I, so rs − r′ s′ = (r − r′ )s + r′ (s − s′ ) ∈ I. Thus, the above multiplication is well defined, it is easy to see that R/I is a ring. Hence, we have the next theorem. 54 2. Rings and Fields Theorem 2.2.3. [Quotient Ring] Let R be a ring and I an ideal of R. Then the operators (r + I) + (s + I) = (r + s) + I and (r + I)(s + I) = rs + I make R/I into a ring with unity 1 + I, called the factor or quotient ring of R by I. The map ϕ : R → R/I defined by ϕ(r) = r + I is an onto ring homomorphism which has kernel I. It is called the canonical projection of R onto R/I. There also are three isomorphism theorems for rings. Their proofs are similar to isomorphism theorems for groups. Hence, we shall just sketch them. Theorem 2.2.4. [First Isomorphism Theorem] If ϕ : R → S is an onto ring homomorphism, then R/ ker ϕ ∼ = im S. Proof. Define ϕ̄ : R/ ker ϕ → S by ϕ̄(r + ker ϕ)ϕ(r) for all r ∈ R. Clearly, ϕ̄ is onto and it is easy to check that ϕ̄ is a ring homomorphism. Moreover, for r, s ∈ R, we have ϕ(r) = ϕ(s) ⇔ ϕ(r − s) = 0 ⇔ r − s ∈ ker ϕ ⇔ r + ker ϕ = s + ker ϕ. Hence, ϕ̄ is an isomorphism. Theorem 2.2.5. [Second Isomorphism Theorem] If S is a subring and I is an ideal of R, then S/(S ∩ I) ∼ = (S + I)/I. Proof. Define ϕ : S → (S + I)/I by ϕ(s) = s + I for all s ∈ S. It is easy to verify that ϕ is a ring homomorphism with kernel S ∩ I and the theorem follows from the first isomorphism theorem. Theorem 2.2.6. [Third Isomorphism Theorem] If I and J are ideals of a ring R such that I ⊆ J, then J/I is an ideal of R/I and (R/I)/(J/I) ∼ = R/J. Proof. Define ϕ : R/I → R/J by ϕ(r + I) = r + J for all r ∈ R. It can be verified that ϕ is a ring homomorphism with kernel J/I and the theorem follows from the first isomorphism theorem. Remark. As for groups, the third isomorphism theorem gives a 1-1 correspondence between the set of ideals of R containing I and the set of ideals of R/I. We end this section by embedding an integral domain into a field. We say that a ring R can be embedded in a ring R′ if there exists a monomorphism (i.e., an injective homomorphism) of R into R′ . Example 2.2.4. A ring R can be embedded in the ring Mn (R) by the diagonal map a 7→ aIn . Theorem 2.2.7. Any ring R without identity can be embedded in a ring R′ with identity. Moreover, R′ can be chosen to be either of characteristic zero or of same characteristic as R. Proof. Consider the rings R × Z and R × Zn defined in Exercises 2.1. They are rings with unity (0, 1) and (0, 1̄), and of characteristic 0 and n, respectively. If char R = 0, we define ϕ : R → R × Z by ϕ(x) = (x, 0) and if char R = n, we define ϕ : R → R × Zn by ϕ(x) = (x, 0̄). It is easy to show that both functions are monomorphisms. This finishes the proof. 55 2.2. Ideals, Quotient Rings and the Field of Fractions We now wish to show that every integral domain can be embedded in a field, called its field of fractions such that every element of the field is a fraction a/b where a and b lie in the integral domain and b 6= 0. There is only one problem to overcome: we might wish to define the field to be the set of all “fraction” a/b, with b 6= 0. But this is not quite right because two different fractions may be the same number. E.g., 1/2 = 2/4 = 3/6. We overcome this problem by defining an equivalence relation on certain pairs of elements in the integral domain. The results are presented in the next theorem. Its proof is routine and omitted. Theorem 2.2.8. [Field of Fractions] Suppose D is an integral domain, and let S be the set of pairs {(r, s) : r, s ∈ D and s 6= 0}. 1. (r, s) ∼ (r′ , s′ ) ⇔ rs′ = r′ s defines an equivalence relation on S. 2. Let [r, s] denote the equivalence class of (r, s) and let Q(D) denote the set of all equivalence classes. Then [r, s] + [r′ , s′ ] = [rs′ + r′ s, ss′ ] and [r, s][r′ , s′ ] = [rr′ , ss′ ] are well defined binary operations on Q(D). 3. The set Q(D) is a field with these operations and D is embedded in Q(D) by the monomorphism r 7→ [r, 1]. The field Q(D) is called the field of fractions or quotient field of D. The equivalence class [r, s] is denoted by r/s. Remark. If R is an entire ring which is not commutative, the construction Q(R) above does not exist in general. Example 2.2.5. Let D be an integral domain and a, b ∈ D. If am = bm and an = bn , for m and n relatively prime positive integers, prove that a = b. Proof. If a = 0, then b = 0 since D has no zero divisor. Assume that a 6= 0. Then b 6= 0. Let F be the field of fractions of D. Since (m, n) = 1, ∃x, y ∈ Z, mx + ny = 1. Thus, in F , we have a = a1 = amx+ny = (am )x (an )y = (bm )x (bn )y = bmx+ny = b1 = b, so a = b in D. Exercises 2.2. 1. An element a of a ring R is nilpotent if an = 0 for some n ∈ N. Show that the set of all nilpotent elements N in a commutative ring R is an ideal, called the nilradical of R. Moreover, prove that R/N has no nonzero nilpotent. 2. Show that a ring R has no nonzero nilpotent element if and only if 0 is the only solution of x2 = 0 in R. 3. Let ϕ : R → S be a homomorphisms of rings. Prove the following statements. (a) If I is an ideal of R and ϕ is onto, then ϕ(I) is an ideal of S. (b) If J is an ideal of S, then ϕ−1 (J) is an ideal of R containing ker ϕ. 4. Let R be a commutative ring and I an ideal of R. Show that √ I = {x ∈ R : ∃n ∈ N, xn ∈ I} is anpideal of R which contains I, called the radical of I. In addition, prove that √ √ √ (a) I= I and (b) if I = R, then I = R. 5. Let R and S be rings and ϕ : R → S be such that (i) ∀r, s ∈ R, ϕ(r + s) = ϕ(r) + ϕ(s) and (ii) ∀r, s ∈ R [ϕ(rs) = ϕ(r)ϕ(s) ∨ ϕ(rs) = ϕ(s)ϕ(r)]. Prove that ∀r, s ∈ R, ϕ(rs) = ϕ(r)ϕ(s) or ∀r, s ∈ R, ϕ(rs) = ϕ(s)ϕ(r). 56 2. Rings and Fields 6. [Chinese Remainder Theorem] If I and J are ideals of a ring R such that I + J = R, prove that R/(I ∩ J) ∼ = R/I × R/J. 7. Let R be a division ring. Prove that any nonzero ring homomorphism ϕ : R → R is 1-1. 8. Let I be an ideal of a ring R and let Mn (I) be the set of n × n matrices with entries in I. Prove that (a) Mn (I) is an ideal of Mn (R) and Mn (R)/Mn (I) ∼ = Mn (R/I), and (b) every ideal of Mn (R) is of the form Mn (I) for some ideal I of R. In particular, if R is a division ring, then the ring Mn (R) has only two ideals. Project 10. Let n ∈ N and n ≥ 2. Define Zn [i] = {a + ib : a, b ∈ Zn } where i2 ≡ −1 (mod n). (a) Prove that Zn [i] is a ring containing Zn as a subring. (b) Determine all units, zero divisors and nilpotent elements in Zn . (c) Determine all units, zero divisors and nilpotent elements in Zn [i]. 2.3 Maximal Ideals and Prime Ideals We have learned that a ring R has two trivial ideals, namely {0} and R itself. In this section, we shall discover properties of maximal ideals and prime ideals. These are two kinds of important ideals in commutative algebra and algebraic geometry. An ideal M of R is maximal if M 6= R and for every ideal J of R, M ⊆ J ⊆ R ⇒ J = M or J = R. Example 2.3.1. In the ring Z, for n ∈ N, nZ is maximal if and only if n is a prime. Proof. Let n be a prime and let J be an ideal of Z such that nZ ⊆ J ⊆ Z. Then J = dZ for some d ∈ N and d | n, so d = 1 or d = n. Hence, J = nZ or J = Z. On the other hand, assume that n = ab for some 1 < a, b < n. Then nZ ⊆ aZ ⊆ Z, aZ 6= nZ and aZ 6= Z, so nZ is not maximal. Remarks. 1. Every ideal I 6= R is contained in some maximal ideal M . Proof. Let I = {J : J 6= R and J is an ideal of R containing I}. Let C = {Jα }α∈Λ be a chain in I . Then ∪C is an ideal of R. If ∪C = R, then 1 ∈ Jα for some α ∈ Λ, so Jα = R, a contradiction. Hence, ∪C is an upper bound of C in I . By, Zorn’s lemma, we have I has a maximal element which turns out to be our desired maximal ideal containing I. 2. If M is a maximal ideal and I is an ideal of R such that I * M , then M + I = R. Proof. Let x ∈ I, ∈ / M . Consider the ideal J = M + Rx which is larger than M . Since M is maximal, J = R. Thus, R = M + Rx ⊆ M + I. 3. If M1 and M2 are distinct maximal ideals, then M1 + M2 = R. In addition, if R is commutative, then M1 M2 = M1 ∩ M2 . 4. If R is commutative, then Ru = R if and only if u is a unit. Theorem 2.3.1. Let R be a commutative ring and M an ideal of R. Then M is a maximal ideal of R if and only if R/M is a field. Proof. Clearly, R/M is a commutative ring with unity 1 + M . Assume that M is a maximal ideal. Let a ∈ / M . Then M + Ra = R, so ∃b ∈ R, 1 = m + ba. Thus, 1 + M = ba + M = (b + M )(a + M ), and hence R/M is a field. Conversely, suppose that R/M is a field. Let M ⊆ J ⊆ R and J 6= M . Then ∃a ∈ J r M . Since R/M is a field and a ∈ / M , ∃b ∈ R, 1 + M = (a + M )(b + M ) = ab + M , so 1 − ab ∈ M ⊆ J. Since a ∈ J, ab ∈ J which implies 1 ∈ J. Hence, J = R. 57 2.3. Maximal Ideals and Prime Ideals An ideal P of R is prime of R if P 6= R and for any ideals A, B of R, AB ⊆ P ⇒ A ⊆ P or B ⊆ P. Theorem 2.3.2. Let P be an ideal of R such that P 6= R. 1. If ∀a, b ∈ R, ab ∈ P ⇒ a ∈ P or b ∈ P , then P is prime. 2. If R is commutative and P is prime, then ∀a, b ∈ R, ab ∈ P ⇒ a ∈ P or b ∈ P . Proof. (1) Assume that AB ⊆ P and A * P . Choose a ∈ A, ∈ / P . Let b ∈ B. Thus, ab ∈ AB ⊆ P , so a ∈ P or b ∈ P . But a ∈ / P , hence B ⊆ P . (2) Let a, b ∈ R be such that ab ∈ P . Since R is commutative, Rab = RaRb ⊆ P , so Ra ⊆ P or Rb ⊆ P . Hence, a ∈ P or b ∈ P . Theorem 2.3.3. Let R be a commutative ring and P an ideal of R. Then P is a prime ideal of R if and only if R/P is an integral domain. Proof. This follows from Theorem 2.3.2 as follows. For an ideal P , P is prime ⇔ ∀a, b ∈ R, ab ∈ P ⇒ a ∈ P or b ∈ P ⇔ ∀a, b ∈ R, (a + P )(b + P ) = 0 + P ⇒ a + P = 0 + P or b + P = 0 + P ⇔ R/P is an integral domain as desired. Theorems 2.3.1 and 2.3.3 are the most useful for characterizing maximal ideals and prime ideals in commutative rings. Corollary 2.3.4. Let R be a commutative ring. 1. Every maximal ideal of R is prime. 2. If R is finite, then every prime ideal of R is maximal. Example 2.3.2. In Z, nZ is prime if and only if n = 0 or n is a prime. Remark. In the ring Z, {0} is a prime ideal which is not maximal. The set of all prime ideals of a commutative ring R is denoted by Spec R, called the spectrum of R. E.g., Spec Z = {pZ : p is a prime} ∪ {{0}}. A local ring is a commutative ring which has a unique maximal ideal. Examples 2.3.3. 1. Z has infinitely many maximal ideals of the form pZ where p is a prime, so it is not a local ring. 2. Every field is a local ring with maximal ideal {0}. 3. Zpn is a local ring with the maximal ideal pZpn for all primes p and n ∈ N. Theorem 2.3.5. Let R be a commutative ring. Then R is a local ring if and only if the nonunits of R form an ideal. Proof. Assume R is a local ring with the maximal ideal M . Let a ∈ R r M . If aR 6= R, then aR is contained in some maximal ideal, so aR ⊆ M which yields a contradiction. Thus, aR = R, so a is a unit. Hence, M is the set of nonunits of R. Conversely, suppose that the nonunits of R form an ideal M of R. Clearly, M is maximal. Let M ′ be another maximal ideal of R. If ∃a ∈ M ′ r M , then a is a unit, so M ′ = R, a contradiction. Thus, M ′ ⊆ M . Since M ′ is maximal, M ′ = M . 58 2. Rings and Fields Corollary 2.3.6. In a finite local ring R, every element is either a unit or a nilpotent element. Example 2.3.4. Fix a prime p and let Zp = {m/n : m, n ∈ Z and p does not divide n}. Then Zp is a subring of Q and is local. Its unique maximal ideal is {pk/n : k, n ∈ Z and p ∤ n}. We shall show an important structure theorem for finite commutative rings in Section 4.6. It says that every finite commutative ring is a direct product of a finite number of local rings. (Corollary 4.6.7). Hence, a local ring turns out to be a core when we study a finite commutative ring. It has many applications coding theory and cryptography. Exercises 2.3. 1. Let R be a ring and I an ideal of R. Prove that the map J 7→ J/I gives a 1-1 correspondence {ideals of R containing I} ←→ {ideals of R/I}. Moreover, this correspondence carries maximal ideals to maximal ideals. 2. Prove Corollary 2.3.6 and Example 2.3.4. 3. Find all ideals, all prime ideals and all maximal ideals of (a) Z12 (b) Z2 × Z4 (c) Q × Q (d) Q × Z (e) Z × Z4 × Z5 . 4. Let R be a commutative ring. If every ideal proper of R is prime, show that R is a field. 5. Show that in a Boolean ring R, every prime ideal P 6= R is maximal. 6. Let R be a commutative ring and b ∈ R a nilpotent element. Prove that u + b is a unit for all units u in R. Project 11 (Chain ring). A ring is called a chain ring if all its ideals form a chain under inclusion. For example, Zpn , p a prime and n ∈ N, is a chain ring. Also, every field is a chain ring. Let R be a finite commutative ring. Prove that R is a chain ring if and only if R is a local ring whose maximal ideal is principal. A finite chain ring arises in algebraic number theory as quotient rings of rings of integers in number fields. It has many applications in coding theory because of the similarity with finite fields. Galois rings in Project 16 are examples for this situation. 2.4 Factorizations From elementary number theory, we know that every positive integer can be decomposed uniquely into a product of prime numbers (Theorem 1.1.5). It is the unique factorization property of the ring Z. In this section, we shall learn about factorizations in any other integral domains. 2.4.1 Irreducible Elements and Prime Elements Let R be a commutative ring and suppose that a, b ∈ R. We say that a divides b and write a | b, if there is an r ∈ R such that ra = b. This definition coincides the divisibility discussed previously in Section 1.1. Remarks. Let R be a commutative ring and a, b ∈ R. 1. a divides b ⇔ b ∈ Ra ⇔ Rb ⊆ Ra. 2. a divides 0 (R0 ⊆ Ra). 3. a ∈ R, 1 divides a (Ra ⊆ R · 1 = R). 4. a divides 1 ⇔ R = Ra ⇔ a is a unit. 5. 0 divides a ⇔ Ra ⊆ R0 ⇔ a = 0. 59 2.4. Factorizations Let R be an integral domain and suppose a, b ∈ R. We say that a and b are associates if a | b and b | a. Theorem 2.4.1. Let R be an integral domain, a, b ∈ R. The following statements are equivalent. (i) a and b are associates. (ii) Ra = Rb. (iii) a = ub for some unit u ∈ R. Proof. (i) ⇒ (iii) If a = 0, then b = 0 and (3) is clear. Suppose then that a 6= 0. Since a | b and b | a, we can write a = ub and b = va. Thus, a = ub = uva, so (uv − 1)a = 0, so uv = 1. Hence, a = ub and u is a unit of R. (iii) ⇒ (ii) If a = ub where u is a unit, then Ra = Rub = (Ru)b = Rb. (ii) ⇒ (i) If Ra = Rb, then a = rb, b = sa, so b | a and a | b. Hence, a and b are associates. Let R be an integral domain. We say that a nonzero nonunit element a in R is an irreducible element or atom if a cannot be expressed as a product a = bc where b and c are nonunits. For example, in Z, p and −p, p a prime number, are irreducible elements. Theorem 2.4.2. Let R be an integral domain and a a nonzero nonunit in R. 1. a is irreducible ⇔ (∀b, c ∈ R, a = bc ⇒ b or c is a unit). 2. If Ra is maximal, then a is irreducible. The converse holds if R is a PID. Proof. (1) It follows directly from the definition. (2) Assume that Ra is maximal. Let b, c ∈ R be such that a = bc. Then Ra ⊆ Rb ⊆ R. Since Ra is maximal, Ra = Rb or Rb = R. If Rb = R, then b is a unit. Let Ra = Rb. Then a = bu for some unit u ∈ R, so bc = bu which implies c = u is a unit since R has no zero divisor. Finally, we assume that R is a PID and a ∈ R is irreducible. Let J be an ideal of R such that Ra ⊆ J ⊆ R. Since R is a PID, J = Rb for some b in R, and so a ∈ Rb. Thus, a = cb for some c ∈ R, so b or c is a unit. Hence, Rb = R or Ra = Rb. A nonzero nonunit element p in R is a prime element if ∀a, b ∈ R, p | ab ⇒ p | a or p | b. Note that a prime number is a prime element in Z by Corollary 1.1.4 (2). Theorem 2.4.3. Let R be an integral domain. (1) For a nonzero nonunit p in R, p is prime ⇔ Rp is a prime ideal. (2) Every prime element is irreducible. The converse holds if R is a PID. Proof. (1) It follows directly from the definition and Theorem 2.3.2. (2) Let p be a prime element. Assume that p = ab for some a, b ∈ R. Then Rab = Rp, so Ra ⊆ Rp or Rb ⊆ Rp. Since Rp = Rab ⊆ (Ra ∩ Rb), Ra = Rp or Rb = Rp, so au = p or bv = p for some units u and v in R. Hence, b = u or a = v is a unit in R. Finally, suppose that R is a PID and p is irreducible. Then Rp is maximal, so it is a prime ideal. Hence, p is prime. 2.4.2 Unique Factorization Domains A unique factorization domain (UFD) is an integral domain R which satisfies: 1. Every nonzero nonunit of R is a product of atoms. 2. If a is a nonzero nonunit of R, then the expression of a as a product of atoms is unique in the following sense: “If a = a1 . . . ar = b1 . . . bs where a1 , . . . , ar , b1 , . . . , bs are atoms, then r = s and there is a reordering bi1 , . . . , bir of b1 , . . . , bs such that a1 and bi1 are associates, a2 and bi2 are associates, . . . , ar and bir are associates”. 60 2. Rings and Fields Examples 2.4.1. 1. The ring of rational integers Z is a UFD by the fundamental theorem of arithmetic (Theorem 1.1.5). Since U (Z) = {±1}, the atoms of Z are ±p where p is a prime. Note that p and −p are associates (e.g., 12 = 2 · 2 · 3 = (−2)(−3) · 2). 2. Let F be a field. Every element of F except 0 is a unit. Hence, every nonzero nonunit of F is uniquely a product of atoms (vacuously!). That is, F has no nonzero nonunits. Theorem 2.4.4. Let R be an integral domain. Then R is a UFD if and only if 1. every nonzero nonunit of R is a product of atoms and 2. every irreducible element is prime. Proof. Suppose R is a UFD. Then (1) holds, by the definition of a UFD. It remains to show that if x is irreducible, then x is prime. Suppose x | bc, and let ax = bc. Write a, b and c as products of atoms, so that a . . . a x = b1 . . . bl c1 . . . cm . | 1 {z k} | {z } | {z } a c b Since these are two factorizations of ax = bc into products of atoms and x is an atom, x must be an associate of some bi or some cj . Hence, x | b or x | c. Thus, x is prime. Conversely, suppose (1) and (2) are given. Then to show R is a UFD, it suffices to show that if a1 . . . ar = b1 . . . bs where the ai and bi are atoms, then r = s and the bi may be arranged so the ai and bi are associates for i = 1, . . . , r. The proof proceeds by induction on r. When r = 1, a1 = b1 . . . bs . Since a1 is prime, a1 divides bi for some i. Assume that a1 | b1 , and let b1 = ua1 . Since b1 is an atom, u must be a unit, so a1 and b1 are associates. Furthermore, a1 = b1 . . . bs = ua1 b2 . . . bs , so 1 = (ub2 ) . . . bs . That is, s = 1 and a1 = b1 . For the inductive step, write a1 . . . ar = b1 . . . bs . Since a1 is prime, a1 divides bi for some i. As above, let b1 = ua1 where u is a unit and a1 and b1 are associates. Then a1 . . . ar = b1 . . . bs = ua1 b2 . . . bs , so a2 . . . ar = ub2 . . . bs . Now the inductive hypothesis applies since we have r − 1 factors on the left. It follows that r = s and after reordering the bi , ai and bi are associates for i = 2, . . . , r. This completes the induction. To obtain more examples of a UFD and an integral domain which is not a UFD, we introduce: Let d be a square free integer. The set √ √ Z[ d] = {x + y d : x, y ∈ Z} is a subring of C.√It is called the √ ring of quadratic integers. Note that if x1 , x2 , y1 , y2 ∈ Z are such that x1 + √ y1 d = x2 + y2 d, then x1 = x2 and y1 = y2 because d is non-square. Define a function N : Z[ d] → Z by √ N (x + y d) = x2 − dy 2 √ It is called the norm map on Z[ d]. for all x, y ∈ Z. √ √ Theorem 2.4.5.√ 1. If α ∈ Z[ d] and N (α) = 0, then α = 0 = 0 + 0 d. 2. ∀α, β ∈ √ Z[ d], N (αβ) = N (α)N (β) and (α | β ⇒ N (α) | N (β)). 3. ∀α ∈ Z[ √d], α is a unit ⇔ N (α) = ±1. √ 4. If α ∈ Z[ d] and N (α) = p is a prime number, then α is irreducible in Z[ d]. 61 2.4. Factorizations Proof. Let x, y ∈ Z be such that x2 − dy 2 = 0. Then x2 = dy 2 . If y 6= 0, then d = x2 /y 2 , so √ d = |x/y| ∈ Q, which is a contradiction. Thus, we must have y√= 0 which also forces x = 0. This proves (1). A direct calculation gives (2). For (3), let α ∈ Z[ d]. Suppose that α is a unit. √ Then αβ = 1 for some β ∈ Z[ d]. Thus, 1 = N (1) = N (αβ) = N (α)N (β), so N (α)√divides 1 in Z. This gives √ N (1) = ±1. Conversely, √ assume √ that N (α) = ±1. Write α√= x + y d. Then ±1 = N (x + y d) = x2 − y 2 d = (x + y d)(x − y d) which implies that x + y d is a unit. Finally, (4) follows from (3). Example 2.4.2. The unit group of the ring Z[i] is {1, −1, i, −i} where i denotes √ −1. √ Remark. The equation x2 − dy 2 = 1 is called the Pell’s equation. Every unit in Z[ d] is a solution of Pell’s equation, or else of x2 − dy 2 = −1, the negative Pell’s equation. If d < 0, then x2 − dy 2 ≥ 0. In this case the negative Pell’s equation has no solutions. In fact, Pell’s equation only has very few solutions in this case, namely two, unless d = −1 when there are four solutions. If d > 0, there are infinitely many solutions to Pell’s equation. The negative Pell’s equation may or may not have solutions. √ Example 2.4.3. Consider the ring Z[ −5]. √ √ 1. 1 − √−5, 1 + −5, 2 and 3 are irreducible elements. √ 2. 1 + −5 and 2 are not prime elements. Hence, Z[ −5] is not a UFD by Theorem 2.4.4. √ √ √ Solution. (1) Assume that 1 − −5 = (a + b −5)(c + d −5) for some a, b, c, d ∈ Z. By taking norms, we have 6 = (a2 + 5b2 )(c2 + 5d2 ), √ a unit. which implies that a2 + 5b2 = 1, 2, 3 or 6. Observe that b = 0 implies a2 = 1, so a + b 5 is√ 2 ≥ 5, so a2 + 5b2 = 6. This forces that c2 + 5d2 = 1 and thus c + d −5 is If b 6= 0, then a2 + 5b √ √ √ a unit. Hence, 1 − −5 is irreducible. Next, assume that 2 = (a + b −5)(c + d −5) for some a, b, c, d ∈ Z. By taking norms, we have 4 = (a2 + 5b2 )(c2 + 5d2 ), which implies that a2 + 5b2 = 1, 2 or 4. If a2 + 5b2 = 2, then 2√is a square √ modulo 5 which is a 2 + 5d2 = 1. Hence, a + b −5 or c + d −5 is a unit and so contradiction. Thus, a2 + 5b2 = 1 or c √ 2 is irreducible. Similarly, 1 + −5 and 3 are irreducible. (2) Note that √ √ (1 + −5)(1 − −5) = 6 = 2 · 3. √ √ √ Then (1 + √−5) | 2 · 3. But, if (1 + −5) | 2 or (1 + −5) | 3,√then 6 | 4√or 6 | 9, which are√absurd. Thus, 1 √ + −5 is not a prime element. Similarly, 2 | (1 + −5)(1 − −5). If 2 | (1 + −5) or 2 | (1 − −5), then 4 | 6, a contradiction. Hence, 2 is not a prime element. Next, we talk about common factors, gcd and lcm of elements in an integral domain (cf. Section 1.1). Let R be an integral domain and suppose a, b ∈ R. A greatest common divisor of a and b, gcd(a, b), is an element d ∈ R which satisfies 1. d | a and d | b and 2. ∀c ∈ R, (c | a ∧ c | b) ⇒ c | d. A least common multiple of a and b, lcm(a, b), is an element m ∈ R which satisfies 1. a | m and b | m and 2. ∀c ∈ R, (a | c ∧ b | c) ⇒ m | c. Remark. +3 and −3 are greatest common divisors of 12 and 15. 60 and −60 are least common multiples of 12 and 15. Thus, the gcd or lcm of two elements is not unique, (however we adopt the above notation anyway, e.g., gcd(12, 15) = 3 and gcd(12, 15) = −3 are both correct!). By their definitions, they are unique up to associates as recorded in the next theorem. 62 2. Rings and Fields Theorem 2.4.6. Let R be an integral domain and let a, b ∈ R. 1. If d and d′ are gcd’s of a and b, then d and d′ are associates. 2. If m and m′ are lcm’s of a and b, then m and m′ are associates. Let R be an integral domain and let Q(R) be the set of atoms of R. Define an equivalence relation on Q(R) by a ∼ b if a and b are associates. Then a set of representative atoms for R is a set P = P(R) which contains exactly one atom from each equivalence class. Example 2.4.4. Q(Z) = {±p | p is a prime} is the set of all atoms in Z. P(Z) = {p | p is a positive prime} is a set of representative atoms. P(Z) = {+2, −3, +5, −7, . . .} is another set of representative atoms. We obtain the next theorem directly from the definition of a UFD. Theorem 2.4.7. Let R be an integral domain and let P be a set of representative atoms for R. Then the following statements are equivalent. (i) R is a UFD. (ii) Every nonzero element of R can be expressed uniquely (up to order of factors) as a = ubi11 · · · bikk , where u is a unit of R, k ≥ 0, i1 , . . . , ik > 0 and b1 , . . . , bk are distinct elements of P. Another important result from R being a UFD is the existence of gcd and lcm for any pair of nonzero elements. We also have the same relation for gcd and lcm as in elementary number theory. Theorem 2.4.8. Let R be a UFD and suppose a, b ∈ R r {0}. 1. a and b have a gcd and an lcm. 2. Let P be a set of representative atoms for R. Then among the gcd’s of a and b there is exactly one which is a product of elements of P. The same is true for the lcm’s of a and b. 3. If a and b are nonzero, gcd(a, b) = r, and lcm(a, b) = s, then ab and rs are associates. In other words, ab . lcm(a, b) = gcd(a, b) Proof. Let P be a set of representative atoms of R, and let b1 , . . . , bk ∈ P be all the atoms of P which occur in either a or b when they are factored as in Theorem 2.4.7. Write a = ubi11 · · · bikk and b = vbj11 · · · bjkk where u and v are units and is , js ≥ 0. Then we derive: min(i ,j ) min(i ,j ) (a) r = b1 1 1 · · · br r r is a gcd for a and b. max(i1 ,j1 ) max(ir ,jr ) (b) s = b1 · · · br is a lcm for a and b. (c) r is the only gcd of a and b which is a product of elements of P, and s is the only lcm of a and b which is a product of elements of P. (d) Since i + j = min(i, j) + max(i, j) for any integers i and j, ab = uvb1i1 +j1 · · · bikk +jk = uvrs. Hence, ab and rs are associates. Remark. Suppose R is an integral domain and Ra + Rb = Rc. Then c = gcd(a, b). The converse does not hold. E.g., Q[s, t], where s and t are indeterminates. Then gcd(s, t) = 1 and Q[s, t] 6= Q[s, t]s + Q[s, t]t. 63 2.4. Factorizations Proof. Since Rc ⊇ Ra and Rc ⊇ Rb, c | a and c | b. Suppose d | a and d | b. Then Rd ⊇ Ra + Rb = Rc, so d | c. Hence, c = gcd(a, b). Now, we shall prove that a PID is also a UFD. The following lemma is a key for R being a PID. Lemma 2.4.9. [Ascending Chain Condition (ACC) for a PIR] Let R be a principal ideal ring. If I1 ⊆ I2 ⊆ . . . is a chain of ideals in R, then ∃m ∈ N, In = Im for all n ≥ m. S Proof. Let I = ∞ n=1 In . Then I is an ideal of R. Since R is a PIR, ∃a ∈ R, (a) = I. Then S∞ a ∈ n=1 In , so ∃m ∈ N, a ∈ Im . Thus, I = (a) ⊆ Im ⊆ I which implies that Im = I. Hence, ∀n ≥ m, In = Im . Lemma 2.4.10. If R is a PID and a is a nonzero nonunit element in R, then there exists an atom p ∈ R such that p | a. Proof. Since a is nonunit, Ra $ R. Then there exists a maximal ideal M of R such that Ra ⊆ M . Since R is a PID, M = Rp for some atom p by Theorem 2.4.2. Since Ra ⊆ Rp, p | a. Theorem 2.4.11. Every PID is a UFD. Proof. Let R be a PID. By Theorems 2.4.3 and 2.4.4, it suffices to show that every nonzero nonunit of R is a product of atoms. Let a ∈ R be nonzero nonunit. By Lemma 2.4.10, there exists an atom p1 dividing a. Write a = p1 b1 for some b1 ∈ R. If b1 is a unit, then a is an atom. If b1 is nonunit, then there exists an atom p2 dividing b1 , so we write a = p1 b1 = p1 p2 b2 . Continuing, we get a strictly ascending chain of ideals (a) ⊂ (b1 ) ⊂ (b2 ) ⊂ · · · . Since R is a PID, this chain must terminate, by the ACC in Lemma 2.4.9, with some br = pr ur where ur is a unit and pr is an atom. Hence, a = p1 p2 . . . pr ur , and so R is a UFD as desired. Finally, we study a generalization of the division algorithm which leads to a special kind of integral domains. An integral domain D is called a Euclidean domain if there exists a map d : D r {0} → N ∪ {0}, called a valuation map, such that 1. ∀a, b ∈ D r {0}, d(a) ≤ d(ab) and 2. ∀a ∈ D, b ∈ D r {0}, ∃q, r ∈ D, a = bq + r with r = 0 or d(r) < d(b). Examples 2.4.5. 1. Any field F is a Euclidean domain with valuation d(a) = 1 for all a 6= 0. 2. From the division algorithm for Z (Theorem 1.1.1), we have Z is a Euclidean domain if we define d(a) = |a| for all a 6= 0. 3. The ring Z[i] = {m + ni : m, n ∈ Z} is called the ring of Gaussian integer. This is a subring of C, hence an integral domain. Its elements can be identified with the set of “lattice points”, that is, points with integral coordinates in the complex plane. If a = m + ni, we put d(a) = aā = |a|2 = m2 + n2 , the norm map. Then d(a) ∈ N and d(ab) = d(a)d(b) ≥ d(a) for all a, b ∈ Z[i] r {0}. To prove that d satisfies the condition of the definition of a Euclidean domain, we note that if b 6= 0, then ab−1 = µ+νi, where µ and ν are rational numbers. Now we can find integers u and v such that |u − µ| ≤ 1/2, |v − ν| ≤ 1/2. Set ε = µ − u, η = ν − v, so that |ε| ≤ 1/2 and |η| ≤ 1/2. Then a = b[(u + ε) + (v + η)i] = bq + r where q = u + vi is in Z[i]. Since r = a − bq, r ∈ Z[i]. Moreover if r 6= 0, then d(r) = |r|2 = |b|2 (ε2 + η 2 ) ≤ |b|2 (1/4 + 1/4) = d(b)/2. Thus, d(r) < d(b). Hence, Z[i] is a Euclidean domain. 64 2. Rings and Fields Theorem 2.4.12. A Euclidean domian is a PID, and hence is a UFD. Proof. Let I be an ideal in a Euclidean domain D. If I = {0}, we have I = (0). Otherwise, let b 6= 0 be an element of I for which d(b) is minimal for the nonzero elements of I. Let a be any element of I. Then a = bq + r for some q, r ∈ D with r = 0 or d(r) < d(b). Since r = a − bq ∈ I and d(r) < d(b), we must have r = 0 by the choice of b in I. Hence, a = bq, so I = (b). √ Example 2.4.6. Let θ = 21 (1 + −19) and Z[θ] = {a + bθ : a, b ∈ Z}. Assume that u = a + bθ is a unit in Z[θ]. Then (a + bθ)(c + dθ) = 1 for some c, d ∈ Z. The squares of absolute value on both sides give ((2a + b)2 + 19b2 )((2c + d)2 + 19d2 ) = 16 which implies b = d = 0 and so ac = 1. Hence, the unit group of Z[θ] = {±1}. By a similar technique, we can show that 2 and 3 are irreducible in Z[θ]. Now, suppose that d is a valuation map on Z[θ]. Choose m ∈ Z[θ] which is nonzero nonunit such that d(m) is minimal. First, we divide 2 by m and get q, r ∈ Z[θ] and 2 = mq + r with d(r) < d(m) or r = 0. This means r = 0, 1 or −1. If r = 0, then m | 2 which forces m = ±2 since 2 is irreducible and m is not a unit. Similarly, if r = −1, then m = ±3. The case r = 1 cannot happen, for if it did, then m | 1, so m is a unit. Next, we divide θ by m in the same way, we obtain q ′ , r′ ∈ Z[θ] and θ = mq ′ + r′ with d(r′ ) < d(m) or r′ = 0. Again, we have r′ = 0, 1 or −1. Thus, one of θ, θ + 1 or θ − 1 is divisible by m. But m = ±2 or ±3 and it is easy to see that none of these quotients is in R. This contradiction tell us that Z[θ] is not a Euclidean domain. Next, we shall show that Z[θ] is a PID. Let I be a nonzero ideal of Z[θ]. Choose b ∈ I so that |b| is as small as possible. We aim to show that I = Z[θ]b. Suppose not. Then there is an element a ∈ I r Z[θ]b. Note that ap − bq ∈ I for all p, q ∈ Z[θ], so if we can find p, q with |ap − bq| < |b| (or equivalently |(a/b)p − q| < 1), then we shall be done. Since we may replace a by any element a′ = a − bq, we can subtract any desired element of R from √ a/b. In particular, we can assume that the imaginary part y of √ a/b = x + iy lies between ± 19/4. Now, if the imaginary part of a/b lies strictly between ± 3/2, then a/b lies at distance less than 1 from some rational integer and we are done. Thus, we may assume the imaginary part of a/b lies between √ √ 3/2 and 19/4 (or the argument is similar). √ the √ √ Hence, √ the imaginary √ √ negative of this, where part of 2(a/b) − (1 + −19)/2 lies between 3 − 19/2 and 0. But 19 < 27 = 3 3, so √ √ √ √ 3/2 > 19/2 − 3 > 0. Therefore, the imaginary part of 2(a/b) − (1 + −19)/2 is sufficient small that the complex number lies at a distance less than 1 for some rational integer. In both cases, we have found elements p, q ∈ Z[θ] such that |ap−bq| < |b| which is a contradiction. Hence, Z[θ] is a PID. √ Remark. In conclusion, recall that Z is an integral domain which is not a field and Z[ −5] is not √ a UFD. Besides, Z[θ] = {a + bθ : a, b ∈ Z}, where θ = (1 + −19)/2, is a PID which is not a Euclidean domain as shown above. Finally, Z[x] (in the next section) is a UFD which is not a PID. Exercises 2.4. 1. If p and q are prime elements in an integral domain R such that p | q, prove that p and q are associates. 2. Let R be a UFD and c a nonzero element in R. Prove that R/Rc contains a nonzero nilpotent element if and only if there is a prime element p ∈ R with p2 | c. 3. Let R be a UFD. If a ∈ R is a nonzero nonunit element, prove that Ra is the product of a finite number of prime ideals. 65 2.4. Factorizations 4. If R is a PID and gcd(a, b) = 1, show that Ra + Rb = R, so 1 = ax + by for some x, y ∈ R. 5. Let R be a PID and suppose that a, b and c are nonzero elements of R such that Ra + Rb = Rc. Show that there exist√u, v ∈ R such that ua + vb = c and Ru + Rv = R. √ √ 6. Prove that 4 + 10 is irreducible but not prime in the ring {a + b 10 : a, b ∈ Z}. Deduce that Z[ 10] is not a UFD. √ 7. Show that the ring Z[ 2] has infinitely many units. (Hint. If u is a unit, so is un for all n ∈ Z.) 8. (a) Let D be a Euclidean domain. Prove that u is a unit in D if and only if d(u) = d(1). (b) Show that ±1 and ±i are units in Z[i] and prove that if a+bi is not a unit in Z[i], then a2 +b2 > 1. 9. Let R be a Euclidean ring and a, b ∈ R, b 6= 0. Prove that there exist q0 , q1 . . . , qn and r1 , . . . , rn in R such that a = q0 b + r1 , b = q1 r1 + r2 , r1 = q2 r2 + r3 , d(r1 ) < d(b), d(r2 ) < d(r1 ), d(r3 ) < d(r2 ), ··· ··· ··· rn−2 = qn−1 rn−1 + rn , rn−1 = qn rn d(rn ) < d(rn−1 ), and if a and b satisfy the above conditions, then rn is a gcd of a and b. This algorithm is called the Euclidean algorithm. Find a gcd of 8 + 6i and 5 − 15i in Z[i] by using the Euclidean algorithm. 10. Let D be a UFD with field of fractions F and suppose α ∈ F . Show that it is possible to write α = a/b with a, b ∈ D and gcd(a, b) = 1. 11. Let R be a PID with field of fractions F , and let S be a ring with R ⊆ S ⊆ F . (a) If α ∈ S, show that α = a/b with a, b ∈ R and 1/b ∈ S. (b) Prove that S is a PID. 12. Let R = {m/2n : m, n ∈ Z and n ≥ 0}. (a) Prove that R is a subring of Q and determine all units of R. (b) Show that 3 is an irreducible element in R. (c) Prove that R is a PID. Project 12 (Prime elements in the ring of Gaussian integers). We have learned that all units in Z[i] are ±1, ±i. In this project, we shall determine all prime elements in Z[i]. Use the norm map, show that up to multiplication by units, the prime elements in Z[i] are of three types: (a) p, where p is a prime in Z satisfying p ≡ 3 (mod 4), (b) π or π̄, where q = ππ̄ is a prime in Z satisfying q ≡ 1 (mod 4), (c) α = 1 + i. Project 13 (Quadratic √ norm Euclidean domains). Find all square free integers d ≡ 2, 3 (mod 4) such that the norm map on Z[ d] satisfies the axiom of a Euclidean function. [Answer. They are −2, −1, 2, 3, 6, 7, 11, 19, 33.] Moreover, for a square free integer d ≡ 1 (mod 4), let ) " √ √ # ( x+y d 1+ d = : x, y ∈ Z, a ≡ b (mod 2) . Z 2 2 Define the norm map on Z h √ i 1+ d 2 by N √ ! x2 − dy 2 x+y d = . 2 4 Find all square free integers d ≡ 1 (mod 4) such that the norm map on Z h √ i 1+ d 2 satisfies the axiom of a h √ i Euclidean function. [Answer. They are −11, −7, −3, 5, 13, 17, 21, 29, 37, 41, 57, 73. Note that Z 1+2 69 is a Euclidean domain but not for norm.] 66 2.5 2. Rings and Fields Polynomial Rings One of familiar topic in elementary algebra is “polynomials”. Algebraic equations usually involve factorization of polynomials. Here, we treat them in a more abstract way with the things that we have studied from the previous section. There will be many important results in this section. 2.5.1 Polynomials and Their Roots Let R be a ring with identity 1 and let x be a symbol called an indeterminate, not representing any element in R. Let R[x] denote the set of all symbols a0 + a1 x + · · · + an xn where n ∈ N ∪ {0}, ai ∈ R, x0 = 1, x1 = x. For i ∈ N, let xi denote 1 · xi . In the symbol a0 + a1 x + · · · + an xn , we may drop ai xi if ai = 0. Each element a0 + a1 x + · · · + an xn is called a polynomial and ai is called the coefficient of xi for i ∈ {1, . . . , n} and a0 is called the constant term. For p(x) = a0 + a1 x + · · · + an xn and q(x) = b0 + b1 x + · · · + bm xm in R[x], we can write p(x) = a0 + a1 x + · · · + ak xk and q(x) = b0 + b1 x + · · · + bk xk where k ≥ max{m, n}, ai = 0 if i > n and bj = 0 if j > m and we define 1. p(x) = q(x) ⇔ ai = bi for all i ∈ {0, 1, . . . , k} k 2. p(x) + q(x) = (a0 + b0 ) + (a1 + b1 )x + · · · + (ak + bP k )x and l 3. p(x)q(x) = c0 + c1 x + · · · + cn+m xn+m , where cl = i=0 ai bl−i (= a0 bl + a1 bl−1 + · · · + al b0 ) for all l ∈ {0, 1, . . . , n + m}. Hence, under the operation defined above R[x] is a ring which has 1 as its identity and contains R as a subring (considered elements as constant polynomials). The ring R[x] is called the ring of polynomials over R. If R is commutative, so is R[x]. Set R[x1 , x2 ] = R[x1 ][x2 ] and R[x1 , . . . , xn ] = R[x1 , . . . , xn−1 ][xn ] if n > 2. If p(x) = a0 + a1 x + · · · + an xn ∈ R[x] and an 6= 0, then the degree of p(x), denoted by deg p(x), is defined to be n. For technical reasons, we define the degree of the zero polynomial to be −∞ and adopt the following conventions: (−∞) < n and (−∞) + n = −∞ = n + (−∞) for every integer n; (−∞) + (−∞) = −∞. Let f (x) = a0 + a1 x + · · · + an xn . If an 6= 0, an is called the leading coefficient of f (x) and f (x) is a monic polynomial if an = 1. If R is commutative and c ∈ R, then f (x) 7→ f (c) := a0 + a1 c + · · · + an cn gives a homomorphism from R[x] to R, called the evaluation at c. In addition, if f (c) = 0, then c is called a root of f (x). The following statements are clearly true. 1. Every unit in R is a unit in R[x]. 2. If f (x) = a0 + a1 x + · · · + am xm , g(x) = b0 + b1 x + · · · + bn xn ∈ R[x] and am bn 6= 0, then deg(f (x)g(x)) = m + n. In particular, if R is an integral domain, we have: Theorem 2.5.1. Let R be an integral domain. 1. R[x] is an integral domain. 2. ∀f (x), g(x) ∈ R[x] r {0}, deg f (x)g(x) = deg f (x) + deg g(x). 3. The set of all units of R[x] is the set of all units of R. In particular, U (Z[x]) = {±1} and U (F [x]) = F r {0}, where F is a field. 4. ∀a ∈ R, a is irreducible in R ⇔ a is irreducible in R[x]. 5. ∀a, b ∈ R, b is a unit ⇒ a + bx is irreducible in R[x]. Proof. (1) and (2) are clear from the above observation. Note that if f (x) is a unit, then f (x)g(x) = 1 for some g(x) ∈ R[x], so deg f (x) + deg g(x) = deg 1 = 0 by (2). This forces that deg f (x) = 0 = deg g(x) which implies that f (x) lies in R and (3) follows. Next, let a ∈ R. If a is irreducible in R[x], then a is clearly irreducible in R. On the other hand, if a = f (x)g(x) for some 67 2.5. Polynomial Rings nonzero nonunits f (x) and g(x) in R[x], we have 0 = deg a = deg f (x) + deg g(x), so this again gives deg f (x) = deg g(x) = 0. This means that f (x) and g(x) indeed lie in R, and thus a is reducible in R. Finally, let a, b ∈ R with b a unit. Then deg(a + bx) = 1. If a + bx = f (x)g(x) for some f (x), g(x) ∈ R[x], then 1 = deg f (x) + deg g(x), so f (x) or g(x) must lie in R, say f (x) = c a constant in R and g(x) = u+vx. Since b = cv is a unit, c is a unit. Hence, a+bx is irreducible. Theorem 2.5.2. [Division Algorithm] Let R be a ring, f (x), g(x) ∈ R[x] and g(x) 6= 0. Assume that the leading coefficient of g(x) is a unit in R. Then ∃ unique q(x), r(x) ∈ R[x] such that f (x) = q(x)g(x) + r(x) where r(x) = 0 or deg r(x) < deg g(x). Proof. If there exists an h(x) ∈ R[x] such that f (x) = h(x)g(x), let q(x) = h(x) and r(x) = 0. Assume that f (x) 6= h(x)g(x) for all h(x) ∈ R[x]. Let S = {deg(f (x) − h(x)g(x)) : h(x) ∈ R[x]} ⊆ N ∪ {0}. Then S 6= ∅. By the Well-Ordering Principle, there exists a polynomial q(x) in R[x] such that deg(f (x) − q(x)g(x)) has the least degree and we may write r(x) for f (x) − q(x)g(x). Then r(x) 6= 0. Assume that deg r(x) ≥ deg g(x). Write r(x) = a0 + a1 x + · · · + an xn , an 6= 0, and g(x) = b0 + b1 x + · · · + bm xm with bm a unit. Since deg r(x) ≥ deg g(x), n − m ≥ 0. Let n−m g(x). Thus, deg s(x) < n and s(x) = r(x) − an b−1 m x n−m n−m )g(x), s(x) = f (x) − q(x)g(x) − an b−1 g(x) = f (x) − (q(x) − an b−1 m x m x so s(x) ∈ S and deg s(x) < deg r(x), a contradiction. To prove that q(x) and r(x) are unique, suppose that q2 (x) and r2 (x) are polynomials such that f (x) = g(x)q2 (x) + r2 (x) where r2 (x) = 0 or deg r2 (x) < deg g(x). Then g(x)q(x) + r(x) = f (x) = g(x)q2 (x) + r2 (x). Subtracting yields g(x)[q(x) − q2 (x)] = r2 (x) − r(x). Since the leading coefficient of g(x) is assumed to be a unit, we have deg(g(x)[q(x) − q2 (x)]) = deg g(x) + deg(q(x) − q2 (x)). Since deg(r2 (x) − r(x)) < deg g(x), this relation can hold only if q(x) − q2 (x) is zero, i.e., q(x) = q2 (x), and hence finally r(x) = r2 (x). For a field R, the leading coefficient of a nonzero polynomial g(x) in R[x] is always a unit in R, so the division algorithm above gives: Corollary 2.5.3. If F is a field, then F [x] is a Euclidean domain with valuation d(p(x)) = deg p(x) for all p(x) ∈ F [x] r {0}. Moreover, F [x] is a PID and a UFD. Theorem 2.5.4. [Remainder Theorem] Let R be a ring and f (x) ∈ R[x]. Then for all c ∈ R, the remainder when x − c divides f (x) is f (c). Proof. Let c ∈ R. By Theorem 2.5.2, there exist unique q(x) ∈ R[x] and r ∈ R such that f (x) = q(x)(x − c) + r. Then f (c) = q(c)(c − c) + r = r. Corollary 2.5.5. Let R be a ring. 1. If f (x) ∈ R[x], c ∈ R and f (c) = 0, then f (x) = q(x)(x − c) for some q(x) ∈ R[x]. 2. If R is commutative, f (x) ∈ R[x] and c ∈ R, then f (c) = 0 ⇔ (x − c) | f (x) in R[x]. 3. Let R be an integral domain, f (x) ∈ R[x], deg f (x) = 2 or 3 and the leading coefficient of f (x) is a unit in R. Then f (x) has a root in R ⇔ f (x) is reducible in R[x]. 68 2. Rings and Fields Proof. (1) and (2) are clear. For (3), assume that c is a root of f (x). Then f (x) = q(x)(x − c) for some q(x) ∈ R[x]. Since deg f (x) is 2 or 3, deg q(x) is 1 or 2, so f (x) is reducible. Conversely, suppose that f (x) = g(x)h(x), where g(x), h(x) ∈ R[x] of degree ≥ 1. Since deg f (x) = 2 or 3, deg g(x) = 1 or deg h(x) = 1. Hence, f (x) has a root in R. Examples 2.5.1. 1. x2 − 3 is irreducible over Q but not over R. 2 2. x + 1 is irreducible over R but not over C since x2 + 1 = (x − i)(x + i). 3. x3 − x + 1 is irreducible over Z3 but reducible over R by the intermediate value theorem. In general, every polynomial of odd degree over R has a root in R (Theorem 5.5.3). 4. x4 + 4 has no roots in R but it can be factored as (x2 − 2x2 + 2)(x2 + 2x2 + 2) in R[x]. Corollary 2.5.6. Let F be a field. 1. If f (x) is a polynomial over F of degree n, then f (x) has at most n roots in F . 2. If f (x) and g(x) are polynomials over F of degree ≤ n such that f (α1 ) = g(α1 ), . . . , f (αn+1 ) = g(αn+1 ) where α1 , . . . , αn+1 are distinct elements of F , then f (x) = g(x). 3. If F is infinite and f (x) and g(x) are polynomials over F such that f (α) = g(α) for all α ∈ F , then f (x) = g(x). Proof. We shall prove (1) by induction on k = deg f (x). It is clear when f (x) is linear. Assume that k > 1 and any polynomials of degree k have at most k roots in F . Suppose that f (x) is of degree k + 1. The statement is true when f (x) has no root in F . Otherwise, let α be a root of f (x) in F . Then f (x) = (x − α)q(x) for some polynomial q(x) ∈ F [x] of degree k. By the inductive hypothesis, q(x) has at most k roots. Hence, f (x) has at most k +1 roots. The remaining statements follow from the first one. Remarks. 1. f (x) = x2 − 1 has four roots in Z12 , namely 1, −1, 5, −5. 2. Corollary 2.5.6 says that two polynomials over an infinite field F which defined the same function on F are identical. This is NOT true if F is finite. Let F = Zp , f (x) = x and g(x) = xp . Then f (α) = g(α) for all α ∈ Zp but f (x) 6= g(x). Theorem 2.5.7. Let F be a field and F [x] the polynomial ring over F . Then linear polynomials are the only atoms in F [x] if and only if each polynomial f (x) ∈ F [x] of positive degree has a root in F . Proof. Suppose that linear polynomials are the only atoms in F [x]. Let f (x) be a polynomial of positive degree over F . Since F [x] is a UFD, f (x) = α1 (x) · · · αk (x), a product of atoms. Each αi (x) is linear, so αi (x) = bi (x − ci ) (bi , ci ∈ F with bi 6= 0). Then (x − ci ) | f (x), so c1 , . . . , ck are roots of f (x) in F . Conversely, assume that every f (x) ∈ F [x] of positive degree has a root in F . Let α(x) be an atom in F [x]. We claim that α(x) is linear. For, let b ∈ F be a root of α(x). Then (x − b) | α(x) so α(x) = (x − b)β(x) for some β(x) ∈ F [x]. Since α(x) is an atom, β(x) must be a unit. That is, β(x) is a constant lying in F r {0}. Thus, α(x) is a linear polynomial. A field F is an algebraically closed field if every non-constant polynomial has a root in F . Example 2.5.2. By the fundamental theorem of algebra (Theorem 5.5.6), the only atoms in C[x] are linear polynomials. Thus, C is an algebraically closed field. Theorem 2.5.8. Let R be an integral domain and f (x) ∈ R[x] a nonzero polynomial. If α1 , . . . , αk are distinct roots of f (x), then (x − α1 ) . . . (x − αk ) divides f (x). 69 2.5. Polynomial Rings Proof. We shall prove this result by induction of k. Corollary 2.5.5 (1) gives the basis step. Assume k > 1. By the inductive hypothesis (x − α1 ) . . . (x − αk−1 ) divides f (x), so let f (x) = (x − α1 ) . . . (x − αk−1 )g(x) for some g(x) ∈ R[x]. Then 0 = f (αk ) = (αk − α1 ) . . . (αk − αk−1 )g(αk ). Thus, g(αk ) = 0 since R is an integral domain, so (x−αk ) | g(x). It follows that (x−α1 ) . . . (x−αk ) divides f (x). 2.5.2 Factorizations in Polynomial Rings When factor a polynomial, we first look for some common factors on its coefficients. For example, 2x3 + 4 = 2(x3 + 2). Taking the gcd of the coefficients out allows us to concentrate on polynomials with no common factor on their coefficients and leads to the following definitions. Let R be a UFD and suppose that f (x) = a0 + a1 x + · · · + an xn is a nonzero polynomial in R[x]. The content of f (x) is the gcd of a0 , . . . , an . We say that f (x) is primitive if the content of f (x) is unit in R, i.e., a0 , . . . , an have no common factor except units. Theorem 2.5.9. [Gauss’ lemma] Let R be a UFD and f (x), g(x) ∈ R[x]. If f (x) and g(x) are primitive, so is f (x)g(x). Proof. Let f (x) = a0 + a1 x + · · · + am xm g(x) = b0 + b1 x + · · · + bn xn f (x)g(x) = c0 + c1 x + · · · + cm+n xm+n . We shall suppose that f (x)g(x) is not primitive and obtain a contradiction. Let a ∈ R be an atom of R which divides all of c0 , . . . , cm+n . Since R is a UFD, every atom is a prime, so Ra is a prime ideal. Then (R/Ra)[x] is an integral domain. Since R[x]/R[x]a ∼ = (R/Ra)[x], R[x]a is a prime ideal. Let − : R[x] → R[x]/R[x]a be the canonical map. Since a divides c0 , . . . , cm+n , f¯(x)ḡ(x) = 0. But a does not divide all of a0 , . . . , am or all of b0 , . . . , bn , since f and g are primitive. Thus, f¯(x) 6= 0, ḡ(x) 6= 0. This is a contradiction since f¯(x)ḡ(x) = 0 and f¯(x), ḡ(x) lie in R[x]/P which is an integral domain. Hence, f (x)g(x) is primitive, as claimed. Theorem 2.5.10. Let R be a UFD and f (x), g(x) nonzero polynomials of R[x]. Then: 1. f (x) is primitive ⇔ the content of f (x) is 1. 2. If a is the content of f , then f (x) = af1 (x) where f1 (x) is primitive. 3. If f (x) = af1 (x) and f1 (x) is primitive, then a is the content of f (x). 4. If a and b are the contents of f (x) and g(x), respectively, then ab is the content of f (x)g(x). Proof. (1), (2) and (3) are immediate from the definition of gcd. For the last statement, by (2), we write f (x) = af1 (x) and g(x) = bg1 (x) where f1 (x) and g1 (x) are primitive. By Gauss’ lemma, f1 (x)g1 (x) is primitive, and f (x)g(x) = af1 (x)bg1 (x) = (ab)(f1 (x)g1 (x)). Hence, ab is the content of f (x)g(x), by (3). 70 2. Rings and Fields Theorem 2.5.11. Let R be a UFD and let F = Q(R) = {r/s : r, s ∈ R, s 6= 0} be its field of quotients. Suppose f (x) is an irreducible polynomial in R[x]. Then f (x), considered as a polynomial in F [x], is irreducible in F [x]. In particular, if f (x) ∈ Z[x] is irreducible over Z, it is irreducible over Q. Proof. Suppose f (x) = g(x)h(x) where g(x) and h(x) are polynomials of positive degree in F [x]. Let g(x) = a0 /b0 + (a1 /b1 )x + · · · + (am /bm )xm and h(x) = c0 /d0 + (c1 /d1 )x + · · · + (cn /dn )xn . Let b be a least common multiple of the bi and d a least common multiple of the dj so that g1 (x) = bg(x) and h1 (x) = dh(x) lie in R[x]. Then bdf (x) = bg(x)dh(x) = g1 (x)h1 (x). By Theorem 2.5.10, let g1 (x) = ug2 (x) and h1 (x) = vh2 (x) where u is the content of g1 (x) and v is the content of h1 (x), and g2 (x) and h2 (x) are primitive polynomials in R[x]. Thus, bdf (x) = g1 (x)h1 (x) = uvg2 (x)h2 (x). Since g2 (x) and h2 (x) are primitive, so is g2 (x)h2 (x) and hence the equation above implies that bd | uv in R. Canceling, we obtain f (x) = wg2 (x)h2 (x) where w = uv ∈ R. bd Therefore, f (x) is reducible in R[x], which proves the theorem. Let R be a UFD and F = Q(R) its field of quotients. Suppose h(x) = a0 /b0 + (a1 /b1 )x + · · · + (an /bn )xn ∈ F [x], where a0 /b0 , a1 /b1 , . . . , an /bn are in “lowest terms”. That is, ai and bi have no common factor. Let b = lcm(b0 , . . . , bn ). Then bh(x) = a0 (b/b0 ) + a1 (b/b1 )x + · · · + an (b/bn )xn is in R[x]. Let a be the content of bh(x). It happens that a = gcd(a0 , . . . , an ), although knowing this is not essential. The main point is that h1 (x) = (b/a)h(x) is a primitive polynomial in R[x]. Moreover, the proof of Theorem 2.5.11 shows that if f (x) ∈ R[x], then h(x) | f (x) in F [x] ⇔ h1 (x) | f (x) in R[x]. In particular, suppose f (x) ∈ R[x], and r/s ∈ F is a root of f (x) where r and s are relatively prime. Then h(x) = x − (r/s) divides f (x) in F [x], so h1 (x) = sx − r divides f (x) in R[x]. Thus, we have: Theorem 2.5.12. Let R be a UFD and F its field of quotients. Suppose f (x) ∈ R[x] where f (x) = a0 + a1 x + · · · + an xn and r/s ∈ F is a root of f (x) where r and s are relatively prime. Then s | an and r | a0 if r 6= 0. Proof. The remarks above show that (sx − r) | (a0 + a1 x + · · · + an xn ) in R[x]. It is easy to see that this implies our results. Remarks. 1. Suppose f (x) = a0 + a1 x + · · · + an xn ∈ Z[x], where a0 6= 0. Then there are only finitely many rationals which can possibly be roots of f (x), namely the fractions r/s where r | a0 and s | an . 71 2.5. Polynomial Rings 2. Note that if an = 1 above, then s = ±1 and r/s ∈ Z. In other words, if an = 1, then every rational root of f (x) is an integer. Another important criterion on irreduciblity in Q[x] is the next theorem. Theorem 2.5.13. [Eisenstein’s Criterion] Let f (x) = a0 + a1 x + · · · + an xn lies in Z[x], and suppose that there is a prime number p such that 1. p ∤ an , 2. p | a0 , . . . , an−1 , and 3. p2 ∤ a0 . Then f (x) is irreducible in Q[x]. Moreover, if f (x) is primitive, then f (x) is irreducible in Z[x]. Proof. We shall suppose that f (x) is reducible in Q[x] and obtain a contradiction. By dividing f (x) by its content, we may assume that f (x) is primitive, this does not affect either the hypothesis or the reducibility of f (x) in Q[x]. By Theorem 2.5.11, f (x) is reducible in Z[x], so let f (x) = g(x)h(x) where g(x) = b0 + b1 x + · · · + bm xm and h(x) = c0 + c1 x + · · · + cn−m xn−m are in Z[x]. Note that since f (x) is primitive, neither g(x) nor h(x) is constant. That is, m ≥ 1 and n − m ≥ 1. Let − : Z[x] → Zp [x] be the canonical projection. Then f¯(x) = ān xn where ān 6= 0̄ since p ∤ an , so ḡ(x)h̄(x) = f¯(x) = ān xn . Since Zp [x] is a UFD, this forms ḡ(x) = b̄m xm , h̄(x) = c̄n−m xn−m , so that b̄0 = c̄0 = 0̄ (i.e., p divides b0 and c0 ). But then p2 | a0 since a0 = b0 c0 , which contradicts part (3) of the hypotheses. Hence, f (x) is irreducible in Q[x] as claimed. Example 2.5.3. f (x) = 2x5 − 6x3 + 9x2 − 15 is irreducible in Q[x] and in Z[x]. Corollary 2.5.14. The pth cyclotomic polynomial Φp (x) = xp − 1 = xp−1 + xp−2 + · · · + x + 1 x−1 is irreducible in Q[x] for any prime p. Proof. The polynomial p p−2 (x + 1)p − 1 p−1 x + ··· + p =x + g(x) = Φp (x + 1) = 1 (x + 1) − 1 satisfies the Eisenstein criterion for the prime p and is thus irreducible in Q[x]. But clearly if Φp (x) = h(x)r(x) were a nontrivial factorization of Φp (x) in Z[x], then Φp (x + 1) = g(x) = h(x + 1)r(x + 1) would give a nontrivial factorization of g(x) in Z[x]. Thus, Φp (x) must also be irreducible in Q[x]. We next wish to prove a famous theorem of Gauss: if R is a UFD, so is R[x]. Recall the criterion given in Theorem 2.4.4: an integral domain is a UFD ⇔ 1. every nonzero nonunit is a product of atoms and 2. every atom is prime. Suppose R is a UFD. We first observe that R[x] is an integral domain, so this presents no problem. We shall establish the criteria above for R[x] (and these show that R[x] is a UFD) by doing three things: (a) We determine all atoms of R[x] (Theorem 2.5.15). (b) We show that they are primes (Theorem 2.5.16). 72 2. Rings and Fields (c) We show that every nonzero nonunit of R[x] is a product of atoms and conclude that R[x] is a UFD (Theorem 2.5.17). Theorem 2.5.15. Let R be a UFD, F its field of quotients and f (x) ∈ F [x]. Then f (x) is an atom of R[x] ⇔ either (1) f (x) ∈ R and f (x) is an atom of R or (2) f (x) is a primitive polynomial of degree n ≥ 1 and f (x) is irreducible in F [x]. Proof. Assume that f (x) is an atom of R[x]. If deg f (x) = 0, then f (x) ∈ R, and clearly f (x) must be an atoms of R. Otherwise, suppose that deg f (x) = n ≥ 1, and let a be the content of f (x). Then f (x) = af1 (x) where f1 (x) is primitive. Since f (x) is irreducible in R[x], a must be a unit in R, so f (x) is primitive. Again, since f (x) is irreducible in R[x], it is also irreducible in F [x] by Theorem 2.5.11. Conversely, assume that (1) and (2) hold. If f (x) is an atom of R, it is clearly an atom of R[x] (Theorem 2.5.1). Suppose f (x) is a primitive polynomial of degree n ≥ 1 and f (x) is irreducible in F [x]. We claim that f (x) is an atom of R[x]. For, suppose not, and let f (x) = g(x)h(x), where g(x) and h(x) are nonunits of R[x]. (a) If g(x) or h(x) lies in R, then f (x) is not primitive, a contradiction. (b) If g(x) and h(x) both have positive degree, then f (x) is reducible in F [x], again a contradiction. Hence, if f (x) is an atom of R[x], it has the form (1) or (2), as required. Theorem 2.5.16. Let R be a UFD and f (x) an atom of R[x]. Then R[x]f (x) is a prime ideal of R[x]. That is, f (x) is a prime element. Proof. We consider separately the two types of atoms in R[x] given in Theorem 2.5.15. Case 1. a ∈ R is an atom of R. Since R is a UFD, every atom is a prime, so Ra is a prime ideal. Then (R/Ra)[x] is an integral domain. Since R[x]/R[x]a ∼ = (R/Ra)[x], R[x]a is a prime ideal, so a is prime. Case 2. f (x) is a primitive polynomial of degree n ≥ 1 and f (x) is irreducible in F [x] where F is the quotient field of R. First we claim that F [x]f (x) ∩ R[x] = R[x]f (x). Clearly, f (x) ∈ F [x]f (x) ∩ R[x]. Conversely, we suppose g(x)f (x) ∈ R[x] with g(x) = a0 /b0 + (a1 /b1 )x + · · · + (an /bn )xn ∈ F [x]. We can find relatively prime a, b ∈ R such that (b/a)g(x) = g1 (x) where g1 (x) is a primitive polynomial in R[x]. (In fact, a = gcd(a0 , a1 , . . . , an ) and b = lcm(b0 , b1 , . . . , bn ) will do, provided each ai and bi are relatively prime.) Thus, (b/a)g(x)f (x) = g1 (x)f (x) ∈ R[x]. By Gauss’ lemma, g1 (x)f (x) is a primitive polynomial. In connection with the above equation, this forces b to be a unit of R, so g(x) = (a/b)g1 (x) ∈ R[x]. Hence, g(x)f (x) ∈ R[x]f (x) which proves our claim. By the second isomorphism theorem, we have R[x]/R[x]f (x) = R[x]/(R[x] ∩ F [x]f (x)) ∼ = (R[x] + F [x]f (x))/F [x]f (x). Since (R[x] + F [x]f (x))/F [x]f (x) ⊆ F [x]/F [x]f (x) which is a field because f (x) is irreducible in F [x], R[x]+F [x]f (x))/F [x]f (x) is an integral domain. Thus, R[x]/R[x]f (x) is an integral domain, so R[x]f (x) is a prime ideal. Therefore, f (x) is prime and this proves the theorem. Theorem 2.5.17. [Gauss] If R is a UFD, so is R[x]. Hence, if R is a UFD, so is R[x1 , . . . , xn ] for all n ∈ N. 73 2.5. Polynomial Rings Proof. We know all the atoms of R[x] by Theorem 2.5.15 and Theorem 2.5.16 tells us that each atom of R[x] is prime. Hence, (by Theorem 2.4.4) to verify that R[x] is a UFD, it remains to show that each nonzero nonunit f (x) ∈ R[x] is a product of atoms. Case 1. deg f (x) = 0, i.e., f (x) ∈ R. Since R is a UFD and every atom of R is an atom of R[x], we can express f (x) as a product of atoms in R, and so in R[x]. Case 2. deg f (x) = n ≥ 1. Let f (x) = f1 (x) . . . fk (x) where (a) each fi (x) has degree ≥ 1 and (b) k is as large as possible. Such a factorization exists because any factorization which satisfies (a) has at most n terms since n = deg f (x) = deg f1 (x) + · · · + deg fk (x) ≥ k. Now, let ai be the content of fi (x), and let fi (x) = ai gi (x) where gi (x) is a primitive polynomial. We claim that gi (x) is an atom in R[x] because if gi (x) = r(x)s(x) where r(x) and s(x) are nonunits, then r(x) and s(x) cannot lie in R, since gi (x) is primitive. In addition, r(x) and s(x) cannot both have positive degree because then we could write f (x) as a product of k + 1 polynomials of positive degree, which violates (b). Thus, each gi (x) is an atom as desired. Hence, f (x) = f1 (x) . . . fk (x) = a1 g1 (x) . . . ak gk (x) = a1 . . . ak g1 (x) . . . gk (x) = ag1 (x) . . . gk (x), where a ∈ R. By Case 1, a can be written as a product of atoms in R[x] and therefore shows that f (x) is a product of atoms in R[x], which proves R[x] is a UFD. Example 2.5.4. Since Z is a UFD, we have Z[x] is also a UFD. However, in the following exercises, we shall know that the ideal (x, 2) of Z[x] is not principal, so Z[x] is an example of a UFD which is not a PID. Exercises 2.5. 1. Let R be a ring. (a) Show that Mn (R[x]) ∼ = Mn (R)[x] for all n ∈ N, where x is an indeterminate in both cases. (b) If I is an ideal of R, prove that I[x], the set of all polynomials with coefficients in I, is an ideal of R[x] and R[x]/I[x] ∼ = (R/I)[x]. 2. Prove the following statements. (a) If R is an integral domain, then x is a prime element in R[x]. (b) In Z[x], (x) is a prime ideal but not a maximal ideal. (c) If F is a field, then (x) is a maximal ideal in F [x]. a b 2 ∼ 3. Let F be a field. Show that F [x]/(x ) = : a, b ∈ F , a subring of M2 (F ). 0 a 4. Let F be a field. Find a ring isomorphism F [x]/(x2 − x) → F × F . 5. Prove that the ideal I = (x, 2) is not a principal ideal of Z[x]. Hence Z[x] is not a PID. In addition, show that I is a maximal ideal in Z[x]. 6. Construct a field of: (a) 125 elements (b) 81 elements. 7. Find all odd prime numbers p such that x + 2 is a factor of x4 + x3 + x2 − x + 1 in Zp [x]. 8. Let p(x) ∈ R[x]. Prove that if p(a + bi) = 0, then p(a − bi) = 0 for all a, b ∈ R. Deduce by the fundamental theorem of algebra that there exist real numbers c, r1 , . . . , rk , a1 , b1 , . . . , am , bm such that p(x) = c(x − r1 ) . . . (x − rk )(x2 − (2a1 )x + (a21 + b21 )) . . . (x2 − (2am )x + (a2m + b2m )). In addition, if p(x) ∈ R[x] is irreducible over R, then deg p(x) = 1 or 2, namely, p(x) = bx + c or p(x) = ax2 + bx + c with b2 − 4ac < 0. 9. Let p be an odd prime. Prove that xn − p is irreducible over Z[i][x]. 10. If R is an intergral domain for which every ideal of R[x] is principal, show that R must be a field. 11. Let D be an integral domain. If ϕ : D[x] → D[x] is an automorphism such that ϕ(a) = a for all a ∈ D, prove that there exist c, d ∈ D with c a unit in D such that ϕ(x) = cx + d. Here x stands for the indeterminate of D[x]. 74 2. Rings and Fields 12. Let R be a UFD and F its field of quotients. Let f (x) = a0 + a1 x + · · · + an xn lies in R[x], and suppose that there is an irreducible element p ∈ R such that (i) p ∤ an , (ii) p | a0 , . . . , an−1 , and (iii) p2 ∤ a0 . Prove that f (x) is irreducible in F [x]. Moreover, if f (x) is primitive, then f (x) is irreducible in R[x]. Project 14 (Units in a polynomial ring). Let R be a commutative ring and f (x) = a0 + a1 x + · · · + an xn in R[x]. Prove that f (x) is a unit in R[x] if and only if a0 is a unit in R and a1 , . . . , an are nilpotent elements in R. (This project generalizes the result in Theorem 2.5.1 (3) to any commutative ring.) Project 15 (Generalized Eisenstein’s criterion). Let P (x) = an xn +an−1 xn−1 +· · ·+a1 x+a0 be a polynomial with integer coefficients. If there exist a prime number p and an integer k ∈ {0, 1, . . . , n − 1} such that p | a0 , a1 , . . . , ak , p ∤ ak+1 and p2 ∤ a0 , then P (x) has an irreducible factor in Z[x] of degree greater than k. Extend this result to a UFD similar to the last question of Exercises 2.5. 2.6 Field Extensions Let F be a field and f (x) a polynomial over F of degree n ∈ N. Then the quotient ring F [x]/(f (x)) = {g(x) + (f (x)) : g(x) ∈ F [x]} = {g(x) + (f (x)) : g(x) ∈ F [x] and g(x) = 0 or deg g(x) < n} = {a0 + a1 x + · · · + an−1 xn−1 + (f (x)) : ai ∈ F } by the division algorithm. Thus, if F is a finite field, then F [x]/(f (x)) is a commutative ring of |F |n elements. In addition, if f (x) is irreducible in F [x], then (f (x)) is a maximal ideal, so F [x]/(f (x)) is a field. Note that F is isomorphic to {c + (f (x)) : c ∈ F }, so we may embed F into F [x]/(f (x)) by using the inclusion map. Examples 2.6.1. 1. R[x]/(x2 + 1) is a field isomorphic to C (with the map f (x) 7→ f (i)). 2. Z11 [x]/(x2 + 3) is a field of 121 elements. Under the above idea, if a polynomial over a field does not possess a a root in its own field, we shall create a bigger field where we can find a root of it. 2.6.1 Algebraic and Transcendental Extensions Before we extend a field, we first determine the smallest one possible according to its characteristic. Let F be a field. The intersection of all subfields of F is the smallest subfield of F , called the prime field of F . Theorem 2.6.1. Let F be a field with the prime subfield P and 1F denote the identity of F . 1. If char F = p, a prime, then P = {n · 1F : n = 0, 1, . . . , p − 1} ∼ = Z/pZ. 2. If char F = 0, then P = {(m · 1F )(n · 1F )−1 : m, n ∈ Z, n 6= 0} ∼ = Q. Proof. Since P is a field, 1F ∈ P , so {n · 1F : n ∈ Z} ⊆ P . Define ϕ : Z → P by ϕ(n) = n · 1F for all n ∈ Z. Then ϕ is a ring homomorphism and im ϕ = {n · 1F : n ∈ Z}, so Z/ ker ϕ ∼ = im ϕ. (1) Assume that char F = p is a prime. Then im ϕ = {n · 1F : n = 0, 1, . . . , p − 1} and p is the smallest positive integer such that p ∈ ker ϕ, so ker ϕ = pZ. Hence, im ϕ ∼ = Z/pZ which is a field, ∼ so P = im ϕ = Z/pZ. 75 2.6. Field Extensions (2) Assume that char F = 0. Then ϕ is a monomorphism. Since {n · 1F : n ∈ Z} ⊆ P and P is a subfield of F , {(m · 1F )(n · 1F )−1 : m, n ∈ Z, n 6= 0} ⊆ P . Define ϕ̄ : Q → P by ϕ̄(m/n) = ϕ(m)ϕ(n)−1 for all m, n ∈ Z, n 6= 0. Then ϕ̄ is a monomorphism and ϕ̄|Z = ϕ. Thus, Q∼ = im ϕ̄ = {(m · 1F )(n · 1F )−1 : m, n ∈ Z, n 6= 0} which is a subfield of P , and hence they are equal. In this section, we require some background in vector spaces. A field K is said to be an extension of a field F if F is a subring of K. Remark. By Theorem 2.6.1, any field can be considered as an extension field of the field Q or Zp for some prime p. Let K be an extension field of F . The degree of K over F , [K : F ], is the dimension of K as a vector space over F . More generally, if a field F is a subring of a ring R, then [R : F ] is the dimension of R as a vector space over F . For example, [C : R] = 2 and [R : Q] is infinite (in fact [R : Q] = |R|). Theorem 2.6.2. If [L : K] and [K : F ] are finite, then [L : F ] is finite and [L : F ] = [L : K][K : F ]. In fact, [L : F ] = [L : K][K : F ] whenever F ⊆ K ⊆ L. Proof. With F ⊆ K ⊆ L, let {βj }j∈J be a basis of K over F and {αi }i∈I a basis of L over K. Every element of L can be written uniquely as a finite linear combination of the elements of {αi }i∈I with coefficients in K, and every such coefficient can be written uniquely as a finite linear combination of the elements of {βj }j∈J with coefficients in F . Hence, every element of L can be written uniquely as a finite linear combination of the elements of {αi βj }i∈I,j∈J with coefficients in K: {αi βj }i∈I,j∈J is a basis of L over F , and [L : F ] = |I × J| = [L : K][K : F ]. Notation. Let K be an extension field of F . 1. If t1 , . . . , tn are indeterminates over F , then F (t1 , . . . , tn ) denotes the field of quotients of the polynomial ring F [t1 , . . . , tn ]. 2. If u1 , . . . , un ∈ K (or S ⊆ K), then F [u1 , . . . , un ] (or F [S]) denotes the subring of K generated by F and u1 , . . . , un (or S), and F (u1 , . . . , un ) (or F (S)) denotes its field of quotients. Theorem 2.6.3. [Classification of Elements in an Extension Field] Let K be a field extension of a field F and let u ∈ K. Then EITHER (a) [F (u) : F ] = ∞ and F [u] ∼ = F (t) where t is an indeterminate OR = F [t], so F (u) ∼ (b) [F (u) : F ] is finite and F [u] = F (u). Proof. Let t be an indeterminate and consider the ring homomorphism ϕ F [t] → K defined by ϕ(t) = u (or ϕ(f (t)) = f (u)). Note that the kernel of ϕ is a prime ideal, since the image of ϕ has no zero divisors. There are two possibilities. (1) ker ϕ = {0}. Then we have (a). (2) ker ϕ 6= {0}. Then ker ϕ = F [t]g(t) where g(t) is a monic prime (i.e., irreducible) polynomial. Since F [t] is a PID, F [t]g(t) is a maximal ideal. Thus, F [u] ∼ = F [t]/F [t]g(t) is a field, so F [u] = F (u). 76 2. Rings and Fields Remarks. 1. If g(t) = g0 + g1 t + · · · + gn−1 tn−1 + tn , then [F (u) : F ] = n and {1, u, . . . , un−1 } is a basis for F (u) over F . 2. Consider R ⊂ C and g(t) = g0 + g1 t + t2 ∈ R[t]. We distinguish three cases. (a) If g12 − 4g0 > 0, then g(t) = (t − a)(t − b) where a, b ∈ R, a 6= b and R[t]/R[t]g(t) is a ring without nonzero nilpotent elements. (b) If g12 − 4g0 = 0, then g(t) = (t − a)2 and R[t]/R[t]g(t) is a ring with nonzero nilpotent elements. (c) If g12 − 4g0 < 0, then R[t]/R[t]g(t) ∼ = C. √ 3. If p is a prime, then t2 − p is irreducible over Q and the fields Q( p) ∼ = Q[t]/(t2 − p) are all distinct. √ √ Proof. Let p and q be distinct primes. Assume that ϕ : Q[ p] → Q[ q] is an isomorphism. Then √ √ ϕ(1) = 1, and so ϕ(r) = r for all r in Q. Let ϕ( p) = a + b q for some a, b ∈ Q. Thus, p = ϕ(p) = √ √ √ √ (ϕ( p))2 = (a + b q)2 = (a + bq) + 2ab q. Since q is not rational, ab = 0. However, if a = 0, then √ √ p = bq which implies q | p. If b = 0, then ϕ( p) = a = ϕ(a), so p = a is rational. Hence, both cases √ √ lead to a contradiction. Therefore, Q[ p] and Q[ q] are not isomorphic. An element in an extension field can be classified according to Theorem 2.6.3 as follows. Let K be an extension field of a field F . An element u ∈ K is algebraic over F in case there exists a nonzero polynomial f (t) ∈ F [t] such that f (u) = 0 and transcendental element over F otherwise. √ √ For example, every complex number is algebraic over R; 3 2 and 1 + 5 ∈ R are algebraic over Q. It has been proved that e and π ∈ R are transcendental over Q (by the LindemannWeierstrass theorem). We show the existence of real numbers transcendental over Q in Corollary 5.2.3. Moreover, most of real numbers are in fact transcendental over Q (see Exercises 2.6). Theorem 2.6.3 yields characterizations of algebraic and transcendental elements. Corollary 2.6.4. Let K be an extension field of a field F and u ∈ K. The following conditions on u are equivalent: (i) u is transcendental over F (if f (t) ∈ F [t] and f (u) = 0, then f = 0); (ii) F (u) ∼ = F (t); (iii) [F (u) : F ] is infinite. Corollary 2.6.5. Let K be an extension field of a field F and u ∈ K. The following conditions on u are equivalent: (i) u is algebraic over F (there exists a polynomial 0 6= f (t) ∈ F [t] such that f (u) = 0); (ii) there exists a monic irreducible polynomial g(t) ∈ F [t] such that g(u) = 0; (iii) [F (u) : F ] is finite. Moreover, in part (ii), we have g(t) is unique; f (u) = 0 if and only if g(t) | f (t); F (u) ∼ = F [t]/(g(t)); and [F (u) : F ] = deg g(t). When u is algebraic over F , the unique monic irreducible polynomial g(t) ∈ F [t] in part (ii) is the minimal polynomial of u. The degree of u over F is deg g(t). An extension field K of a field F is algebraic in case every element of K is algebraic over F . For example, C is an algebraic extension of R, but R is not algebraic over Q. Note that if [K : F ] is finite, then K is an algebraic extension because [F (u) : F ] ≤ [K : F ] < ∞ for all u ∈ K. An extension field E of a field F is said to be a simple extension of F if E = F (α) for some α ∈ E. √ √ √ √ Example 2.6.2. Prove that Q( 2, 3) = Q( 2 + 3) is a simple extension. √ √ √ √ √ √ √ √ 3 ∈ 3), we have K ⊆ Q( 2, 3). For Solution. Let K = Q( 2 + √3). Since √ 2 2+ √ √ Q( 2, √ √ another inclusion, note that ( 2 + 3) = 5 + 2 2 3, so 2 3 ∈ K. Thus, √ √ √ √ √ √ √ √ √ √ √ √ √ √ 2 = ( 2 + 3) 2 3 − 2( 2 + 3) and 3 = 3( 2 + 3) − ( 2 + 3) 2 3 √ √ are in K. Hence, Q( 2, 3) = K. 77 2.6. Field Extensions Theorem 2.6.6. If L is an algebraic extension of K and K is an algebraic extension of F , then L is algebraic extension over F . Proof. Let u ∈ L. Since L is algebraic over K, there exists f (x) = a0 + a1 x + · · · + an xn ∈ K[x] such that f (u) = 0. Since K is algebraic over F , a0 , a1 , . . . , an are algebraic over F , so [F (a0 , a1 , . . . , an ) : F ] is finite. For, let E = F (a0 , a1 , . . . , an ). Then [E : F ] = [F (a0 ) : F ] n Y [F (a0 , a1 , . . . , ai ) : F (a0 , a1 , . . . , ai−1 )], i=1 a0 is algebraic over F and ai is algebraic over F (a0 , . . . , ai−1 ) for all i ∈ {1, . . . , n}. Since f (x) ∈ E[x], u is algebraic over E, so [E(u) : E] is finite by Corollary 2.6.5. Thus, [F (u) : F ] ≤ [E(u) : F ] = [E(u) : E][E : F ] < ∞. Hence, u is algebraic over F . Corollary 2.6.7. For a, b ∈ K, if a and b are algebraic over F of degree m and n, respectively, then a ± b, ab and a/b (if b 6= 0) are all algebraic over F of degree ≤ mn. Hence, the set of all algebraic elements of K over F is a subfield of K and is an algebraic extension over F . Proof. By Corollary 2.6.5, [F (a) : F ] = m and [F (b) : F ] = n. Since b is algebraic over F , b is algebraic over F (a), so [F (a)(b) : F (a)] ≤ n. Thus, by Theorem 2.6.2, [F (a, b) : F ] = [F (a)(b) : F ] = [F (a)(b) : F (a)][F (a) : F ] ≤ mn. Since a ± b, ab, ab−1 (if b 6= 0) are in F (a, b) which is a finite extension, they are all algebraic over F of degree ≤ mn. Example 2.6.3. Consider Q ⊂ C. Let A = {z ∈ C : z is algebraic over Q}. By Corollary 2.6.7, A is algebraic over Q. Assume that [A : Q] = n is finite. Let f (x) = xn+1 − 3. It is irreducible over Q by Eisenstien’s criterion. Let α ∈ C be such that f (α) = 0. Then α ∈ A ans so Q ⊂ Q[α] ⊂ A. But [Q[α] : Q] = n + 1 > [A : Q], which is a contradiction. Hence, [A : Q] is infinite. This provides an example of infinite algebraic field extensions. 2.6.2 More on Roots of Polynomials We conclude this chapter by working more on roots of polynomials. The theorem of Kronecker assures us that we may obtain an extension field of F in which the polynomial p(x) ∈ F [x] has a root. Theorem 2.6.8. If F is a field and G is a finite subgroup of the multiplicative group of nonzero elements of F , then G is a cyclic group. In particular, the multiplicative group of all nonzero elements of a finite field is cyclic. Proof. If G = {1}, then G is cyclic. Assume that G 6= {1}. Since G is a finite abelian group, G∼ = Z/(m1 ) ⊕ · · · ⊕ Z/(mk ) where k ≥ 1, m1 > 1 and m1 | · · · | mk . Since mk k X i=1 Z/(mi ) = 0, u is a root of the polynomial xmk − 1 ∈ F [x] for all u ∈ G. By Corollary 2.5.6, this polynomial has at most mk distinct roots in F , so |G| ≤ mk . Hence, we must have k = 1 and G ∼ = Z/(m1 ) which is a cyclic group. 78 2. Rings and Fields Remark. The finite multiplicative subgroup of a division ring may not be cyclic. E.g., Q8 = {±1, ±i, ±j, ±k} is a subgroup of the ring of real quaternions H and Q8 is not cyclic. Let R be an integral domain and f (x) ∈ R[x]. If α is a root of f (x), then there exist m ∈ N and g(x) ∈ R[x] such that f (x) = (x − α)m g(x) and g(α) 6= 0. m is called the multiplicity of the root α of f (x) and if m > 1, α is called a multiple root of f (x). If f (x) = a0 + a1 x + · · · + an xn ∈ R[x], we define f ′ (x) ∈ R[x], the derivative of f (x), to be the polynomial f ′ (x) = a1 + a2 x + · · · + nan xn−1 . We record the immediate properties of the derivative of polynomials in the next lemma Lemma 2.6.9. If f (x) and g(x) are polynomials over an integral domain R and c ∈ R, then 1. (cf (x))′ = cf ′ (x), 2. (f (x) + g(x))′ = f ′ (x) + g ′ (x), 3. (f (x)g(x))′ = f (x)g ′ (x) + f ′ (x)g(x), 4. ((f (x))n )′ = n(f (x))n−1 f ′ (x) where n ∈ N. Characterizations of polynomials with multiple roots using derivatives are as follows. Theorem 2.6.10. Let E be an extension of a field F and f (x) ∈ F [x]. 1. For α ∈ E, α is a multiple root of f (x) if and only if α is a root of both f (x) and f ′ (x). 2. If f (x) and f ′ (x) are relatively prime, then f (x) has no multiple root. 3. If f (x) is irreducible over F having a root in E, then f (x) has no multiple root in E if and only if f ′ (x) 6= 0. Proof. (1) is clear. (2) Since f (x) and f ′ (x) are relatively prime, there exist h(x) and k(x) in F [x] such that 1 = f (x)h(x) + f ′ (x)k(x). If α ∈ E is a multiple root of f (x), by (1), f (α) = 0 = f ′ (α), so 1 = 0, a contradiction. (3) Since f (x) is irreducible, f ′ (x) 6= 0 and deg f ′ (x) < deg f (x), we have f (x) and f ′ (x) are relatively prime, so f (x) has no multiple roots. Conversely, if f ′ (x) = 0, then f (α) = 0 = f ′ (α) for some α ∈ E since f (x) has a root in E. Hence, by (1), α is a multiple root of f (x). Theorem 2.6.11. [Number of Roots] If f (x) ∈ F [x] and deg f (x) = n > 1, then f (x) can have at most n roots counting multiplicities in any extension field of F . Proof. We shall prove the theorem by induction on the degree of f (x). If deg f (x) = 1, then f (x) = ax + b for some a, b ∈ F and a 6= 0. Then −b/a is the unique root of f (x) and −b/a ∈ F , so we are done. Let deg f (x) = n > 1 and assume that the result is true for all polynomials of degree < n. Let E be an extension field of F . If f (x) has no roots in E, then we are done. Let r ∈ E be a root of f (x) of multiplicity m ≥ 1. Then there exists q(x) ∈ E[x] such that f (x) = (x − r)m q(x) and q(r) 6= 0. Then deg q(x) = n − m. By the inductive hypothesis q(x) has at most n − m roots in E counting multiplicities. Hence, f (x) has at most m+(n−m) roots in E counting multiplicities. Theorem 2.6.12. [Kronocker] If p(t) ∈ F [t] is irreducible in F [t], then there exists an extension field E of F such that [E : F ] = deg p(t) and p(t) has a root in E. Proof. We use the discussion at the beginning of the section to prove this theorem. Let E = F [x]/(p(x)) where x is an indeterminate. Since p(x) is irreducible, E is a field containing {a + (p(x)) : a ∈ F } as a subfield. But F ∼ = {a + (p(x)) : a ∈ F } by ϕ : a 7→ a + (p(x)), so E can be considered as an extension field of F by considering a as a + (p(x)) for all a ∈ F . Then E = F [x]/(p(x)) = F (t̄) where t̄ = x + (p(x)) is a root of p(t). Since E = F (t̄) and p(t) is irreducible over F , [E : F ] = [F (t̄) : F ] = deg p(t) by Corollary 2.6.5. 79 2.6. Field Extensions Corollary 2.6.13. If p(t) ∈ F [t] is a nonconstant polynomial, then there exists a finite extension field E of F containing a root of p(t) and [E : F ] ≤ deg p(t). Proof. Since F [t] is a UFD, p(t) has an irreducible factor in F [t], say p1 (t). By Theorem 2.6.12, there exists an extension field E of F such that E contains a root of p1 (t) and [E : F ] = deg p1 (t). Hence, [E : F ] ≤ deg p(t) and E contains a root of p(t). Exercises 2.6. 1. If u ∈ K is algebraic of odd degree over F , prove that F (u2 ) = F (u). 2. Let a, b ∈ K be algebraic over F of degree m and n, respectively. Prove that if m and n are relatively prime, then [F (a, b) : F ] = √ √ √ mn.√ 3. Show that the degree of 2 + 3 over Q is 4 and the degree of 2 + 3 5 is 6. 4. Let p be a prime and let v ∈ C satisfy v 6= 1, v p = 1 (e.g., v = cos(2π/p) + i sin(2π/p)). Show that [Q(v) : Q] = p − 1. 5. Let E = Q(u) where u3 − u2 + u + 2 = 0. Express (u2 + u + 1)(u2 − u) and (u − 1)−1 in the form au2 + bu + c where a, b, c ∈ Q. 6. Let E be an algebraic extension of a field F . Show that any subring of E/F is a subfield. Hence prove that any subring of a finite dimensional extension field E/F is a subfield. 7. Let E = F (u), u transcendental and let K 6= F be a subfield of E/F . Show that u is algebraic over K. 8. Let u and v be positive irrational numbers such that u is algebraic over Q and v is transcendental over Q. (a) Show that v is transcendental over Q[u]. (b) Classify whether the following elements are algebraic or transcendental over Q. √ √ 1 u (iii) v (ii) (i) u+v 9. (a) Show that there are countably many irreducible polynomials in Q[x]. (b) Let A be the set of all real numbers that are algebraic over Q. Show that A is countable, so that R r A is uncountable. 10. Let R be an integral domain and f (x) a nonconstant polynomial. Prove that: (a) If char R = 0, then f ′ (x) 6= 0. (b) If char R = p, a prime, then f ′ (x) = 0 ⇔ ∃a0 , a1 , . . . , an ∈ R, f (x) = a0 + a1 xp + · · · + an xnp . 11. Suppose that F is a finite field and f (x) ∈ F [x] a nonconstant. If f ′ (x) = 0, prove that f (x) is reducible over F . 12. Let F be a finite field with q elements. Prove that if K is an extension field of F and b ∈ K is m algebraic over F , then bq = b for some m ∈ N. 13. A complex number α is called an algebraic integer if it is a root of a monic polynomial f (x) = xn + an−1 xn−1 + · · · + a1 x + a0 whose coefficients are in Z. Let A = {α ∈ C : α is an algebraic integer}. Prove that A ∩ Q = Z. 14. Let f (x) = x2 + x + 2 be a polynomial in Z3 [x] and E = Z3 [x]/(f (x)). (a) Show that f (x) is irreducible in Z3 [x], and so E is a field of 9 elements extending Z3 . (b) Find the characteristic of E and [E : Z3 ]. (c) Find the multiplicative inverse of 1 + x + (f (x)). (d) How many generators of the cyclic multiplicative group E r {0}? 15. Let E1 and E2 be subfields of a field K. The composite field of E1 and E2 , denoted by E1 E2 , is the smallest subfield of K containing both E1 and E2 . Prove that if [K : F ] is finite, then [E1 E2 : F ] ≤ [E1 : F ][E2 : F ]. 16. Let α be an irrational number. If α is a common root of f (x) = x3 + ax + b and g(x) = x2 + cx + d for some a, b, c, d ∈ Q, prove that: (a) g(x) is a factor of f (x) (b) c is a root of f (x). Project 16 (Galois ring). Let p be a prime and n, r ∈ N. Let f (x) be a monic irreducible polynomial in Zp [x] of degree r. Consider this polynomial as a polynomial in Zpn [x]. 80 2. Rings and Fields (a) Prove that the quotient ring R = Zpn [x]/(f (x)) is a commutative ring of pnr elements that contains the ring Zpn ∼ = {c + (f (x)) : c ∈ Zpn } as a subring. (b) Prove by the first isomorphism theorem that R/(p + (f (x))) ∼ = Zp [x]/(f (x)). Deduce that M = (p + (f (x)) is a maximal ideal of R. (c) Prove that R r M is the unit group R× . Conclude that M is the unique maximal ideal of R and so R is a local ring. The ring Zpn [x]/(f (x)) is called a Galois ring. It is a ring extension of the ring Zpn similar to a Galois field that is a field extension of the field Zp . This finite ring has many parallel properties to the finite field and has many applications in algebraic graph theory and algebraic coding theory. Project 17 (More on the exponents). Let G be a finite group. Recall the exponent of G defined before Theorem 1.7.13 in order to obtain information on the structure of a finite abelian group. Prove that: (a) exp G = lcm{o(a) : a ∈ G} where o(a) is the order of a in G. (b) If G = G1 × G2 , then exp G = lcm(exp G1 , exp G2 ). × (c) Let n ≥ 2. Clearly, exp Zn = n. Compute exp Z× n . The exponent of Zn is the Carmichael λ-function which was first introduced in 1910 (see [22]). (d) Find the exponent of the unit group of the Galois ring in Project 16. (e) Somer and Křížek [40, 41] used the Carmichael λ-function as a main tool to study the digraph of the kth power mapping of Zn . Meemark and Wiroonsri [36, 37] replaced it with the exponent of the group to obtain a general way to study this graphs. Their work on this digraphs influenced many articles. Unfortunately, they did not have the formula for the exponent explicitly. Let q be a prime power and f (x) be a monic polynomial in Fq [x] of degree ≥ 1. Compute the exponent of the unit group of the ring Fq [x]/(f (x)). (f) Compute the exponent of the unit group of the ring Zpn [x]/(f (x)m ) where m ∈ N and f (x) is a monic irreducible polynomial in Zp [x] of degree r considered as a polynomial in Zpn [x]. (g) [Open] Determine the exponent of the unit group of a finite local ring. More ring theory will be in terms of modules in Chapter 4. We shall classify extension fields and talk about the fundamental theorem of Galois theory in Chapter 5. 3 | Advanced Group Theory Deeper results of groups are presented in this chapter. Various kinds of series of a group are studied in the first three sections. A solvable group gets its name from the Galois group of a polynomial p(x) and solvability by radicals of the equation p(x) = 0. A nilpotent group can be considered as a generalization of an abelian group. A linear group gives an example of an infinite simple group. Finally, we discuss how to construct a group from a set of objects and presentations. 3.1 Jordan-Hölder Theorem The ideas of normal series of a group and solvability that arose in Galois theory yield invariants of groups (the Jordan-Höder theorem), showing that simple groups are, in a certain sense, building towers of finite groups. A subnormal series of a group G is a finite sequence H0 , H1 , . . . , Hn of subgroups of G such that Hi ⊳ Hi+1 (although not necessarily normal in G) for all i with H0 = {e} and Hn = G. The groups Hi+1 /Hi are called the factors associated with the series. A subnormal series is called a normal series of G if Hi ⊳ G for all i. Examples 3.1.1. 1. {0} < 8Z < 4Z < Z and {0} < 9Z < Z are normal series of Z. 2. {(1)} < A3 < S3 is a normal series of S3 . 3. {(1)} < A4 < S4 , {(1)} < V4 < S4 and {(1)} < V4 < A4 < S4 are normal series of S4 . Here V4 = {(1), (12)(34), (13)(24), (14)(23)}. 4. {(1)} < {(1), (12)(34)} < V4 < A4 < S4 is a subnormal series of S4 which is not a normal series. A subnormal [normal] series {Kj } is a refinement of a subnormal [normal] series {Hi } of a group G if {Hi } ⊆ {Kj }. Example 3.1.2. The series {0} < 72Z < 9Z < 3Z < Z is a refinement of the series {0} < 9Z < Z. Two subnormal [normal] series {Hi } and {Kj } of the same group G are isomorphic if there is a one-to-one correspondence between the collections of factor groups {Hi+1 /Hi } and {Kj+1 /Kj } such that corresponding factor groups are isomorphic. Clearly, two isomorphic subnormal [normal] series must have the same number of groups. Example 3.1.3. The two series of Z15 , {0} < h5i < Z15 and {0} < h3i < Z15 are isomorphic. The following theorem is fundamental to the theory of series. Theorem 3.1.1. [Schreier] Two subnormal [normal] series of a group G have isomorphic refinements. Example 3.1.4. Find isomorphic refinements of the normal series {0} < 8Z < 4Z < Z and {0} < 9Z < Z. 81 82 3. Advanced Group Theory Consider the refinement {0} < 72Z < 8Z < 4Z < Z of {0} < 8Z < 4Z < Z and the refinement {0} < 72Z < 18Z < 9Z < Z of {0} < 9Z < Z. In both cases the refinements have four factor groups isomorphic to Z4 , Z2 , Z9 , and 72Z or Z. The order in which the factor groups occurs is different to be sure. Recall the following fact. Theorem 3.1.2. If N is a normal subgroup of G, and if H is any subgroup of G, then HN = N H is a subgroup of G. Furthermore, if H ⊳ G, then HN ⊳ G. To prove Schreier’s theorem, we shall need the following lemma developed by Zassenhaus. This lemma is also called the butterfly lemma since the diagram which accompanies the lemma has a butterfly shape. Lemma 3.1.3. [Zassenhaus] Let H and K be subgroups of a group G and let H ∗ and K ∗ be normal subgroups of H and K respectively. Then 1. H ∗ (H ∩ K ∗ ) is a normal subgroup of H ∗ (H ∩ K). 2. K ∗ (H ∗ ∩ K) is a normal subgroup of K ∗ (H ∩ K). 3. H ∗ (H ∩ K)/H ∗ (H ∩ K ∗ ) ∼ = K ∗ (H ∩ K)/K ∗ (H ∗ ∩ K) ∼ = (H ∩ K)/[(H ∗ ∩ K)(H ∩ K ∗ )]. H ◆◆ ♣ K✭ ◆◆◆ ♣♣♣ ✭✭ ◆◆◆ ♣ ✖✖✖ ♣ ✭✭ ◆◆◆ ♣♣♣ ◆ ✭✭ ✖✖✖ ♣♣♣ ✭✭ ✖✖ K ∗ (H ∩ K) H ∗ (H ∩ K) ❘❘❘ ✭✭ ❧❧ ❘❘❘ ❧ ❧ ✖✖✖ ❘❘❘ ❧❧❧ ✭✭ ❧ ❘ ❧ ✖ ❘❘❘ ❧ ✭✭ ❧❧❧ ✖✖✖ ✭✭ H ∩ K ✭✭ ✖✖✖ ✭✭ ✖✖✖ ✭✭ ✖ K ∗ (H ∗ ∩ K) H ∗ (H ∩ K ∗ ) ✭✭ ✖✖✖ ✸ ❑ ❘ ❘❘❘ s ☛ ❧❧ ✸✸ ❑❑❑ ✭✭ s ❧ ❘ ❧ s ☛ ❘ ❧ ❘❘❘ ✸✸ ❑❑❑ ❧❧ ss ☛☛ ✭✭ ✖✖✖ ❧ s ❘ ❧ ❑❑ ❘❘❘ ❧ ✸✸ ss ❑❑ ✭✭ ❧❧❧ ☛☛ ss ✖✖✖ ✸ ☛ ❑ s ∗ ∗ ❑❑ s ✸✸ ☛ ✭✭ (H ∩ K)(H ∩ K ) s ❑ s ☛ ◗◗◗ ❑❑ ✸✸ ♠ ☛ ss ✭ ✖✖✖ ♠ ❑ s ♠ ◗ ☛ ❑❑ ◗◗◗ ✸✸ ☛ ss ♠♠♠ ❑❑ ✭✭✭ ♠ ◗ ☛ ♠ ◗ ✖✖✖ ssss ✸ ♠ ◗◗◗ ☛ ❑❑ ✸✸ ♠♠ s ◗◗◗ ☛☛ ♠♠♠ ✸ ♠ ☛ ◗ H∗ ■ K∗ ♠ ◗ ✸✸ ☛ ◗◗◗ ♠♠ ■■ ✉ ♠ ☛ ✉ ♠ ◗ ✸ ☛ ◗◗◗ ■■ ✉ ♠♠ ■■ ✉✉ ◗◗◗ ✸✸ ☛☛ ♠♠♠♠♠ ■■ ✉✉ ✸ ☛ ◗ ✉ ♠ ◗ ✉ ☛ ♠ H ∩ K∗ H∗ ∩ K Proof. We first note that H ∗ (H ∩ K), H ∗ (H ∩ K ∗ ), K ∗ (H ∩ K) and K ∗ (H ∗ ∩ K) are groups. It is easy to show that H ∗ ∩ K are H ∩ K ∗ are normal subgroups of H ∩ K. Apply Theorem 3.1.2 to H ∗ ∩K and H∩K ∗ as normal subgroups of H∩K, we have L = (H ∗ ∩K)(H∩K ∗ ) is a normal subgroup of H ∩ K. Thus we have the lattice of subgroups shown above. Let φ : H ∗ (H ∩ K) → (H ∩ K)/L be defined as follows. For h ∈ H ∗ and x ∈ H ∩ K, let φ(hx) = xL. We show φ is well defined and a homomorphism. Let h1 , h2 ∈ H ∗ and x1 , x2 ∈ H ∩K. 83 3.1. Jordan-Hölder Theorem −1 ∗ ∗ If h1 x1 = h2 x2 , then h−1 2 h1 = x2 x1 ∈ H ∩ (H ∩ K) = H ∩ K ⊆ L, so x1 L = x2 L. Thus φ is ∗ ∗ well defined. Since H is normal in H, there is h3 in H such that x1 h2 = h3 x1 . Then φ((h1 x1 )(h2 x2 )) = φ((h1 h3 )(x1 x2 )) = (x1 x2 )L = (x1 L)(x2 L) = φ(h1 x1 )φ(h2 x2 ) Thus, φ is a homomorphism. Obviously φ is onto (H ∩ K)/L. Finally if h ∈ H ∗ and x ∈ H ∩ K, then φ(hx) = xL = L if and only if x ∈ L, or if and only if hx ∈ H ∗ L = H ∗ (H ∗ ∩ K)(H ∩ K ∗ ) = H ∗ (H ∩ K ∗ ). Hence, ker φ = H ∗ (H ∩ K ∗ ). Another similar result follows by symmetry. Proof of Schreier’s theorem. Let G be a group and let {e} = H0 < H1 < H2 < · · · < Hn = G and {e} = K0 < K1 < K2 < · · · < Km = G be two subnormal series for G. For i where 0 ≤ i ≤ n − 1, we form the chain of (not necessarily distinct) groups Hi = Hi (Hi+1 ∩ K0 ) ≤ Hi (Hi+1 ∩ K1 ) ≤ · · · ≤ Hi (Hi+1 ∩ Km ) = Hi+1 . We refine the first subnormal series by inserting the above chain between Hi and Hi+1 . In a symmetric fashion, for 0 ≤ j ≤ m − 1, we insert the chain Kj = Kj (Kj+1 ∩ H0 ) ≤ Kj (Kj+1 ∩ H1 ) ≤ · · · ≤ Kj (Kj+1 ∩ Hn ) = Kj+1 between Kj and Kj+1 . Thus we get two refinement having mn terms. By Zassenhaus’s Lemma, we have Hi (Hi+1 ∩ Kj+1 )/Hi (Hi+1 ∩ Kj ) ∼ = Kj (Kj+1 ∩ Hi+1 )/Kj (Kj+1 ∩ Hi ) for 0 ≤ i ≤ n − 1 and 0 ≤ j ≤ m − 1. Hence, this two refinements are isomorphic. For normal series, where all Hi and Kj are normal in G, we merely observe that all the groups Hi (Hi+1 ∩ Kj ) and Kj (Kj+1 ∩ Hi ) are normal in G, so the same proof applies. A normal subgroup M (6= G) is called a maximal normal subgroup of G if there exists no normal subgroup N , other than G or M , such that M ⊳ N ⊳ G. Recall that a group G is simple if G and {e} are the only normal subgroups of G. For example, Zp , p a prime, and An , n 6= 4, are simple. We also have an obvious fact. Theorem 3.1.4. G is a simple abelian group if and only if G is cyclic of prime order. The next criterion follows directly from the third isomorphism theorem (Theorem 1.4.4). Theorem 3.1.5. M is a maximal normal subgroup of a group G if and only if G/M is simple. A subnormal series {Hi } of a group G is a composition series if all the factor groups Hi+1 /Hi are simple. A normal series {Hi } of G is a principal or chief series if all the factor groups Hi+1 /Hi are simple. Observe that by Theorem 3.1.5 Hi+1 /Hi is simple if and only if Hi is a maximal normal subgroup of Hi+1 . Thus for a composition series, each Hi must be a maximal normal subgroup of Hi+1 . To form a composition series of a group G, we just look for a maximal normal subgroup Hn−1 of G, then for a maximal normal subgroup of Hn−1 , and so on. If this process terminates in finite number of steps, we have a composition series. Hence, we have first shown: 84 3. Advanced Group Theory Theorem 3.1.6. If G is a finite group, then G has a composition series. Note that by Theorem 3.1.5 a composition series cannot have any further refinement. To form a principal series, we have to hunt for a maximal normal subgroup Hn−1 of G, then for a maximal normal subgroup of Hn−1 that is also normal in G, and so on. The main theorem is as follows. Theorem 3.1.7. [Jordan-Hölder] Any two composition [principal] series of a group G are isomorphic. Proof. Let {Hi } and {Ki } be two composition [principal] series of G. By Schreier’s theorem, they have isomorphic refinements. But since all factor groups are already simple, Theorem 3.1.5 shows that neither series has any further refinement. Hence, {Hi } and {Ki } must already be isomorphic. Examples 3.1.5 (Examples of composition series). 1. If G is simple, then {e} ⊳ G is the only normal series of G. It is a composition series for G and its associated factor is G = G/{e}. 2. If n 6= 4, then {(1)} < An < Sn is a composition series of Sn . 3. Z has many normal series. For example, let m1 , . . . , mn be positive integers. Then Z > m1 Z > m1 m2 Z > · · · > m1 m2 . . . mn Z > {0} is a normal series for Z whose associated factors are Zm1 , Zm2 , . . . , Zmn , Z. Note that since any nontrivial subgroup of Z is isomorphic to Z, any normal series for Z must have one associated factor isomorphic to Z. Hence, Z has no composition series. 4. Let p be prime and G = Zp × Zp . If (x, y) 6= (0, 0) in G, then h(x, y)i ∼ = Zp and {(0, 0)} < h(x, y)i < G is a composition series for G. The composition factors are G/h(x, y)i ∼ = Zp and 2 − 1)/(p − 1) = p + 1 h(x, y)i/(0, 0) ∼ Z , i.e., Z with multiplicity 2. Note that G has (p = p p subgroups of order p, so G has p + 1 distinct composition series. But in all cases they have the same composition factors: Zp with multiplicity 2. 5. Let p and q be prime and G = Zp × Zq = hai × hbi. Then the only proper subgroup of G are hai = Zp and hbi = Zq . Thus G has two composition series {e} < hai < G and {e} < hbi < G In both cases, the associated composition factors are Zp and Zq both with multiplicity one. 6. Consider Zp3 , Zp2 × Zp and Zp × Zp × Zp . In any composition series for these groups the same composition factors, namely Zp with multiplicity 3, occur. Exercises 3.1. 1. Suppose G has precisely two subgroups. Show that G has prime order. 2. A proper subgroup M of G is maximal if whenever M ⊆ H ⊆ G, we have H = M or H = G. Suppose G is finite and has only one maximal subgroup. Show that |G| is a power of prime. 3. Let G = Z36 . Consider two normal series {0} < h12i < h3i < Z36 and {0} < h18i < Z36 . Find two isomorphic chains and exhibit the isomorphic factor groups as described in the proof of Schreier’s Theorem. 4. Find a composition series for the dihedral group D4 = {σ, ρ : σ 4 = ρ2 = e and ρσρ−1 = σ −1 } and for the quaternion group Q = {±1, ±i, ±j, ±k}. Determine the composition factor in each case. 5. Prove that if G has a composition [resp. principal] series and if N is a proper normal subgroup of G, then there exists a composition [resp. principal] series containing N . Hence, show that N and G/N have composition [principal] series. 6. Show that if H0 = {e} < H1 < H2 < · · · < Hn = G is a subnormal [normal] series of G, and if Hi+1 /Hi is of finite order si+1 , then G is of finite order s1 s2 . . . sn . 7. Show that an infinite abelian group can have no composition series. 85 3.2. Solvable Groups 3.2 Solvable Groups Let G be a group. For g, h ∈ G, [g, h] = ghg −1 h−1 is called a commutator of G. The derived subgroup of G, denoted by G′ , is the group generated by all commutators of elements of G, i.e., G′ = hghg −1 h−1 : g, h ∈ Gi. The n-th derived subgroup of G, denoted by G(n) is defined inductively by G(0) = G and G(n) = (G(n−1) )′ for all n ≥ 1. Theorem 3.2.1. Let G be a group. 1. If N is a subgroup, then (N is normal and G/N is abelian) if and only if G′ ⊆ N . 2. G′ is a normal subgroup of G and G/G′ is abelian. 3. Every homomorphism θ : G → A, where A is an abelian group, factors through G/G′ . More precisely, there is a map θ̄ : G/G′ → A such that θ = θ̄ ◦ π, where π : G → G/G′ is the canonical projection. Proof. (1) Assume that N is normal and G/N is abelian. Let x, y ∈ G. Then xyN = yxN , so xyx−1 y −1 ∈ N . Thus G′ ⊆ N . Conversely, suppose that G′ ⊆ N . Let x, y ∈ G and n ∈ N . Then xnx−1 n−1 ∈ G′ ⊆ N which implies that xnx−1 ∈ N n = N . Hence, N ⊳ G. Since (xy)(yx)−1 = xyx−1 y −1 ∈ G′ ⊆ N , xyN = yxN , so G/N is abelian. (2) follows from 1 by taking N = G′ . (3) Define θ̄(xG′ ) = θ(x) for all x ∈ G. Clearly, θ = θ̄ ◦ π and is a homomorphism. Since θ(G′ ) = {e}, θ̄ is well defined. Remark. The quotient G/G′ is the largest abelian homomorphic image of G. The derived series of a group G is the sequence of groups G = G(0) ≥ G(1) ≥ G(2) ≥ · · · ≥ G(n) ≥ . . . . A group G is said to be solvable (of derived length ≤ n) if G(n) = {e} for some n. A solvable group arises from the study of the Galois group of a polynomial in order to obtain a criterion to determine if it is solvable by radicals. We shall see this in Section 5.7. A subgroup H of a group G which is invariant under all automorphisms, that is, ϕ(H) ≤ H for all ϕ ∈ Aut G, is called a characteristic subgroup of G. Using the inner automorphisms ϕa (x) = axa−1 for all a ∈ G, we deduce that every characteristic subgroup is normal in G. Lemma 3.2.2. Let ϕ : G → H be a surjective homomorphism. Then ϕ(G(i) ) = H (i) for every i ≥ 0. Also, G(i) is a characteristic subgroup for all i, and is thus normal in G. Proof. We have ϕ([x, y]) = [ϕ(x), ϕ(y)], and since ϕ is onto, we see that ϕ maps the set of commutators in G onto those in H. It follows that ϕ(G′ ) = H ′ , and repeated application of this argument yields that ϕ(G(i) ) = H (i) , as required. That the terms of the derived series of G are characteristic follows from the first part of the lemma when we take H = G and ϕ ∈ Aut G. Theorem 3.2.3. Let G be a group. Then G is solvable if and only if G has a subnormal series with abelian factors. Proof. If G is solvable, G = G(0) > G(1) > · · · > G(n) = {e} is a subnormal series with abelian factors. Conversely, suppose G = G0 > G1 > · · · > Gm = {e} is a subnormal series for G with abelian factors. Since Gi /Gi+1 is abelian, Gi+1 ≥ G′i . We claim Gi ≥ G(i) for i = 0, 1, . . . , m by induction on i. For i = 0, G0 = G = G(0) . Assume Gi ≥ G(i) . Then Gi+1 ≥ G′i ≥ (G(i) )′ = G(i+1) , which completes the induction. Hence, {e} = Gm ≥ G(m) , so G(m) = 1 and G is solvable. 86 3. Advanced Group Theory Remark. From Lemma 3.2.2, we know that G(i) ⊳ G for all i. Then the above derived series G = G(0) > G(1) > · · · > G(n) = {e} is indeed a normal series with abelian factors for G. Also, if G is solvable, its derived length, dl(G), is the smallest positive integer n such that G(n) = {e}. Examples 3.2.1 (Examples of solvable groups). 1. An abelian group G is solvable of derived length 1 because G′ = {e}. In addition, the groups with derived length 1 are exactly the abelian groups. Hence, a group G is abelian if and only if G is solvable of derived length 1. 2. Let Dn be the dihedral group of order 2n, i.e., Dn = {σ, ρ : σ n = ρ2 = e and ρσρ−1 = σ −1 }. Here, σ is the 2π/n rotation and ρ is the reflection of the regular n-gon. For example, (2) D1 = Z2 , D2 = Z2 × Z2 and D3 = S3 . Then Dn′ = hσ 2 i, an abelian group. Thus, Dn = {e}. For n = 1 or 2, Dn is abelian and hence has derived length one. For n ≥ 3, Dn is solvable of derived length two. Proof. Observe that Dn = {e, σ, σ 2 , . . . , σ n−1 , ρ, ρσ, ρσ 2 , . . . , ρσ n−1 }. For x, y ∈ Dn , we distinguish four cases σ k σ l σ −k σ −l = e (ρσ k )(ρσ l )(σ −k ρ−1 )(σ −l ρ−1 ) = ρσ k σ k σ −l σ −l ρ−1 = σ −2k σ 2l xyx−1 y −1 = (ρσ k )σ l (σ −k ρ−1 )σ −l = σ −l σ −l = σ −2l σ k (ρσ l )σ −k (σ −l ρ−1 ) = σ k σ k = σ 2k . This implies that Dn′ ⊆ hσ 2 i. On the other hand, we have σ 2 = ρσ −1 ρ−1 σ. Thus, Dn′ = hσ 2 i. 3. The groups S1 = {(1)} and S2 = Z2 are abelian groups. The group S3 = D3 is solvable of (3) derived length two. Since S4′ = A4 , A′4 = V4 and V4′ = {(1)} = S4 , we can conclude that (2) the group S4 is solvable of derived length 3. For n ≥ 5, Sn′ = An and A′n = Sn = An since An is simple and non-abelian. Therefore Sn ≥ An ≥ An ≥ . . . is the derived series of Sn and Sn is not solvable for n ≥ 5. These facts are important in Galois theory (Section 5.7) and relate to the famous formula for the solution of quadratic, cubic and quartic equations (by using square roots, cube roots, etc.), and the historic proof by Abel in 1824 that there are no such formula for the quintic equation. Proof. It is easy to see that any group of order two in A4 are not normal. Since A4 has more than one Sylow 3-subgroup, any subgroups of A4 of order three are not normal. Moreover, A4 has no subgroup of order six (see Exercises 1.5). Hence, the normal subgroups of A4 are A4 , V4 and {(1)}. Note that S4′ is a subgroup of A4 . Moreover, it is normal in A4 . Since S4 and S4 /V4 are not abelian, S4′ must be A4 . Since A4 is not abelian and A4 /V4 is abelian, we have A′4 = V4 . Hence, S4 ⊲ A4 ⊲ V4 ⊲ {(1)} is the derived series of S4 . Next, let n ≥ 5 and K = Sn′ ⊳ Sn . Then K ∩ An ⊳ Sn , so K ∩ An ⊳ An . Since An is simple, K ∩ An = {(1)} or K ∩ An = An . But K ⊆ An and K 6= {(1)} (since Sn is non-abelian), we get K = An . Hence, Sn′ = An . The following theorem is often useful to decide if a group is solvable. Theorem 3.2.4. 1. If G is solvable and H is a subgroup of G, then H is solvable. 2. If G is solvable and N is a normal subgroup G, then G/N is solvable. 3. A homomorphic image of a solvable group is solvable. 4. If N ⊳ G and N and G/N are solvable, then G is solvable and dl(G) ≤ dl(N ) + dl(G/N ). 5. If G and H are solvable, then G × H is solvable. 3.3. Nilpotent Groups 87 Proof. (1) Since H (i) ≤ G(i) for all i, H (n) = {e} if G(n) = {e}. (2) The application of Lemma 3.2.2 to the canonical homomorphism π : G → G/N yields that (G/N )(i) = π(G(i) ) for all i, and hence if G(n) = {e}, we have (G/N )(n) = {N }. (3) follows from (2). (4) Let dl(N ) = n and dl(G/N ) = m. Since the canonical homomorphism ϕ : G → G/N maps G(m) to (G/N )(m) = {N }, we see that G(m) ⊆ N . Thus G(m+n) = (G(m) )(n) ⊆ N (n) = {e}, and hence G is solvable. (5) follows from (4). Some additional conditions under which finite groups are solvable are as follows. Theorem 3.2.5. Let G be a finite group. 1. [Burnside] If |G| = pa q b for some primes p and q, then G is solvable. 2. [Philip Hall] If for every prime p dividing |G| we factor the order of G as |G| = pa m where (p, m) = 1, and G has a subgroup of order m, then G is solvable, i.e., if for all primes p, G has a subgroup whose index equals the order of a Sylow p-subgroup, then G is solvable—such subgroups are called Sylow p-complements. 3. [Feit-Thompson] If G is odd, then G is solvable. 4. [Thompson] If for every pair of elements x, y ∈ G, hx, yi is a solvable groups, then G is solvable. Burnside’s and Philip Hall’s Theorems were proved by using Character Theory. The proof of the Feit-Thompson Theorem takes 255 pages of hard mathematics (Solvability of groups of odd order, Pacific Journal of Mathematics, 13 (1963), pp. 775–1029). Thompson’s Theorem was first proved as a consequence of 475-page paper (that in turn relies ultimately on the Feit-Thompson Theorem). Exercises 3.2. 1. (a) Give an example of a normal subgroup of G which is not characteristic. (b) Prove that Z(G) is a characteristic subgroup of G. (c) If H is a characteristic subgroup of N and N ⊳ G, show that H ⊳ G. 2. Show that if G is a solvable simple group, then G is abelian. 3. Let {e} = H0 < H1 < H2 < · · · < Hn−1 < Hn = G be a composition for G. Prove that G is solvable if and only if the composition factors Hi+1 /Hi all have prime order. Deduce that if G is solvable with a composition series, then G is finite. 4. Find a composition series of S3 × S3 . Is S3 × S3 solvable? 5. Show that a group of order 1995 is solvable. 6. Let p < q < r be primes and let G1 be a group of order pq and let G2 be a group of order pqr. Prove that both of them are solvable. [Hint. G1 has a unique subgroup of order q.] 7. Let G be a group of order 495 = 32 · 5 · 11. (a) Prove that a Sylow 5-subgroup or a Sylow 11-subgroup of G is normal in G. (b) Let P be a Sylow 5-subgroup and Q a Sylow 11-subgroup of G. Prove that P Q is normal in G. (c) Prove that G is solvable. 8. Prove (without using the Feit-Thompson Theorem) that the following statements are equivalent: (i) every group of odd order is solvable (ii) the only simple groups of odd order are those of prime order. 3.3 Nilpotent Groups In this section, we shall introduce a class of groups whose structure, next to those of abelian groups, is most amenable to analysis. We begin by generalizing the notion of a commutator. 88 3. Advanced Group Theory If A and B are subsets of G, [A, B] is the subgroup of G generated by all commutators [a, b] = aba−1 b−1 where a ∈ A and b ∈ B, that is, [A, B] = h[a, b] : a ∈ A and b ∈ Bi. Note that [A, B] = [B, A]. Example 3.3.1. G′ = [G, G], G(2) = [G′ , G′ ], . . . , G(n+1) = [G(n) , G(n) ]. The lower central series of a group G is defined inductively by Γ1 (G) = G and Γn+1 (G) = [G, Γn (G)] for all n ≥ 1, so we get G = Γ1 (G) ≥ Γ2 (G) ≥ . . . and Γn (G) is called the n-th term of the lower central series of G. A group G is said to be nilpotent of class ≤ n if Γn+1 = {e}. Remarks. 1. Since Γ2 (G) = [G, G] = G′ , G is abelian if and only if G is nilpotent of class ≤ 1. 2. Note that the derived series commences G = G(0) ≥ G(1) ≥ . . . while the lower central series commences G = Γ1 (G) ≥ Γ2 (G) ≥ . . . . Note however that G is abelian if and only if {e} = [G, G] = G′ = Γ2 (G), so G is abelian ⇔ G is solvable of length ≤ 1 ⇔ G is nilpotent of class ≤ 1. Examples 3.3.2 (Examples of nilpotent groups). 1. S3 has the derived series S3 > A3 > {(1)} and has the lower central series S3 > A3 ≥ A3 ≥ . . . , so S3 is solvable (of length 2) but not nilpotent. 2. S4 has the derived series S4 > A4 > V4 > {(1)} and has the lower central series S4 > A4 ≥ A4 ≥ . . . , so S4 is solvable (of length 3) but not nilpotent. 3. Dn = hρ, τ : ρn = τ 2 = e and τ ρτ −1 = ρ−1 i has the derived series Dn > hρ2 i > {e} and has a lower central series Dn ≥ hρ2 i ≥ hρ4 i ≥ hρ8 i ≥ . . . . Hence, Dn is solvable (of length 2) r unless D1 or D2 which is abelian. But Dn is nilpotent if and only if ρ2 = e for some r if and only if n is a power of 2. Theorem 3.3.1. Let G be a group. Then Γn+1 (G) ≥ G(n) for all n ≥ 0. Hence, a nilpotent group is solvable. Therefore, Sn is not nilpotent for all n ≥ 5. Proof. We shall use induction on n. For n = 0, Γ1 (G) = G = G(0) . For the inductive step, we suppose Γn+1 (G) ≥ G(n) . Thus Γn+2 (G) = [G, Γn+1 (G)] ≥ [G(n) , G(n) ] = G(n+1) . Finally, assume that G is nilpotent. Then Γn+1 (G) = {e} for some n, so G(n) = {e}. Hence, G is solvable. Remark. In fact, we have Γ1 (G) ≥ G(0) , Γ2 (G) ≥ G(1) , Γ4 (G) ≥ G(2) , Γ8 (G) ≥ G(3) , . . . , Γ2n (G) ≥ G(n) , . . . but this is more difficult to prove. 89 3.3. Nilpotent Groups Recall that if N is a normal subgroup of G, then H ↔ H/N gives a 1-1 correspondence between subgroups of G containing N and subgroups of G/N . Moreover, this correspondence carries normal subgroups to normal subgroups. Now let Z(G) denote the center of a group G. Then Z(G) is a normal subgroup of G and Z(G/Z(G)) is a normal subgroup of G/Z(G). Hence, Z(G/Z(G)) = Z2 (G)/Z(G) where Z2 (G) is a normal subgroup of G containing Z(G). We generalize this construction to make the following definition. The upper central series of a group G is defined inductively by Z0 (G) = {e} and Zn+1 (G)/Zn (G) = Z(G/Zn (G)) for all n ≥ 1, so we get {e} = Z0 (G) ≤ Z1 (G) ≤ Z2 (G) ≤ . . . and Zn (G) is called the n-th term of the upper series of G. Remarks. 1. Z1 (G) is the center of G and Zi+1 (G)/Zi (G) is the center of G/Zi (G). 2. Zi+1 (G)/Zi (G) = Z(G/Zi (G)) is equivalent to Zi+1 (G) = {g ∈ G : [G, g] ≤ Zi (G)} because Zi+1 (G)/Zi (G) =Z(G/Zi (G)) ⇐⇒ ∀g ∈ G, [g ∈ Zi+1 (G) ⇔ ∀x ∈ G, gxZi (G) = xgZi (G)] ⇐⇒ ∀g ∈ G, [g ∈ Zi+1 (G) ⇔ ∀x ∈ G, xgx−1 g −1 ∈ Zi (G)] ⇐⇒ ∀g ∈ G, [g ∈ Zi+1 (G) ⇔ [G, g] ⊆ Zi (G)] ⇐⇒ Zi+1 (G) = {g ∈ G : [G, g] ≤ Zi (G)}. 3. We can show by induction that Zi (G) is a characteristic subgroup of G for all i ∈ N. A subnormal series G = G1 ≥ G2 ≥ . . . is called a central series for G if [G, Gi ] ≤ Gi+1 for all i. Remarks. 1. Since [G, Γi (G)] = Γi+1 (G), the lower central series is a central series for G. 2. Note that the condition Zi+1 (G)/Zi (G) = Z(G/Zi (G)) implies the inclusion [G, Zi+1 (G)] ≤ Zi (G). Thus, if Zn (G) = G for some n, then the upper central series (in reverse order) is a central series for G: G = Zn (G) ≥ Zn−1 (G) ≥ · · · ≥ Z1 (G) ≥ Z0 (G) = {e}. Now, we wish to collect equivalence definitions of a nilpotent group in terms of lower central series, upper central series and central series. Theorem 3.3.2. Let G be a group. 1. If G = G1 ≥ G2 ≥ G3 ≥ . . . is a central series for G, then Gn ≥ Γn (G) for all n. 2. G has a central series G = G1 > G2 > · · · > Gn+1 = {e} if and only if G is nilpotent of class ≤ n. Proof. (1) We shall use induction on n. For n = 1, G1 = G = Γ1 (G). For the inductive step, we suppose Gn ≥ Γn (G). Then Gn+1 ≥ [G, Gn ] ≥ [G, Γn (G)] = Γn+1 (G). (2) If G = G1 ≤ G2 ≤ . . . ≤ Gn+1 = {e} is a central series for G, then {e} = Gn+1 ≥ Γn+1 (G), so G is nilpotent of class ≤ n. Conversely, if G is nilpotent of class ≤ n, then G = Γ1 (G) ≥ . . . ≥ Γn+1 (G) = {e} is a central series of the required length. 90 3. Advanced Group Theory Theorem 3.3.3. Let G be a group. 1. Suppose G = G1 ≥ G2 ≥ . . . ≥ Gn+1 = {e} is a central series for G. Then Zk (G) ≥ Gn−k+1 for all k ∈ {0, 1, . . . , n}. 2. If Zn (G) = G, then G = Zn (G) ≥ Zn−1 (G) ≥ . . . ≥ Z1 (G) ≥ Z0 (G) = {e} is a central series for G. 3. G is nilpotent of class ≤ n if and only if Zn (G) = G. Proof. (1) We shall show that Zk (G) ≥ Gn−k+1 by induction on k. For k = 0, Z0 (G) = {e} = Gn+1 . Suppose Zk (G) ≥ Gn−k+1 . Let g ∈ Gn−(k+1)+1 = Gn−k , then [G, g] ≤ Gn−k+1 ≤ Zk (G), so g ∈ Zk+1 (G). Hence, Zk+1 (G) ≥ Gn−(k+1)+1 . (2) Since Zi+1 (G)/Zi (G) = Z(G/Zi (G)), [G, Zi+1 (G)] ≤ Zi (G). Hence, the given series is a central series. (3) follows from (1) and (2) using Theorem 3.3.2. Suppose that G is nilpotent of class ≤ n and that G = G1 ≥ G2 ≥ . . . ≥ Gn+1 = {e} is any central series for G. Theorems 3.3.2 and 3.3.3 show that we have the following inclusions G= G= G= ZnS(G) | G 1 S | Γ1 (G) ≥ ≥ ≥ Zn−1 S (G) | G 2 S | Γ2 (G) ≥ ··· ≥ ≥ ··· ≥ ≥ ··· ≥ ZkS(G) | Gn−k+1 S | Γn−k+1 (G) ≥ ··· ≥ ≥ ··· ≥ ≥ ··· ≥ Z0S(G) | GS n+1 | Γn+1 (G) = {e} = {e} = {e} In other words, of all central series for G, the upper central series has the largest groups and the lower central series has the smallest groups. We can restate some of the conclusions of Theorems 3.3.2 and 3.3.3 as follows. Theorem 3.3.4. Let G be a group. Then the following statements are equivalent. (i) G is nilpotent of class ≤ n. (ii) Γn+1 (G) = {e}. (iii) G has a central series G = G1 ≥ G2 ≥ . . . ≥ Gn+1 = {e}. (iv) Zn (G) = G. Next, we shall see that a finite nilpotent group behaves like a finite abelian group. We show that it is a direct product of its Sylow p-subgroups. We recall Theorem 1.6.2. Theorem 3.3.5. Let p be a prime. If G 6= {e} is a finite p-group, then Z(G) 6= {e}. We can thus prove another important fact. Theorem 3.3.6. Let G be a finite p-group. Then G is nilpotent, and hence G is solvable. Proof. Consider the upper central series {e} = Z0 (G) ≤ Z1 (G) ≤ Z2 (G) ≤ Z3 (G) ≤ . . . . If Zi (G) 6= G, then G/Zi (G) is a p-group, so Zi+1 (G)/Zi (G) = Z(G/Zi (G)) 6= {Zi (G)}. That is, Zi+1 (G) Zi (G). Since G is finite, the central series cannot increase for all i. Hence, Zn (G) = G for some n, so G is nilpotent. Theorem 3.3.7. Let G be a nilpotent group and let {e} < Z1 (G) < · · · < Zn (G) = G be the upper central series of G. Suppose H is a subgroup of G and define inductively N0 (H) = H, N1 (H) = N (H) = {g ∈ G : gHg −1 ⊆ H}, the normalizer of H and Nk+1 = N (Nk (H)) for all k ≥ 0. Then Nn (H) = G. 3.3. Nilpotent Groups 91 Proof. We shall prove by induction on i that Ni (H) ≥ Zi (G). For i = 0, N0 (H) = H ≥ {e} = Z0 (G). Suppose Ni (H) ≥ Zi (G). Let g ∈ Zi+1 (G). Then [g, G] ⊆ Zi (G). To show that g ∈ N (Ni (H)), let x ∈ Ni (H). Then gxg −1 x−1 ∈ Zi (G) ≤ Ni (H), so gxg −1 ∈ Ni (H)x = Ni (H). Hence, g ∈ Ni+1 (H). From the above theorem, we can deduce the following: Theorem 3.3.8. Suppose G is nilpotent and H is a proper subgroup of G. Then N (H) H. Before we discuss the main characterization theorem, we study some auxiliary results. Theorem 3.3.9. 1. If G is nilpotent and H is a subgroup of G, then H is nilpotent. 2. If G is nilpotent and N is a normal subgroup G, then G/N is nilpotent. 3. If G and H are nilpotent, then G × H is nilpotent. Proof. (1) and (2) are analogous to the proofs of 3.2.4 for G is solvable. (3) Suppose that G and H are nilpotent. Then there exist r, s > 0 so that Γr (G) = {eG } and Γs (H) = {eH }. Thus Γk (G × H) = Γk (G) × Γk (H) = {(eG , eH )} where k = max{r, s}. Hence, G × H is nilpotent. Finally, we shall that a finite nilpotent group behaves like a finite abelian group as we have seen in Theorem 1.7.12. This theorem characterizes all finite nilpotent groups. Theorem 3.3.10. [Finite Nilpotent Groups] Let G be a finite group. Then the following statements are equivalent. (i) G is nilpotent. (ii) All Sylow p-subgroups of G are normal in G. (iii) G is the direct product of its Sylow p-subgroups. Proof. (i) ⇒ (ii). Assume that G is nilpotent. Recall Theorem 1.6.10 that if P is a Sylow psubgroup, then N (N (P )) = N (P ). But Theorem 3.3.8 asserts that if H is a proper subgroup of G, then N (H) H. Thus we must have N (P ) = G, that is, P is normal in G since P ⊳ N (P ). (ii) ⇒ (iii). Note that if a Sylow p-subgroup P of G is normal in G, then it is the unique Sylow p-subgroup of G. Let p1 , p2 , . . . , pk be the distinct prime divisors of |G| and let Pi be the Sylow pi -subgroup of G. Suppose x ∈ Pi and y ∈ Pj where i 6= j. Then xyx−1 y −1 ∈ Pi ∩ Pj = {e}, so x and y commute. It follows that φ : P1 × · · · × Pk → G defined by φ(x1 , . . . , xk ) = x1 . . . xk is a homomorphism. It is easy to show that φ is a bijection. Hence, G is the direct product of its Sylow p-subgroups. (iii) ⇒ (i). A finite p-group is nilpotent (Theorem 3.3.6) and a finite direct product of nilpotent groups is nilpotent (Theorem 3.3.9). Hence, if G is the direct product of its Sylow p-subgroups, then G is nilpotent. The next corollary is just a restatement of Theorem 1.7.12. Corollary 3.3.11. A finite abelian group is the direct product of its Sylow subgroups. Exercises 3.3. 1. (a) [P. Hall] Let G be a group and x, y, z ∈ G. Write [x, y, z] for [[x, y], z]. Prove that [x, y −1 , z]y [y, z −1 , x]z [z, x−1 , y]x = e. (b) Let X, Y, Z ⊆ G and assume [X, Y, Z] = {e} = [Y, Z, X]. Prove that [Z, X, Y ] = {e}. 2. Prove that if N ≤ Z(G) and N and G/N are nilpotent, then G is nilpotent. Give an example of a group G with a normal subgroup N such that N and G/N are nilpotent but G is not nilpotent. 92 3. Advanced Group Theory 3. Let G be nilpotent of class 3. Show that if v ∈ G′ and x ∈ G, then v −1 xv = cx where c ∈ Z(G). Deduce that G′ is abelian. 4. Show that if G is a nilpotent group and N is a normal subgroup of G where N 6= {e}, then N ∩Z(G) 6= {e}. 5. Prove that if M is a maximal subgroup of a nilpotent group G, then M is normal and |G/M | = p where p is a prime. (A maximal subgroup is a proper subgroup which is not contained in any other proper subgroup. Infinite groups needs not possess maximal subgroups.) 6. Prove that if G is a nilpotent group and N is a minimal normal subgroup of G ({e} = 6 N is normal and simple), then N ≤ Z(G) and |N | = p for some p. Project 18 (Metabelian groups). A group G is metabelian if it admits a proper normal subgroup N such that both N and G/N are abelian. Prove the following statements. (a) All abelian groups are metabelian. (b) A group G is metabelian if and only if G′′ = {e}. Deduce that if G is a metabelian group, then G is solvable. Give an example of a solvable group which is not metabelian. (c) Every subgroup of a metabelian group is metabelian. (d) All nilpotent groups of class 3 or less are metabelian. 3.4 Linear Groups In this section, we talk about linear groups over a field. They have many interesting properties and provide us an example of an infinite simple group (Jordan-Moore’s theorem). Let K be a field and Mn (K) be the set of n × n matrices with entries in K. Then Mn (K) is a ring. Let GLn (K) denote the set of multiplicatively invertible elements in Mn (K), called the general linear group of degree n, that is, GLn (K) = Mn (K)× = {A ∈ Mn (K) : det(A) 6= 0}. Since det(AB) = det A det B, det : GLn (K) → K × is a homomorphism (of two groups). Its kernel consists of determinant one matrices, denoted by SLn (K) and called the special linear group of degree n. It is a normal subgroup of GLn (K) with quotient GLn (K)/SLn (K) isomorphic to K × = K r {0}. Geometrically, let V be a vector space over K of dimension n. Upon choosing a basis of V , we can represent all linear transformations from V to V via n × n matrices with entries in K. Then GLn (K) represents the invertible linear transformations on V , i.e., those which are one-to-one or equivalently those which are onto. Theorem 3.4.1. Let K be a field. Then Z(GLn (K)) = {λIn : λ ∈ K × } and Z(SLn (K)) = {λIn : λ ∈ K and λn = 1}, where In is the n × n identity matrix. Proof. For M to be in the center of G = GLn (K), it must commute with every N in G. In particular, M commutes with the elementary matrices. Recall that multiplying M on the left by an elementary matrix corresponds to performing an elementary row operation; multiplying M on the right by an elementary matrix corresponds to performing an elementary column operation. Thus, multiplying the ith row of M by a nonzero a gives you the same matrix as multiplying the ith column of M by a. This implies that the matrix is diagonal. Then, since interchanging the ith and jth row of M gives us the same matrix as swapping the ith and jth column of M , the ith entry along the diagonal must equal the jth entry along the diagonal, for all i and j. Therefore, 93 3.4. Linear Groups M must be a multiple of In . Finally, it is easy to see that all nonzero multiples of In do commute with all N ∈ G. Hence, the theorem is proved for GLn (K). For the center of SLn (K), we need to use the elementary matrices Xij (a), i 6= j, whose entries are the same as that of the identity matrix In except for an a ∈ K in the (i, j) location. It is obtained by performing the row operation Ri + aRj , i 6= j or the column operation Cj + aCi on In . Clearly, Xij (a) ∈ SLn (K) for all a ∈ K and i 6= j. If M is in the center of SLn (K), then M must commute with Xij (1) for all i 6= j, so the ith and jth columns and rows must be all zeros except for the (i, i) and (j, j) entries which must be equal. Moreover, the product of the diagonal entries is the determinant which is equal to 1. From the above theorem, the center of GLn (K) consists of scalar matrices λIn with λ ∈ K × and the center of SLn (K) consists of scalar matrices λIn with λ ∈ K and λn = 1. They are normal and lead to the next definitions. The quotient group GLn (K)/Z(GLn (K)) = PGLn (K), called the projective linear group of degree n. The quotient SLn (K)/Z(SLn (K)) = PSLn (K) is called the projective special linear group of degree n. If K is finite, we may determine the cardinality of each linear group as follows. Theorem 3.4.2. If |K| = q < ∞, then |GLn (K)| = (q n − 1)(q n − q)(q n − q 2 ) . . . (q n − q n−1 ). Proof. Let A ∈ GLn (K). Then the columns of A are linearly independent vectors in K n . Thus the first column of A can be any nonzero vectors in K n . The second column must not be multiple of the first column, and the jth column must not be a linear combination of the previous j − 1 columns for all j = 2, . . . , n. By the product rule, we obtain the theorem. Corollary 3.4.3. Let K be a finite field with q elements. Then |SLn (K)| = |PGLn (K)| = (q n − 1)(q n − q)(q n − q 2 ) . . . (q n − q n−2 )q n−1 and |PSL2 (K)| = |SL2 (K)| if charK = 2 and |PSL2 (K)| = |SL2 (K)|/2 if charK 6= 2. Proof. They follow from their definitions and Theorem 3.4.2. Lemma3.4.4. Let a field. The groupSL2 (K) is generated by the union of the two subgroups K be 1 λ 1 0 : λ ∈ K and : µ ∈ K . Hence, every matrix, in SL2 (K) is a finite product of 0 1 µ 1 matrices which either upper triangular or lower triangular and which have 1’s along the diagonal. These matrices are called unipotent matrices or transvections. a b Proof. Let ∈ SL2 (K). Assume that c 6= 0. Perform the following row/column transformac d tions: d−1 d−1 R2 C1 R1 + 1−a C2 + 1−d R2 −cR1 1 0 a b 1 1 c c c c / / / . 0 1 c d because ad − bc = 1 c d 0 1 Thus, Hence, 1 0 1 −c 1 0 1−a c 1 a b is a product of transvections. c d a b c d 1 0 1−d c 1 1 0 = . 0 1 94 3. Advanced Group Theory a+b b If c = 0, then d 6= 0 and the matrix ∈ SL2 (K) can be treated as in the first case. d d However, a b a+b b 1 0 = 0 d d d −1 1 and the result follows. Theorem 3.4.5. Let K be a field. The elementary matrices Xij (a), defined in the proof of Theorem 3.4.1, generate SLn (K). Proof. If n = 1, then SL1 (K) = {1} is trivial. Lemma 3.4.4 gives the case n = 2. For n > 2, the theorem follows from the mathematical induction in a similar manner. Lemma 3.4.6. The elementary matrices Xij (a), defined in the proof of Theorem 3.4.1, are commutators in SLn (K) except in the case n = 2 and (|K| = 2 or 3). Proof. If n ≥ 3, this is easy since there is a third index k and [Xik (a), Xkj (a)] = Xij (a). If n = 2, we use the commutator relation 1 β 1 (α2 − 1)β α 0 , = . 0 1 0 1 0 α−1 However, given any λ ∈ K, the equation λ = (α2 − 1)β can be solved for β if and only if there exists a nonzero α ∈ K so that α2 6= 1 (i.e., α 6= ±1). This works as long as K × has at least three elements. Corollary 3.4.7. Let K be a field. If n ≥ 2, then SLn (K) is not solvable except in the cases SL2 (F2 ) and SL2 (F3 ). Remark. SL2 (F2 ) ∼ = A4 . They are solvable groups. = S4 and PSL2 (F3 ) ∼ = S3 , SL2 (F3 ) ∼ = PSL2 (F2 ) ∼ Moreover, PSL2 (F2 ) and PSL2 (F3 ) are not simple. The following theorem was proved by C. Jordan in 1870 for |K| prime. In 1893, after F. Cole discovered a simple group G of order 504, E. H. Moore recognized G as PSL2 (F8 ), and then proved the simplicity of PSL2 (K) for all K of size ≥ 4. It provides an example of infinite simple groups. Theorem 3.4.8. [Jordan-Moore] Let K be a field with |K| ≥ 4. Then PSL2 (K) is a simple group. Proof. Using the third isomorphism theorem, it suffices to prove that a normal subgroup N of SL2 (K) containing a matrix other than ±I2 must be all of SL2 (K). Let A 6= ±I2 be a matrix in N . Then there is a vector ~v in K 2 so that ~v and A~v are linearly independent over K. This means that 0 b {~v , A~v } is a basis of K 2 . The matrix representation of A with respect to this basis is (since 1 d A~v = 0 · ~v + 1 · A~v and A(A~v ) = b · ~v + d · A~vfor some b, d ∈ K). Since det A = 1, we actually 0 −1 0 −1 have b = −1. That is, A is conjugate to . Since N is normal, is also in N . 1 d 1 d Our strategy is to show that N contains all unipotent elements in SL2 (K) by repeatedly using −1 −1 CB ∈ N for all C ∈ SL (K) and B ∈ N ”. First, apply this trick with B = A the fact that 2 “C B α 0 × (α ∈ K ) to get and C = 0 α−1 −2 α d(α−2 − 1) −1 −1 ∈ N. C A CA = 0 α2 95 3.4. Linear Groups Next, repeat the fact with B′ 1 µ α−2 d(α−2 − 1) ′ and C = = (µ ∈ K), we get 0 1 0 α2 C ′−1 B ′−1 1 µ(α4 − 1) CB = ∈ N. 0 1 ′ ′ We get all upper triangular unipotent elements in N as long as there exists an α ∈ K × such that α4 6= 1. This happens if |K| ≥ 6 since the polynomial x4 − 1 has at most four distinct roots in K × × 4 or if |K| = 4 since F× 4 is cyclic of order 3 and α = α for all α ∈ F4 . Observe that −1 0 −1 1 µ 0 −1 1 0 = 1 0 0 1 1 0 µ 1 for all µ ∈ K × . This proves that N = SL2 (K) if |K| ≥ 4 and |K| 6= 5. −2 α d(α−2 − 1) ∈ N for all α ∈ K × . It remains to deal with the case K = F5 . We still have 0 α2 2 −1 −2d −1 −2d 1 −d Take α = 2 to get ∈ N , and hence = ∈ N . Two cases are 0 −1 0 −1 0 1 possible: 1 −d (a) d 6= 0. The powers of give all upper triangular unipotent elements. By conjugating 0 1 0 −1 with , the lower triangular ones appear. Thus, N = SL2 (K). 1 0 δ 1 0 −1 ′′ . We then perform the standard trick with B = A and C = (b) d = 0, so A = −1 0 1 0 (δ ∈ F× ), so that 5 1 −δ ′′−1 −1 Aδ = C A CA = ∈ N. −δ δ 2 + 1 Since δ 6= 0, this element is not in the center. Note that itstrace is δ 2 + 2 is never zero. Choose 0 −1 (as at the beginning of the δ = 1, say. Then A1 ∈ N and A is conjugate to A′ = 1 3 proof and the trace remains the same under conjugation). Apply Case (a), to A′ , and the proof is complete. Exercises 3.4. 1. Show that there is no non-abelian finite simple group of order less than 60. (Hint. We may focus on groups of the following orders: 24, 30, 40, 48, 54 and 56.) 2. Suppose G is a simple group of order 60. Show that: (a) G has a subgroup A of order 12 (b) A has exactly five different conjugates (c) there is an injective homomorphism from G to S5 (d) both A5 and H contain every element of S5 of the form g 2 and therefore every 5-cycle and every 3-cycle (e) H = A5 . Deduce that any simple group of order 60 must be isomorphic to A5 and hence PSL2 (F4 ) and PSL2 (F5 ) are isomorphic to A5 . Project 19 (The groups GL2 (Z/N Z) and SL2 (Z/N Z)). In this project, we determine the structure and the cardinality of the groups GL2 (Z/N Z) and SL2 (Z/N Z). (a) Prove that for any integer N , the map SL2 (Z) → SL2 (Z/N Z) obtained by reducing the matrix entries modulo N is a surjective group homomorphism. 96 3. Advanced Group Theory (b) Prove that for positive integers M and N , the maps (“reduction modulo N ”) from SL2 (Z/M N Z) to SL2 (Z/N Z) and from GL2 (Z/M N Z) to GL2 (Z/N Z) are surjective group homomorphisms. (c) What is the kernel of the homomorphism GL2 (Z/pe Z) → GL2 (Z/pZ)? (d) What are the order of the groups GL2 (Z/pe Z) and SL2 (Z/pe Z)? (e) Let N = pe11 . . . perr be the prime factorization of the positive integer N . Show that the reductions e modulo pj j , j = 1, . . . , r, give isomorphisms GL2 (Z/N Z) ∼ = Y e GL2 (Z/pj j ) and j SL2 (Z/N Z) ∼ = Y e SL2 (Z/pj j ). j (f) What are the order of the groups GL2 (Z/N Z) and SL2 (Z/N Z)? 3.5 Free Groups and Presentations There is a basic method of defining a group G, called a presentation of G by generators and defining relations. We have used this method without defining it precisely. For example, hai means the cyclic group generated by a. If a happened to be an element of some larger group G, then hai means the subgroup of G generated by hai. It could be infinite cyclic or finite cyclic. More generally, if we were working a particular group G, and a1 , . . . , ak ∈ G, then ha1 , . . . ak i denoted the subgroup of G generated by a1 , . . . ak . However, when we were not talking about subgroups of a particular group G, then the brackets h i had a different meaning as shown by the following examples. Examples 3.5.1. 1. hai ∼ = Zn . = Z and ha : an = ei ∼ n m 2. ha, b : a = e, b = e, ab = bai = ha, b : an = e, bm = e, aba−1 b−1 = ei ∼ = Zn × Zm . −1 ∼ 3. ha1 , . . . , ak : an1 1 = · · · = ank k = e, ai aj a−1 a = e if i = 6 ji Z × · · · × Znk . = n1 i j −1 −1 ∼ 4. ha1 , . . . , ak : ai aj ai aj = e if i 6= ji = Z × · · · × Z (k copies). 5. Dn = ha, b : an = b2 = e, bab−1 = a−1 i = ha, b : an = b2 = e, bab−1 a = ei is the dihedral group of order 2n. −1 6. D∞ = ha, b : b2 = e, bab−1 = a−1 i = ha, b(" : b2 = e, bab # a = ei is)the infinite dihedral group. 1 0 0 7. ha, b : a[a, b] = [a, b]a, b[a, b] = [a, b]bi ∼ = by a 7→ Observe that c = a−1 b−1 ab = " " 1 0 0 1 0 0 1 1 0 0 1 0 0 0 1 1 0 1 # # p 1 0 and q r 1 : p, q, r ∈ Z . An isomorphism is given b 7→ " 1 0 0 0 1 0 0 1 1 # . . In each of the above examples the data inside the brackets h i is sufficient to describe a group, that is, it gives the multiplication table for a groups. We call such an expression a presentation for the group. It turns out that every group has a presentation and every presentation defines a group. However, it is generally difficult to decide if a group defined by a presentation is isomorphic to an explicitly given group. Let A be any (not necessarily finite) set of elements ai for i ∈ I. We think of A as an alphabet set and of the ai as letters in the alphabet set. Any symbol of the form ani with n ∈ Z is a syllable and a finite string w of syllables written in juxtaposition is a word. We also introduce the empty word 1, which has no syllables. A word on A is reduced if w = 1 or the string ai a−i or a−i ai does not appear in w for all a ∈ A and i ∈ N. Let A be a set. Write F [A] for the set of all reduced words formed from our alphabet A. For convenience, we may let F [∅] = {1}. We make F [A] into a group by the juxtaposition w1 w2 of 3.5. Free Groups and Presentations 97 two words w1 and w2 with reduction of strings ai a−i or a−i ai (if any) for all a ∈ A and i ∈ N. It is called the free group generated by A. Example 3.5.2. The only example of a free group that has occurred before is Z, which is free on one generators. Clearly, every free group is infinite. Example 3.5.3. F2 = hx, yi. The element of F2 are all words in x and y. More precisely, F2 is the disjoint union of the following seven sets. 1. {1} 4. {xi1 y j1 . . . xik y jk : k > 0, ir , jr ∈ Z r {0}} 2. {xi : i ∈ Z r {0}} 5. {xi1 y j1 . . . xik y jk xik+1 : k > 0, ir , jr ∈ Z r {0}} i 3. {y : i ∈ Z r {0}} 6. {y j1 xi1 . . . y jk xik : k > 0, ir , jr ∈ Z r {0}} 7. {y j1 xi1 . . . y jk xik y jk+1 : k > 0, ir , jr ∈ Z r {0}} Let G be a group and let A be a subset of G such that hAi = G. If G is isomorphic to F [A] under the map ϕ : G → F [A] such that ϕ(a) = a for all a ∈ A, then G is said to be free on A. A group is free if it is free on some nonempty set A. Theorem 3.5.1. [Universal Mapping Property of a Free Group] Let A be a nonempty set. Suppose H is any group and there is a function φ : A → H. 1. There is a unique homomorphism Φ : F [A] → H extending φ. 2. If im φ generates H, then Φ : F [A] → H is a surjection. 3. If G is a group and θ : G → F [A] is an onto homomorphism, then there is a homomorphism Φ : F [A] → G such that θ ◦ Φ = id F [A] , the identity map on F [A]. Proof. (1) is clear and (2) follows immediately from (1). (3) Since θ is onto, for each a ∈ A, there is a ga ∈ G such that θ(ga ) = a. By (1), there is a unique homomorphism Φ : F [A] → H with Φ(a) = ga for all a ∈ A. Then θ ◦ Φ : F [A] → F [A] is the identity map. Similarly, we can show that Corollary 3.5.2. Let S be a set. Then there is a unique free group on S. Proof. Let G1 and G2 be free groups on S. Then S is a subset of both G1 and G2 . Consider the inclusion maps ι1 : S → G1 and ι2 : S → G2 and the result follows from the uniqueness of the universal mapping property. Corollary 3.5.3. Every group H is a homomorphic image of a free group. Proof. Let A be a set for which there exists a bijection φ : A → H (e.g., take A = H and φ = id H ), and let G = F [A]. By the universal mapping property, there is an onto homomorphism Φ : G → H extending φ. Therefore, G/(ker Φ) ∼ = H. We refer the reader to reference textbooks for proofs of the next three theorems. They are stated simply to inform us of these interesting facts. Theorem 3.5.4. If a group G is free on A and also on B (not necessarily finite), then the sets A and B have the same number of elements; that is, any two sets of generators of a free group have the same cardinality. We shall prove this theorem for the finite basis case with some result on finitely generated free abelian group (Corollary 4.2.7) in the next chapter. If G is free on a set A, the number of elements in A is called the rank of G. 98 3. Advanced Group Theory Theorem 3.5.5. Two free groups are isomorphic if and only if they have the same rank. Theorem 3.5.6. [Schreier] A nontrivial proper subgroup of a free group is free. This is not trivial to prove. There is a nice proof of this result using covering spaces (cf. J.-P. Serre, Trees, Springer-Verlag, 1980). Example 3.5.4. Let yl = xl yx−l for l ≥ 0. Then yl , l ≥ 0, are free generators for the subgroup of F2 = hx, yi that they generate. This illustrates that although a subgroup of a free group is free, the rank of the subgroup may be much greater than the rank of the whole group! φ θ / / K be a sequence of groups homomorphisms. We say that it is exact at H Let G H if im θ = ker φ. A short exact sequence of groups is a sequence of groups and homomorphisms 1 /G θ /H φ /K /1 which is exact at G, H and K. In other words, if θ is 1-1, φ is onto and im θ = ker φ. Remark. If N is a normal subgroup of G, then 1 → N → G → G/N → 1 is exact. Conversely, if ι 1 → N → G → H → 1 is exact, then N is normal in G and H ∼ = G/N . Thus short exact sequences are just another notation for normal subgroups and factor groups. A presentation for a group G is an expression G = hg1 , . . . , gr : w1 = · · · = wt = 1i where w1 , . . . , wt are words in g1 , . . . , gr such that the following two properties are satisfied: (1) g1 , . . . , gr generate G and (2) the conditions that w1 = w2 = · · · = wt = 1 are sufficient to define the multiplication table of G. Here, g1 , . . . , gr are called generators of G in the presentation and w1 , w2 , . . . , wt are called defining relations. Note that the free group of rank n is the group Fn = hx1 , . . . , xn : i given by a presentation with n generators and zero defining relation. Remark. The elements of hx1 , . . . , xn i are words in x1 , . . . , xn . Suppose w = w(x1 , . . . , xn ) is any such word. Then if G is any group, we can think of w as a function G × · · · × G → G −1 such that (g1 , . . . , gn ) 7→ w(g1 , . . . , gn ). For example, if w(x1 , x2 ) = [x1 , x2 ] = x1 x2 x−1 1 x2 , then w(g1 , g2 ) = g1 g2 g1−1 g2−1 . Remark. If we have Fn = hx1 , . . . , xn i is a free group and G = hy1 , . . . , yn : w1 = · · · = wt = 1i, then the universal mapping property of a free group says that φ(xi ) = yi defines an onto homomorphism φ : Fn → G. This means we have a short exact sequence 1 / ker φ ι / Fn φ /G / 1. What is the kernel of φ? ker φ is a normal subgroup of Fn and contains wi (x1 , . . . , xn ) for i = 1, . . . , t. In fact, ker φ is the smallest normal subgroup of Fn which contains wi (x1 , . . . , xn ) for i = 1, . . . , t. Let G be a group and S a subset of G. The normal closure of S in G, denoted by hSiG , is the smallest normal subgroup of G containing S. It is the subgroup of G generated by all conjugates of elements of S by elements of G. That is, hSiG = hxyx−1 : x ∈ G and y ∈ Si and so 99 3.5. Free Groups and Presentations Theorem 3.5.7. Let G = hx1 , . . . , xn : w1 = · · · = wt = 1i. Then G ∼ = F/N where N is the normal closure of {w1 , . . . , wt } in the free group F = F [{x1 , . . . , xn }]. Example 3.5.5. Consider the free group F2 = hx, yi. Let G = hx, y : xyx−1 y −1 = 1i ∼ = F2 /N . ′ −1 −1 ′ Since G is abelian, F2 ⊆ N . But xyx y ∈ N and N is the smallest, so N = F2 . Example 3.5.6. Consider the quaternion group Q8 = ha, b : a4 = 1, a2 = b2 , ba = a3 bi of order / hai, eight. We shall determine the structure of Q8 /Q′8 . Since |a| = 4 and ba = a3 b, hai ⊳ Q8 and b ∈ ′ 2 − −1 ′ 2 ′ so Q8 /hai = {hai, bhai}. Then Q8 ⊆ hai. Since a = a 1bab ∈ Q8 , ha i ⊆ Q8 ⊆ hai. In addition, ba2 b−1 = a−2 . Thus, ha2 i is normal in Q8 . Since |Q8 /ha2 i| = 4, it is abelian. Hence, Q′8 = ha2 i. Note that a2 Q′8 = b2 Q′8 = Q′8 . Therefore, Q8 /Q′8 ∼ = Z2 × Z2 . Theorem 3.5.8. [von Dyck’s Theorem /fon dike/] Let G be given by a presentation G = hx1 , . . . , xn : w1 = · · · = wt = 1i. Suppose H is any group which satisfies: 1. H is generated by h1 , . . . , hn and 2. wi (h1 , . . . , hn ) = 1 for i = 1, . . . , t. Then there is a unique onto homomorphism φ : G → H for which φ(xi ) = hi . Proof. By Theorem 3.5.7, G ∼ = F/N , where F is a free group on {x1 , . . . , xn } and N is the normal closure of {w1 , . . . , wt }. By the assumption N ⊆ ker φ, so φ induces a (well defined) homomorphism xi = xi N 7→ hi for all i ∈ {1, . . . , n}. Example 3.5.7. Classify all groups G of order six. Proof. Since 6 = 2 · 3, G contains elements a and b such that |a| = 2 and |b| = 3 and G = ha, bi. Since hbi is normal in G, aba−1 ∈ hbi. Thus, aba−1 = b or aba−1 = b−1 . If aba−1 = b, then G is abelian, so G ∼ = Z6 . Assume that aba−1 = b−1 . Then G = ha, b : a2 , b3 , aba−1 = b−1 i. Note that S3 = h(12), (123)i and (12)(123)(12)−1 = (132) = (123)−1 . By von Dyck’s theorem, there is an onto homomorphism from G to S3 . But |G| = 6 = |S3 |, G ∼ = S3 . To close this section, we introduce a graphical representation of a group given by a set of generators and relations. The idea was originated by Cayley in 1878. It provided a method of visualizing a group and connects two important branches of modern mathematics—groups and graphs. This also has many applications to computer science. Let G be a finite group and S a nonempty subset of G. To avoid loops, we shall assume that e∈ / S. The Cayley digraph of G with generating set S is a digraph Cay(G, S) such that 1. each element of G is a vertex of Cay(G, S), and 2. for x, y ∈ G, there is a directed edge from x to y if and only if x−1 y ∈ S. / x2 / ··· / xn a directed path. A We call the finite sequence of vertices such that x1 directed graph is connected if for every pair of vertices x, z, there is a directed path from x to z. A maximal connected subdigraph of a digraph is called a connected component. Example 3.5.8. For the group (Z6 , +), we consider the following two digraphs. The first one is connected while the second one is not. The second one has two connected components. Cay(Z6 , {1̄}) / 1̄ ❁ @ 0̄ ✂✂ ✂✂ ✂ ✂ ✂✂ ❁❁ ❁❁ ❁❁ ❁ ❁❁ ❁❁ ❁❁ ❁ ✂ ✂✂ ✂ ✂✂ ✂✂ 5̄ ^❁ 4̄ o 3̄ Cay(Z6 , {2̄}) 0̄O ◆◆◆ ♣♣8 1̄ 2̄ ♣♣◆◆ ♣♣♣ ◆◆◆◆◆ ♣ ♣ ◆◆◆ ♣♣ ◆& ♣♣♣ 5̄ f◆◆◆ ♣ 2̄ ♣ ♣ ◆◆◆ ♣ ♣ ♣ ◆◆◆ ◆◆◆♣♣♣♣♣ x♣♣♣ ◆◆ 4̄ 3̄ 100 3. Advanced Group Theory Example 3.5.9. For the group (S3 , ◦), we consider the following three digraphs. The first and the second ones are not connected. They have three connected components and two connected component, respectively. The last one is connected. Note also that since (12)−1 = (12), some edges are undirected. Cay(S3 , {(12)}) (132) (23) (1) Cay(S3 , {(123)}) Cay(S3 , {(12), (123)}) ♣ ♣♣♣ ♣ ♣ ♣ ♣♣♣ ♣♣♣ (12) (13) ❖ ❖❖❖ ❖❖❖ ❖❖❖ ❖❖❖ (123) (132) `❇❇ ⑦ ❇❇ ⑦⑦ ❇❇ ⑦ ⑦ ❇❇ ⑦⑦ ❇❇ ⑦ ⑦ ❇❇ ⑦ (23) ⑦ ❇❇ < ❋❋ ⑦ ① ⑦ ❇❇ ❋ ① ❋❋ ① ⑦⑦ ❇❇ ① ⑦ ❋❋ ① ⑦ ① ❇❇ ⑦ ❋" ① ① ⑦ ❇❇ ⑦ ❇❇ ⑦⑦ (12) o (13) ⑦ ❇❇ ⑦ ⑦ ❇❇ ⑦ ❇❇ ⑦⑦ ⑦ ❇❇ ~⑦⑦ / (123) (1) (132) `❇❇ ⑦ ❇❇ ⑦⑦ ❇❇ ⑦ ⑦ ❇❇ ⑦⑦ ❇❇ ⑦ ⑦ ❇❇ ⑦ (23) ⑦ ❇❇ ❋❋ ①< ⑦⑦ ❇❇ ❋ ① ⑦ ❋ ① ⑦ ❋❋ ❇❇ ⑦ ①① ⑦ ❋ ① ❇❇ ⑦ ❋" ① ① ⑦ ❇❇ ⑦ ⑦ ❇❇ ⑦ (12) o (13) ❖ ⑦ ❖❖❖ ❇❇❇ ⑦ ♣♣♣ ⑦ ❖ ❖❖❖ ❇❇ ⑦⑦ ♣♣♣ ❖❖❖ ❇❇ ⑦⑦♣♣♣♣ ❖❖ ❇ ⑦ ~⑦♣♣♣ / (123) (1) For a non-empty subset S of a group G, we let S −1 = s−1 : s ∈ S. Observe that if we assume further that e ∈ / S and S = S −1 , then the Cayley graph Cay(G, S) is an undirected graph without loops. Hence, it is a graph. A regular graph is a graph such that each vertex has the same number of neighbors which is called the degree of a regular graph. Note that for each x ∈ G, we have x xs for all s ∈ S. From the cancellative property in G, each vertex has |S| neighbors, so Cay(G, S) is a regular graph of degree |S|. Theorem 3.5.9. Let G be a group and S a nonempty subset of G such that e ∈ / S and S = S −1 . Then 1. Cay(G, S) is a regular graph of degree |S|. 2. For x, z ∈ G, there is a path from x to z if and only if x−1 z ∈ hSi. 3. The number of connected components of Cay(G, S) is the index [G : hSi]. 4. Cay(G, S) is connected if and only if hSi = G. 101 3.5. Free Groups and Presentations Proof. Clearly, (4) follows from (3) and (3) follows from (2), respectively. To prove (2), let x, z ∈ G. Since S = S −1 , hSi = {s1 s2 . . . sk : k ∈ N ∪ {0} and s1 , s2 , . . . , sk ∈ S}. Also, any path from x to z is given by x xs1 ··· xs1 s2 xs1 s2 . . . sk = z for some s1 , s2 , . . . , sk . Thus, there is a path from x to z if and only if x−1 z = s1 s2 . . . sk ∈ hSi. Finally, we provides some examples of Cayley graph arising from number theory and ring theory. Example 3.5.10. For n ≥ 3, consider the additive group (Zn , +). We know that for ā ∈ Zn , × × × × ā ∈ Z× n ⇔ −ā ∈ Zn , so Zn = −Zn . The Cayley graph Xn = Cay(Zn , Zn ) is called the unitary Cayley graph. More generally, if R is a finite ring and consider (R, +) is an additive group, then the unitary Cayley graph for R is the Cayley graph XR = Cay(R, R× ). Example 3.5.11. Let R be a finite commutative ring with unity 1 6= 0. Consider the exact sequence of groups 1 / KR / R× θ / (R× )2 /1 where θ : a → a2 is the square mapping on R× with kernel KR and (R× )2 = {a2 : a ∈ R× }. Note that KR consists of the identity and all elements of order two in R× . Let TR = KR (R× )2 . Since −1 ∈ TR , the Cayley graphs HR = Cay(R, TR ) is undirected. This graph is a subgraph of the unitary Cayley graph. It is called the restricted unitary Cayley graphs induced from the square mapping. It is a generalization of Paley graphs (see Project 25). Exercises 3.5. 1. (a) Prove that the derived group of a free group consists of those words in which the −2 sum of the exponents for each generator is equal to zero (e.g., x1 x−1 2 x1 x2 x1 ). (b) Let F be a free group generated by x1 , x2 , . . . , xr . Show that each element of F/F ′ is of the form ′ ′ ∼ r ′ mr 1 m2 (xm 1 x2 . . . xr )F . Now use (a) to show that F/F = Z , i.e., F/F is the free abelian group of rank r. 2. Determine the structure of G/G′ , when G is given by (i) a6 = b2 = (ab)2 = 1; (ii) a6 = 1, b2 = (ab)2 = a3 . 3. Show that if G is generated by a and b subject to the relations a−1 ba = b2 and ab = ba2 , then G = {1}. 4. Let G be a group. For a, b ∈ G, let [a, b] = aba−1 b−1 and ab = bab−1 . (a) Prove that [a, bc] = [a, b][a, c]b for all a, b, c ∈ G. (b) If H = hx, y, z ∈ G : [x, y] = y, [y, z] = z and [z, x] = xi, show that H = {e}. 5. If G is a non-abelian group of order eight, show that G is isomorphic to D4 or Q8 . 6. Let D4 = ha, b : a2 = b2 = (ab)4 = 1i. Draw Cay(D4 , {a, b}). Why is this graph undirected? “Algebraic Graph Theory” is a branch of mathematics in which algebraic methods are applied to problems about graphs. There are three main branches of algebraic graph theory, involving the use of linear algebra, the use of group theory, and the study of graph invariants. The first branch of algebraic graph theory involves the study of graphs in connection with linear algebra. Mainly, it studies the spectrum of the adjacency matrix, or the Laplacian matrix of a graph (this part of algebraic graph theory is also called spectral graph theory). Secondly, algebraic graph theory involves the study of graphs in connection to group theory, particularly automorphism groups and geometric group theory. The focus is placed on various families of graphs based on symmetry such as symmetric graphs, vertex-transitive graphs, edge-transitive graphs, distancetransitive graphs, distance-regular graphs, and strongly regular graphs. Finally, the third branch of algebraic graph theory concerns algebraic properties of invariants of graphs, and especially the 102 3. Advanced Group Theory chromatic polynomial, the Tutte polynomial and knot invariants. The references and related work can be found in Godsil and Royle’s celebrated book [14]. There are many graphs arising from number theory and finite field theory such as Cayley graphs, symplectic graphs (Subsection 4.7.2), Paley graphs (see Project 25) and functional digraphs (e.g., [36, 37]). Moreover, these graphs can be extended to general results in abstract integral domains, e.g., function fields, PID or even UFD (see, [27] and the following project). Project 20 (gcd-graph). We consider a unique factorization domain D. Let c ∈ D be a nonzero nonunit element. Assume that the commutative ring D/(c) is finite. Let C be a set of proper divisor of c. Define the gcd-graph, Dc (C), to be a graph whose vertex set is the quotient ring D/(c) and edge set is {{x + (c), y + (c)} : x, y ∈ D and gcd(x − y, c) ∈ C}. The gcd considered here is unique up to associate. Prove that that Dc ({1}) = GD/(c) = Cay(D/(c), D/(c)× ), the unitary Cayley graph in Example 3.5.10. This gcd-graph on a quotient ring of a unique factorization domain (UFD) introduced in [27] generalizes a gcd-graph or an integral circulant graph (i.e., its adjacency matrix is circulant and all eigenvalues are integers) defined over Zn , n ≥ 2, (see [29, 39]). An integral circulant graph can also be considered as an extension of a unitary Cayley graph. Project 21 (Energy of a graph). Let G be a graph with vertex set {v1 , v2 , . . . , vn }. The adjacency matrix of G, denoted by A(G), is the n × n matrix given by ( 1 if there is an edge joining vi and vj , aij = 0 otherwise. The eigenvalues and eigenvectors of a graph G are defined to be the eigenvalues and eigenvectors of its adjacency matrix A(G). The sum of absolute values of all eigenvalues of a graph G is called the energy of G and denoted by E(G). Let R be a finite local ring with unique maximal ideal M of size m. For k, l ∈ N, we write 0k×l and Jk×l for the k × l matrix whose all entries are 0 and 1, respectively. We also use ~0k = 0k×1 and ~1k = Jk×1 . (a) Prove that the adjacency matrix of the unitary Cayley graph Cay(R, R× ) is given by 0m×m Jm×m Jm×m · · · Jm×m Jm×m 0m×m Jm×m · · · Jm×m × A(Cay(R, R )) = Jm×m Jm×m 0m×m · · · Jm×m . .. .. .. . . .. .. . . . Jm×m Jm×m Jm×m · · · 0m×m (b) Compute the eigenvalues and eigenvectors of the unitary Cayley graph Cay(R, R× ) and determine its energy. The energy is a graph parameter introduced by Gutman (see [25] and [21] for a good survey) arising from the Hückel molecular orbital approximation for the total π-electron energy. Nowadays, the energy of graph is studied for purely mathematical interest. 4 | Modules and Noetherian Rings Modules can be considered as a generalization of vector spaces. It is like we study linear algebra over a ring. In this chapter, we first cover basic concepts of modules. Next, we work on free modules. Projective and injective modules are introduced. We also present the proof of the structure theorems for modules over a PID. Finally, we talk about Noetherian and Artinion rings. Noetherian rings have a lot of applications in algebraic geometry and algebraic number theory. Each ring R that we consider will be assumed to contain a multiplicative identity element, which will be denoted by 1. We shall therefore regard the possession of such an identity as one of the defining conditions of the ring concept and also assume 1 6= 0. 4.1 Modules The definition of a module is similar to a vector space. However, now our scalars are in a ring. Let R be a ring. We say that M is a (left) R-module provided: 1. (M, +) is an additive abelian group 2. there is a multiplication R × M → M which satisfies for all α, β ∈ R and u, v ∈ M , (a) α(u + v) = αu + αv, (b) (α + β)u = αu + βu and (c) α(βu) = (αβ)u 3. if 1 is the unity of R, then 1 · u = u for all u ∈ M . Remark. Note that we abuse notations by not distinguishing between the addition in M or in R and the multiplication in R or the multiplication R × M → M . A right R-module can be defined analogously. Examples 4.1.1. 1. If R = F , a field, an F -module is just a vector space over F . 2. Any abelian group A is a Z-module, where the action of Z is given by for a ∈ A, 0 · a = 0A , n · a = |a + a + {z· · · + a} if n > 0 and n n · a = (−a) + (−a) + · · · + (−a) if n < 0. | {z } −n 3. Let F be a field and R = Mn (F ) the ring of n × n matrices over F . Let V = F n be ndimensional vector space of n × 1 column vectors over F . Then V is an R-module where the multiplication R × V → V is given by A · ~v = A~v (matrix multiplication). 4. Let R be a ring. Then R is an R-module with the usual multiplication R × R → R. More generally, any left ideal A of R is a left R-module. In fact, a subset A of R is a left ideal in R if and only if the left multiplication R × A → A makes A into a left R-module. That is, the set of left ideals of R is the set left R-modules of R. Hence, if R is a ring, then R can be viewed as an R-module, called a regular left [right] R-module, and is denoted by R R [RR ]. 103 104 4. Modules and Noetherian Rings We collect basic terminologies about modules in the following definitions. Let R be a ring. We say that N is an R-submodule or submodule of an R-module M if N is a subgroup of M as an additive group and the multiplications R × M → M and R × N → N agree on N . Let R be a ring. The direct sum of R-modules M and N is the abelian group direct sum of M and N M ⊕ N = {(m, n) : m ∈ M, n ∈ N } with the action of R on M ⊕ N given by r(m, n) = (rm, rn). One often writes m + n in place of (m, n). Let R be a ring and let M and N be R-modules. 1. A map f : M → N is an R-module homomorphism provided (a) f : M → N is a homomorphism of abelian groups and (b) if r ∈ R and m ∈ M , then f (rm) = rf (m). 2. We call a diagram of R-module homomorphisms M f /N g /P exact if im f = ker g. More generally, a sequence of R-modules and homomorphisms ··· / M1 f1 / M2 f2 / M3 / ··· that may be finite or run to infinity in either direction is called exact if for any three consec/ Mi+1 / Mi+2 is exact. An exact sequence of the utive terms the subsequence Mi form / M′ 0 f /M g / M ′′ /0 is called a short exact sequence. This means that f is a monomorphism (1-1), g is an epimorphism (onto) and ker g = im f . 3. If N is a submodule of M , then the quotient group (M/N, +) can be made into an R-module by defining r(x + N ) = rx + N . It is called a factor module of M by N . 4. Let f : M → N be a homomorphism of R-modules. The kernel of f is ker f = {m ∈ M : f (m) = 0N } and the cokernel of f is N/im f . They are clear that ker f and im f are R-submodules of M and N , respectively. Evidently, f is surjective if and only if coker f = 0. In any case, we have 0 / ker f /M f /N / coker f /0 is exact. Remark. The isomorphism theorems also hold for R-modules and their homomorphisms. Note however that the first isomorphism theorem will say a bit more, because coker f = N/im f is an R-module. This is not the case with homomorphisms of groups or rings: If f : G → H is a group homomorphism, then f (G) = im f is not in general a normal subgroup of H, hence H/im f is not in general a group. And if f : R → S is a ring homomorphism, then f (R) = im f is never an ideal in S (unless it is all of S), so S/im f is not a ring. The isomorphism theorems can be stated as theorems about commutative diagrams and exact sequences. The use of diagrams to describe module homomorphisms is very common, we now give the isomorphism theorems in their diagram theoretic versions. Note that many homomorphisms are projections or injections implicitly defined by the diagram. The proofs of the isomorphism theorems are left as exercises. 105 4.1. Modules Theorem 4.1.1. [First Isomorphism Theorem] Let M and N be R-modules. Then the following diagram of R-modules has an exact row and a commutative square. 0 / ker f f /M /N O π / coker f /0 i ∼ = f¯ M/ ker f / im f Theorem 4.1.2. [Second Isomorphism Theorem] Let N1 and N2 be submodules of an Rmodule N . Then there is a commutative diagram with exact rows in which the vertical map of the right is an isomorphism. 0 / N1 ∩ N2 / N2 / N2 /(N1 ∩ N2 ) /0 ∼ = / N1 0 / N1 + N2 / (N1 + N2 )/N1 /0 Theorem 4.1.3. [Third Isomorphism Theorem] If N2 ≤ N1 ≤ N are R-modules, then the following diagram is commutative and has exact rows: 0 / N1 /N / N/N1 /0 id / N1 /N2 0 / N/N2 / N/N1 /0 That is, N/N1 ∼ = (N/N2 )/(N1 /N2 ). Theorem 4.1.4. Let N1 and N2 be submodules of an R-module N . Then the following diagram is commutative and has exact rows and columns. 0 0 0 0 / N1 ∩ N2 / N2 / N2 /(N1 ∩ N2 ) /0 0 / N1 /N / N/N1 /0 0 / N1 /(N1 ∩ N2 ) / N/N2 / N/(N1 + N2 ) /0 0 0 0 Proof. The commutativity and the exactness of the top two rows and the two left columns are clear. The exactness of the third row and right column come, respectively, from the isomorphisms (N1 + N2 )/N2 ∼ = N2 /(N1 ∩ N2 ). = N1 /(N1 ∩ N2 ) and (N1 + N2 )/N1 ∼ The next theorem is widely used in mathematics. It is proved by the technique called “diagram chasing”. 106 4. Modules and Noetherian Rings Theorem 4.1.5. [5-Lemma] Suppose the following diagram is commutative and has exact rows. A1 α1 f1 / A2 α2 f2 B1 α3 f3 β1 / A3 / B2 α4 f4 β2 / A4 / B3 f5 β3 / A5 / B4 β4 / B5 If f1 , f2 , f4 and f5 are isomorphisms, so is f3 . More precisely, 1. if f1 is onto and f2 and f4 are 1-1, then f3 is 1-1, and 2. if f5 is 1-1 and f2 and f4 are onto, then f3 is onto. Proof. (1) Assume f1 is onto and f2 and f4 are 1-1. Suppose x ∈ A3 and f3 (x) = 0. We shall show that x = 0. Since f4 (α3 (x) = β3 (f3 (x)) = β3 (0) = 0 and f4 is 1-1, α3 (x) = 0, so x ∈ ker α3 = im α2 from the exactness of the top row. Thus, x = α2 (y) for some y ∈ A2 . Then 0 = f3 (x) = f3 (α2 (y)) = β2 (f2 (y)), so f2 (y) ∈ ker β2 = im β1 from the exactness of the bottom row. Thus, f2 (y) = β1 (z) for some z ∈ B1 . Since f1 is onto, there is a u ∈ A1 with f1 (u) = z. Then f2 (y) = β1 (z) = β1 (f1 (u)) = f2 (α1 (u)), so y = α1 (u) since f2 is 1-1. Hence, x = α2 (y) = α2 (α1 (u)) = 0 since α2 α1 = 0 by the exactness of the top row. (2) Assume f5 is 1-1 and f2 and f4 are onto. Let x ∈ B3 . We must find w ∈ A3 with f3 (w) = x. Since f4 is onto, we can choose y ∈ A4 with f4 (y) = β3 (x). Then f5 (α4 (y)) = β4 (f4 (y)) = β4 (β3 (x)) = 0 from the bottom row is exact. But f5 is 1-1, so α4 (y) = 0. Since the top row is exact, y = α3 (z) for some z ∈ A3 . Then β3 (x) = f4 (y) = f4 (α3 (z)) = β3 (f3 (z)), so β3 (x − f3 (z)) = 0. Thus, there is a u ∈ B2 with β2 (u) = x − f3 (z) from the bottom row is exact. Since f2 is onto, there is a v ∈ A2 with f2 (v) = u. Hence, x − f3 (z) = β2 (u) = β2 (f2 (v)) = f3 (α2 (v)), so x = f3 (z + α2 (v)) = f3 (w) where w = z + α2 (v). That is, f3 is onto. α β /L /M /N / 0 be a short exact Theorem 4.1.6. [Split Exact Sequence] Let 0 sequence of R-modules. Then the following three conditions are equivalent. (i) There exists an isomorphism M ∼ = L ⊕ N in which α : l 7→ (l, 0) and β : (l, n) 7→ n. (ii) There exists a section of β, that is, a homomorphism s : N → M such that β ◦ s = id N . (iiii) There exists a retraction of α, that is, a homomorphism r : M → L such that r ◦ α = id L . If this happens, the sequence is a split exact sequence. Proof. (i) ⇒ (ii) or (iii) is easy. (ii) ⇒ (i). The given section s is clearly injective because it has a left inverse; we claim that M = α(L) ⊕ s(N ). To see this, any m ∈ M is of the form m = (m − s(β(m))) + s(β(m)), where the second term is obviously in s(N ); since β ◦ s = id N , the first term is clearly in ker β, and by exactness this is α(L). Furthermore, α(L) ∩ s(N ) = {0}, since if n ∈ N is such that s(n) ∈ α(L) = ker β then n = β(s(n)) = 0. (iii) ⇒ (i) is similar to (ii) ⇒ (i) and left as an exercise. For finite dimensional vector spaces over a field, every subspace has complement, so every short exact sequence splits. Whether an exact sequence splits or not depends on what ring it is considered over. For example, 0 is split over k but not over k[x]. / k[x]x / k[x] / k[x]/k[x]x /0 107 4.2. Free Modules and Matrices Proof. It is easy to see that k[x] ∼ = k ⊕ k[x]x as k-vector spaces. Note that k is a k[x]-module, where the ϕ scalar multiplication is given by (a0 + a1 x + · · · + an xn )c = a0 c for all c, ai ∈ k. Assume that k[x] ∼ = k ⊕ k[x]x as k[x]-modules and ϕ : 1 7→ (a0 , a1 x + · · · + an xn ). Then for all m ∈ N ∪ {0} and b0 , b1 , . . . , bm ∈ k, ϕ (b0 + b1 x + · · · + bm xm ) 7→ (b0 a0 , (b0 + b1 x + · · · + bm xm )(a1 x + · · · + an xn )). Since ϕ is 1-1, a0 6= 0. Since ϕ is onto, a2 , . . . , an = 0 and a1 6= 0. Thus, we reduce the above map to ϕ (b0 + b1 x + · · · + bm xm ) 7→ (b0 a0 , a1 (b0 x + b1 x2 + · · · + bm xm+1 )). Consider (0, x) in k + k[x]x. If ϕ(b0 + b1 x + · · · + bm xm ) = (0, x), then b0 a0 = 0 and a1 b0 = 1 which is impossible because a0 and a1 are nonzero. Hence, ϕ is not onto which is a contradiction. Exercises 4.1. 1. Let M1 and M2 be R-submodules of an R-module M . Define φ : M1 ×M2 → M1 +M2 by φ(m1 , m2 ) = m1 + m2 . Prove that φ is an isomorphism if and only if M1 ∩ M2 = {0}. Let R be a ring, I an ideal of R and M an R-module. Prove that ( n ) X IM := ri xi : n ≥ 1, ri ∈ I, xi ∈ M i=1 is an R-submodule of M and M/IM is an R/I-module by the scalar multiplication defined by (r + I)(x + IM ) := rx + IM . 2. Complete the proof of Theorem 4.1.6. 3. (a) If φ : M → M be an R-module homomorphism such that φ ◦ φ = φ, prove that M = ker φ ⊕ im φ. (b) If α : M → N and β : N → M are R-module homomorphisms such that β ◦ α = id M , prove that N = im α ⊕ ker β. 4.2 Free Modules and Matrices Like a finite dimensional vector space over a field, we shall see in this section that a free module (i.e., a module with basis) over a commutative ring behaves in a similar way. Let M1 , . . . , Mk be R-modules. The direct sum of M1 , . . . , Mk is the set of k-tuples {(m1 , . . . , mk ) : mi ∈ Mi } with the following operations: (m1 , . . . , mk ) + (n1 , . . . , nk ) = (m1 + n1 , . . . , mk + nk ) r(m1 , . . . , mk ) = (rm1 , . . . , rmk ), r ∈ R. The direct sum of M1 , . . . , Mk is denoted by M1 ⊕ · · · ⊕ Mk or k M Mi . i=1 Let M1 , . . . , Mk be submodules of an R-module M . The sum of M1 , . . . , Mk is the set {m1 + · · · + mk : mi ∈ Mi for all i}, denoted by M1 + · · · + Mk or n X Mi . It is a i=1 MX i with m1 + · · · independent if for any mi ∈ condition is equivalent to Mi ∩ j6=i submodule of M . We say that M1 , . . . , Mk are + mk = 0, we have m1 = · · · = mk = 0M . This Mi = {0M } for all i. 108 4. Modules and Noetherian Rings Theorem 4.2.1. Let M1 , . . . , Mk be submodules of an R-module M . Then 1. The map φ : M1 ⊕ · · · ⊕ Mk → M defined by φ(m1 , . . . , mk ) = m1 + · · · + mk is an R-module homomorphism whose image is M1 + · · · + Mk . 2. φ is one-to-one if and only if M1 , . . . , Mk are independent submodules of M . In case φ is an isomorphism, we say that M is the internal direct sum of the submodules M1 , . . . , Mk . Let M be an R-module and X a subset of M . The submodule of M generated by X, denoted by RX, is the set of all finite sums {r1 x1 + · · · + rk xk : ri ∈ R and xi ∈ X}. If RX = M , we say that X generates or spans M . If some finite subset {x1 , . . . , xk } of M generates M , we say that M is finitely generated M and we write M = Rx1 + · · · + Rxk . If M is generated by a single element, i.e., if M = Rx for some x ∈ M , M is said to be cyclic. We say that x1 , . . . , xk ∈ M are linearly independent over R if for any r1 , . . . , rk ∈ R with r1 x1 + · · · + rk xk = 0M , we have r1 = · · · = rk = 0. A subset X (possibly infinite) of M is linearly independent if every finite subset of X is linearly independent. We say that a set X is linearly dependent if it is not linearly independent. Remarks. 1. By convention, the empty set is linearly independent and R∅ = {0M }. 2. If x ∈ M , {x} is a linearly independent set if and only if Rx ∼ = R as left R-modules. In particular, if we take M = R, a left R-module, then {x} is a linearly independent set (where x ∈ R) if and only if x is not a right zero divisor, i.e., a 6= 0 ⇒ ax 6= 0. · · ⊕ R} as left 3. If {x1 , . . . , xk } is a linearly independent set, then Rx1 + · · · + Rxk ∼ = R | ⊕ ·{z R-modules. 4. Any subset of a linearly independent set is a linearly independent set. k If an R-module M is generated by a linearly independent set X, we say that M is the free R-module on the set X and that X is a basis for M . If X = {x1 , . . . , xk } is a finite set, we say that M is the finitely generated free module spanned by x1 , . . . , xk . Now let M = Rx1 + · · · + Rxn be the free R-module on the set X = {x1 , . . . , xn }. Suppose N is any left R-module and y1 , . . . , yn are any elements of N . Let us define a map φ : M → N by φ(r1 x1 + · · · + rn xn ) = r1 y1 + · · · + rn yn . Then φ is a homomorphism of left R-modules such that φ(xi ) = yi for all i. In fact, we could also define a homomorphism even if X were infinite. The point is that any set map X → N gives rise to an R-module homomorphism M → N . More precisely, Theorem 4.2.2. [Universal Mapping Property of a Free Module] Let R be a ring, X a set and M = M (X) the free R-module on the set X. Let i : X → M be defined by i(x) = 1 · x for all x ∈ X. (i may be thought of as an inclusion map.) Suppose N is an R-module and α : X → N is a set map. Then there exists a unique R-module homomorphism θ : M → N such that θ ◦ i = α. i / M ❇❇ ❇❇ α ❇❇❇ θ X❇ N Hence, any module is a homomorphic image of a free module. 109 4.2. Free Modules and Matrices Next, let us consider homomorphism of finitely generated free R-modules. Suppose M and N are free R-modules with bases X = {x1 , . . . , xm } and Y = {y1 , . . . , yn } where M = Rx1 + · · · + Rxm and N = Ry1 + · · · + Ryn . Let φ : M → N be an R-module homomorphism. Then φ will be completely defined as soon as we specify φ(x1 ), . . . , φ(xm ). Moreover, by the above theorem, any choice of φ(x1 ), . . . , φ(xm ) is possible. Hence, φ ↔ (φ(x1 ), . . . , φ(xm )) is a 1-1 correspondence between the set of R-module homomorphisms φ : M → N and N ×· · ·×N (m-copies). We have not written N ⊕ · · · ⊕ N because so far this correspondence is only a 1-1 correspondence of sets. We do not know if any structure is preserved. Let M and N be R-modules. The set of R-module homomorphisms from M to N is denoted by homR (M, N ). Remarks. 1. homR (M, N ) is an abelian group with the addition given by (φ + θ)(m) = φ(m) + θ(m). 2. If R is commutative, then we can make homR (M, N ) into a left R-module by defining (rφ)(m) = rφ(m). Note that rφ : M → N is really an R-module homomorphism, for if m ∈ M, s ∈ R, then (rφ)(sm) = r(φ(sm)) = r(sφ(m)) = (rs)φ(m) = (sr)φ(m) = s(rφ(m)) = s[(rφ)(m)]. However, this computation makes it clear that the commutativity of R is essential. If R is not commutative, there is no natural way to make homR (M, N ) into a left R-module. Let us restate the remarks above in the next theorem. Theorem 4.2.3. Let M and N be left R-modules. 1. homR (M, N ) is an abelian group (or Z-module) with addition (φ + θ)(m) = φ(m) + θ(m). 2. If R is commutative, homR (M, N ) is a left R-module, where (rφ)(m) = rφ(m). 3. If M = Rx1 + . . . Rxm is the free R-module with basis x1 , . . . , xm , then homR (M, N ) −→ N ⊕ · · · ⊕ N φ 7−→ (φ(x1 ), . . . , φ(xm )) is an isomorphism of abelian groups. If R is commutative, it is an isomorphism of R-modules. For k ≥ 1 and a ring R, let Rk denote the R-module of k × 1 column vectors over R. Now let us return to free R-modules M = Rx1 + · · · + Rxm and N = Ry1 + · · · + Ryn . As noted earlier, if φ : M → N is an R-module homomorphism, then φ is completely determined by φ(x1 ), . . . , φ(xm ), and φ 7→ (φ(x1 ), . . . , φ(xm )) is an isomorphism of abelian groups, and it is an isomorphism of R-modules if R is commutative. Since N = Ry1 + · · · + Ryn is free on y1 , . . . , yn every element of N can be uniquely expressed in the form y = r 1 y1 + . . . r n y n . In particular, we can write φ(x1 ) = a11 y1 + a21 y2 + · · · + an1 yn φ(x2 ) = a12 y1 + a22 y2 + · · · + an2 yn .. . φ(xm ) = a1m y1 + a2m y2 + · · · + anm yn . 110 4. Modules and Noetherian Rings In this way, we have abelian group isomorphisms homR (M, N ) −→ φ N ⊕ ··· ⊕ N −→ n × m matrices over R a11 a12 . . . a1m a21 a22 . . . a2m 7 → (φ(x1 ), . . . , φ(xm )) 7−→ . − .. . .. . an1 an2 . . . anm Moreover, in case R is commutative, this is an isomorphism of left R-modules. Next, let R be a commutative ring and M = Rx1 + · · · + Rxm , N = Ry1 + · · · + Ryn and P = Rz1 + · · · + Rzp be finitely generated free modules over R with the indicated free generators. Let α : M → Rm , β : N → Rn and γ : P → Rp be the R-module isomorphisms r1 s1 t1 .. .. .. α(r1 x1 + · · · + rm xm ) = . , β(s1 y1 + · · · + sn yn ) = . , γ(t1 z1 + · · · + tp zp ) = . . rm sn tp Write Ruv for the R-module of u × v matrices over R. For each R-module homomorphism φ : M → N , we define a11 a12 . . . a1m a21 a22 . . . a2m [φ] = . .. ∈ Rnm .. . an1 an2 . . . anm implicitly from the equations φ(x1 ) = a11 y1 + a21 y2 + · · · + an1 yn φ(x2 ) = a12 y1 + a22 y2 + · · · + an2 yn .. . φ(xm ) = a1m y1 + a2m y2 + · · · + anm yn . Similarly, for R-module homomorphisms θ : N → P and τ : M → P define [θ] ∈ Rpn and [τ ] ∈ Rpm , respectively. Then φ 7→ [φ], θ 7→ [θ] and τ 7→ [τ ] are isomorphisms of R- modules homR (M, N ) ∼ = Rnm , homR (N, P ) ∼ = Rpn and homR (M, P ) ∼ = Rpm , respectively. Moreover, we obtain the following theorem. Theorem 4.2.4. Let R be a commutative ring. Under the above set-up we have: 1. Each matrix [φ] ∈ Rnm defines a homomorphism [φ] : Rm → Rn by left multiplication of an n × m matrix by an m × 1 matrix. The same is true for [θ] : Rn → Rp and [τ ] : Rm → Rp . 2. The following diagram is commutative M φ /N α ∼ = β ∼ = Rm [φ] / Rn θ /P γ ∼ = [θ] / Rp : [θφ] In particular, [θ][φ] = [θφ] where the left product is multiplication of matrices. Recall the following fact about matrices: Let R be a commutative ring and suppose A ∈ Mn (R). Then A is invertible ⇔ A is left invertible ⇔ A is right invertible ⇔ det A is a unit in R. In particular, if AC = I, then CA = I. Moreover, we have 111 4.2. Free Modules and Matrices Theorem 4.2.5. Let R be a commutative ring and let [φ] ∈ Rmn and [θ] ∈ Rnm . Suppose [φ][θ] = Im and [θ][φ] = In are identity matrices of sizes m × m and n × n, respectively. Then m = n. Proof. Assume that m > n. Then m = n + r for some r ∈ N. Write An×n [φ] = and [θ] = Cn×n Dn×r , Br×n so AC AD In 0 [φ][θ] = = . BC BD 0 Ir Thus, 0 = C(AD) = (CA)D = In D = D which contradicts BD = Ir . Hence, m ≤ n. Similarly, we obtain a contradiction if m < n. Therefore, m = n. Theorem 4.2.6. Let R be a commutative ring and suppose that M = Rx1 + · · · + Rxm and N = Ry1 + · · · + Ryn are free R-modules with indicated generators. If M and N are isomorphic R-modules, then m = n. Proof. Let φ : M → N be an isomorphism with inverse θ : N → M . By Theorem 4.2.4, we can identify M with m × 1 column vectors and N with n × 1 column vectors and obtain a commutative diagram M φ /N ∼ = Rm θ /M [θ] ∼ = [φ] / Rn ∼ = / Rm 9 [θ][φ] In other words, [φ] is an n × m matrix and [θ] is an m × n matrix with [φ][θ] = In and [θ][φ] = Im . Hence, m = n by Theorem 4.2.5. If M = Rx1 + · · · + Rxm is a free R-module on x1 , . . . , xm over a commutative ring R, m is called the rank of M . In particular, if R = Z, then M is a free abelian group on x1 , . . . , xm , and hence we have shown: Corollary 4.2.7. If F is a finitely generated free abelian group, then any two bases of F have the same number of elements. Using this corollary, we can verify that if a group G is free on A and also on B, which are finite sets, then the sets A and B have the same number of elements. It is Theorem 3.5.4 for the finite basis case. Proof of Theorem 3.5.4 for the finite basis case. Assume that G is a free group on A and also on B, where A and B are finite sets. By Exercise 1, G/G′ is a free abelian group of rank |A| and |B|, respectively. By Corollary 4.2.7, |A| = |B|. Remarks. 1. As we have seen that subgroups of a free (abelian) group are free. This is not true for general R-modules. For example, let R = Z6 . Then R R is a free R-module generated by {1}. N = {0, 2, 4} is an R-submodule of R R. Since ∅ does not span N , ∅ is not a basis. If B 6= ∅ is a basis of N , then 0 ∈ / B, so 2 or 4 are in B. Since 3 · 2 = 0 and 3 · 4 = 0 where 3 6= 0, where B is not linearly independent. Hence, submodules of a free module may not be free. 2. In the case of free abelian groups and vector spaces, it is true that any two bases of have the same cardinality. This is not true in general as shown in the following example. 112 4. Modules and Noetherian Rings Example 4.2.1. Let S be a ring and F a free S-module with infinite denumerable basis {e1 , e2 , e3 , . . . }. Let R = homS (F, F ). Then R is a ring with identity 1R , so {1R } is a basis for R R. Next, we define f1 , f2 ∈ R as follows: f1 (e2n ) = en , f1 (e2n−1 ) = 0 and f2 (e2n ) = 0, f2 (e2n−1 ) = en . To show that {f1 , f2 } spans R R, let g ∈ R. Define g1 , g2 ∈ R by g1 (en ) = g(e2n ) and g2 (en ) = g(e2n−1 ). Then (g1 f1 + g2 f2 )(e2n−1 ) = g1 f1 (e2n−1 ) + g2 f2 (e2n−1 ) = g2 (en ) = g(e2n−1 ) and (g1 f1 + g2 f2 )(e2n ) = g1 f1 (e2n ) + g2 f2 (e2n ) = g1 (en ) = g(e2n ). Thus, g = g1 f1 + g2 f2 . Next we shall prove that {f1 , f2 } is linearly independent over R. Let h1 , h2 ∈ R such that h1 f1 + h2 f2 = 0. Then for any n ≥ 1, h1 (en ) = h1 (en ) + 0 = h1 f1 (e2n ) + h2 f2 (e2n ) = (h1 f1 + h2 f2 )(e2n ) = 0 and h2 (en ) = 0 + h2 (en ) = h1 f1 (e2n−1 ) + h2 f2 (e2n−1 ) = (h1 f1 + h2 f2 )(e2n−1 ) = 0, so h1 = h2 = 0. Hence, {f1 , f2 } is linearly independent and so it is a basis of R R. Exercises 4.2. 1. Show that Q is not a free Z-module. 2. Show that M is a cyclic left R-module if and only if it is isomorphic to R/I (considered as a left R-module) for some left ideal I of R. P 3. Show L that {ei }i∈I is a basis of a left R-module M if and only if (ri )i∈I 7→ i∈I ri ei is an isomorphism of i∈I R R onto M . 4. Prove that the module R R in Example 4.2.1 has a basis with m elements for every positive integers m. 5. Let R be a ring and M, N and N ′ R-modules. Then homR (M, N ) and homR (M, N ′ ) are Z-modules. For an R-module homomorphism f : N → N ′ , we define hom(M, −)(f ) : homR (M, N ) → homR (M, N ′ ) by hom(M, −)(f )(h) = f ◦ h for all h ∈ homR (M, N ). Show that (a) hom(M, −)(f ) is a Z-module homomorphism from homR (M, N ) to homR (M, N ′ ). f g (b) If 0 → N → N ′ → N ′′ is exact, then 0 / homR (M, N ) hom(M,−)(f ) / homR (M, N ′ ) hom(M,−)(g) / homR (M, N ′′ ) is exact. f g In a similar manner, one can prove that exactness of N → N ′ → N ′′ → 0 implies exactness of 0 / homR (N ′′ , M ) hom(−,M )(g) / homR (N ′ , M ) hom(−,M )(f ) / homR (N, M ) where hom(−, M )(f )(h) = h ◦ f for all h ∈ homR (N ′ , M ) and hom(−, M )(g)(h) = h ◦ g for all h ∈ homR (N ′′ , M ). 6. Let R be a ring, I a proper ideal of R and F a free R-module with a basis X. Then F/IF is a free R/I-module with a basis of cardinality |X|. 4.3 Projective and Injective Modules The concept of projective modules is a generalization of the idea of a free module. Injective modules, introduced by Baer, are dual to that of projective modules. We follow [6] for this section. Let R be a ring. An R-module P is called projective if given any diagram P f M p //N 113 4.3. Projective and Injective Modules there exists a homomorphism g : P → M such that P g M f ~ //N p is commutative. In other words, given an epimorphism p : M → N , then any homomorphism f : P → N can be factored as f = pg for some g : P → M . 0 / homR (M, N ′ ) i / N′ We recall that for any module M , if 0 hom(M,i) p /N / N ′′ hom(M,p) / homR (M, N ) / 0 is exact, then / homR (M, N ′′ ) is exact. Now suppose M = P is projective. Then given f ∈ hom(P, N ′′ ) there exists a g ∈ homR (P, N ) such that homR (P, p)(g) = pg = f . Thus, in this case, hom(P, p) is surjective and so we actually have the exactness of / homR (M, N ′ ) 0 hom(M,i) / homR (M, N ) i / N′ as a consequence of the exactness of 0 hom(M,p) p /N / N ′′ / homR (M, N ′′ ) /0 /0. The converse holds also. Suppose hom(P, −) is exact and suppose M p / / N . Let K = ker p. p i /M /N / 0 where i is the inclusion /K Then we have the exact sequence 0 map. Applying the exactness of hom(P, −), we obtain the property of a projective module. Therefore, Theorem 4.3.1. Let P be an R-module. Then P is projective if and only if for any R-modules N, N ′ and N ′′ , if 0 0 / N′ i / homR (P, N ′ ) /N hom(P,i) p / N ′′ / 0 is a short exact sequence, then / homR (P, N ) hom(P,p) / homR (P, N ′′ ) /0 is also a short exact sequence of Z-modules. By Theorem 4.2.2, we have: Theorem 4.3.2. Every free module is projective. Example 4.3.1. Q is not a projective Z-module. Proof. Let F be a free Z-module with countable basis X = {x1 , x2 , . . . }. Define g : X → Q by 1 n g : xn 7→ for all n ∈ N. Then g induces a Z-module homomorphism from F to Q. Since g(mxn ) = n ∈ N, g is onto. Assume that Q is projective. Q h F id Q g //Q m n for all m ∈ Z and 114 4. Modules and Noetherian Rings X Then there exists an h : Q → F such that gh = id Q . Suppose h(1) = ai xi (with all but finite i Y X ai = 0). Let k = 1 + |ai | and assume that h(k −1 ) = bi xi (again, with all but finite ai = 0). i i,ai 6=0 Then so X i X kbi xi = k i X bi xi = kh(k −1 ) = h(1) = i X ai xi , i (ai − kbi )xi = 0. Since X is linearly independent, ai = kbi for all i which implies k | ai for all i. This forces k = 1 and ai = 0 for all i. Thus, h is the zero map which contradicts gh = id Q . Hence, Q is not projective. How close are projective modules to being free? We shall give two important characterizations of projective modules as follows. Theorem 4.3.3. The following properties of a module P are equivalent: (i) P is projective. /N /P / 0 splits. /M (ii) Any short exact sequence 0 (iii) P is a direct summand of a free module (that is, there exists a free module F isomorphic to P ⊕ P ′ for some P ′ ). Proof. (i) ⇒ (ii). Let 0 /M f /N g /P / 0 be exact and consider the diagram P id P N //P g By hypothesis we can fill this in with g ′ : P → N to obtain a commutative diagram. Then gg ′ = id P and the given short exact sequence splits. (ii) ⇒ (iii). Since any module is a homomorphic image of a free module (Theorem 4.2.2), we p i / P′ /F /P / 0 where F is a free module. If P have a short exact sequence 0 satisfies property (ii), then this exact sequence splits and hence F ∼ = P ⊕ P ′. (iii) ⇒ (i). We are given that there exists a sequence 0 is free. Now suppose we have a diagram / P′ i /F p /P / 0 with F P f M q //N Combining the two diagrams, we obtain 0 / P′ i p / /F o ❆❆ ′ P ❆❆i ❆❆ f f p ❆❆ //N M /0 q where pi′ = id P (since the top line splits). Since F is free, hence projective, we can fill in g : F → M to obtain f p = qg. Then f = f id P = f pi′ = qgi′ and gi′ : P → M make P ⑤⑤ ⑤⑤ f ⑤ ⑤ ⑤ ~⑤ M q //N gi′ 115 4.3. Projective and Injective Modules commutative. Hence, P is projective. Of particular interest are the modules that are finitely generated and projective. The theorem gives the following characterization of these modules. Corollary 4.3.4. A module P is finitely generated and projective if and only if P is a direct summand of a free module with a finite base. Proof. If P is a direct summand of a free module F with finite base, then P is projective. Moreover, P is a homomorphic image of F , so P has a finite set of generators (the images of the base under an epimorphism of F onto P ). Conversely, suppose P is finitely generated and projective. Then the first condition implies that we have an exact sequence 0 → P ′ → F → P → 0 where F is free with finite base. The proof of the theorem shows that if P is projective, then F ∼ = P ⊕ P ′ , so P is a direct summand of a free module with finite base. The concept of a projective module has a dual obtained by reversing the arrows in the definition as follows. An R-module Q is called injective if given any diagram of homomorphisms 0 i /N /M f Q there exists a homomorphism g : M → Q such that the diagram obtained by filling in g is commutative. In other words, given f : N → Q and a monomorphism i : N → M there exists a g : M → Q such that f = gi. With a slight change of notation, the definition amounts to this: Given an exact sequence i 0 → N ′ → N , the sequence homR (N, Q) hom(i,Q) / homR (N ′ , Q) i /0 p is exact. Since we know that exactness of 0 → N ′ → N → N ′′ → 0 implies exactness of 0 / homR (N ′′ , M ) hom(p,M ) / homR (N, M ) hom(i,M ) / homR (N ′ , M ) , it is clear that Q is injective if and only if hom(−, Q) is exact in the sense that it maps any short exact sequence 0 → N ′ → N → N ′′ → 0 into a short exact sequence of Z-module 0 → homR (N ′′ , Q) → homR (N, Q) → homR (N ′ , Q) → 0. It is easily seen also that the definition of injective is equivalent to the following: If N is a submodule of a module M , then any homomorphism of N into Q can be extended to a homomorphism of M into Q. Another result, which is easily established by dualizing the proof of the analogous result on projective (Theorem 4.3.3), is that if Q is injective, then any short exact sequence 0 → Q → M → N → 0 splits. The converse of this holds also. However, the proof requires the dual of the easy result that any module is a homomorphic image of a projective module (in fact, a free module). The dual statement is that any module can be embedded in an injective one. We shall see that this is the case, but the proof will turn out to be fairly difficult. Theorem 4.3.5. [Baer] A right module Q is injective if and only if any homomorphism of a right ideal I of R into Q can be extended to a homomorphism of R into Q. 116 4. Modules and Noetherian Rings Proof. Obviously, the condition is necessary. Now suppose it holds and suppose M is a module and f is a homomorphism of a submodule N of M into Q. Consider the set {(g, M ′ )} where M ′ is a submodule of M containing N and g is a homomorphism of M ′ into Q such that g|N = f . We define a partial order in the set {(g, M ′ )} by declaring that (g1 , M1′ ) ≥ (g2 , M2′ ) if M1′ ⊃ M2′ and g1 |M2′ = g2 . It is clear that any totally ordered subset has an upper bound in this set. Hence, by Zorn’s lemma, there exists a maximal (g, M ′ ); that is, we have an extension of f to a homomorphism g of M ′ ⊃ N which is maximal in the sense that if g1 is a homomorphism of an M1′ ⊃ M ′ such that g1 |M ′ = g, then necessarily M1′ = M ′ . We claim that M ′ = M . Otherwise, there is an x ∈ M, ∈ / M ′ and so xR + M ′ is a submodule of M properly containing M ′ . Now let I = {s ∈ R : xs ∈ M ′ }. Then I = ann(x + M ′ ) in M/M ′ , so I is a right ideal of R. If s ∈ I, then xs ∈ M ′ , so g(xs) ∈ Q. It is immediate that the map h : s 7→ g(xs) is a module homomorphism of I into Q. Hence, by hypothesis, h can be extended to a homomorphism k of R into Q. We shall use this to obtain an extension of g to a homomorphism of xR + M ′ to Q. The elements of xR + M ′ have the form xr + y, r ∈ R, y ∈ M ′ . If we have a relation xs + y ′ = 0, s ∈ R, y ′ ∈ M ′ , then s ∈ I. Then k(s) = h(s) = g(xs) = −g(y ′ ). Thus, xs + y ′ = 0 for s ∈ R, y ′ ∈ M ′ , implies that k(s) + g(y ′ ) = 0. It follows that xr + y 7→ k(r) + g(y), r ∈ R, y ∈ M ′ , is a well defined map. For, if xr1 + y1 = xr2 + y2 , ri ∈ R, yi ∈ M ′ , then xs + y ′ = 0 for s = r1 − r2 , y ′ = y1 − y2 . Then k(s) + g(y ′ ) = 0 and k(r1 − r2 ) + g(y1 − y2 ) = 0. Since k and g are homomorphisms, this implies that k(r1 ) + g(y1 ) = k(r2 ) + g(y2 ). It is immediate that the map rx+y 7→ k(r)+g(y) is a module homomorphism of xR+M ′ into Q extending the homomorphism g of M ′ . This contradicts the maximality of (g, M ′ ). Hence, M ′ = M and we have proved that if f is a homomorphism of a submodule N of M into Q, then f can be extended to a homomorphism of M into Q. Hence, Q is injective. For certain “nice” rings, the concept of injectivity of modules is closely related to the simpler notion of divisibility, which we proceed to define. If a ∈ R, then the module M is said to be divisible by a if the map x 7→ xa of M into M is surjective. A module is called divisible if it is divisible by every a 6= 0. It is clear that if M is divisible by a or if M is divisible, then any homomorphic image of M has the same property. In some sense injectivity is generalization of divisibility, for we have Theorem 4.3.6. 1. If R has no zero divisors 6= 0, then any injective R-module is divisible. 2. If R is a ring such that every right ideal of R is principal (= aR for some a ∈ R), then any divisible R-module is injective. Proof. (1) Suppose R has no zero-divisors 6= 0 and let Q be an injective R-module. Let x ∈ Q, r ∈ R, r 6= 0. If a, b ∈ R and ra = rb, then a = b. Hence, we have a well defined map ra 7→ xa, a ∈ R, of the right ideal rR into Q. Clearly this is a module homomorphism. Since Q is injective, the map ra 7→ xa can be extended to a homomorphism of R into Q. If 1 7→ y under this extension, then r = 1r 7→ yr. Since r = r1 7→ x1 = x, we have x = yr. Since x was arbitrary in Q and r was any non-zero element of R, this shows that Q is divisible. (2) Suppose R is a ring in which every right ideal is principal. Let M be a divisible R-module and let f be a homomorphism of the right ideal rR into M . If r = 0, then f is the zero map and this can be extended to the zero map of R. If r 6= 0 and f (r) = x ∈ M , then there exists a y in M such that x = yr. Then a 7→ ya is a module homomorphism of R into M and since 117 4.3. Projective and Injective Modules rb 7→ yrb = xb = f (r)b = f (rb), a 7→ ya is an extension of f . Thus, any module homomorphism of a right ideal of R into M can be extended to a homomorphism of R. Hence, M is injective by Baer’s criterion. If R satisfies both conditions stated in the theorem, then an R-module is injective if and only if it is divisible. In particular, this holds if R is a PID. We can use this to construct some examples of injective modules. Examples 4.3.2. 1. Let R be a subring of a field F and regard F as an R-module in the natural way. Evidently F is a divisible R-module. Hence, if K is any R-submodule of F , then F/K is a divisible R-module. In particular, Q is an injective Z-module which is not projective. 2. Let D be a PID, F its field of fractions. If r ∈ D, then the D-module F/rD is divisible and hence is injective by Theorem 4.3.6. Our next objective is to prove that any module can be embedded in an injective module, that i is, given any M there exists an exact sequence 0 → M → Q with Q is injective. The first step in the proof we shall give is as follows. Lemma 4.3.7. Any abelian group can be embedded in a divisible group (= a divisible Z-module). Proof. First let F be a free abelian group with base {xα } and F ′ the vector space over Q with {xα } as base. Then F is embedded in F ′ and it is clear that F ′ is divisible. Now let M be an arbitrary abelian group. Then M is isomorphic to a factor group F/K of a free abelian group F . Hence, F ′ /K is a divisible group and F/K ∼ = M is a subgroup. An immediate consequence of this and Theorem 4.3.6 is the next corollary. Corollary 4.3.8. Any Z-module can be embedded in an injective Z-module. Now for an arbitrary R-module M , we have the isomorphism of M onto homR (R, M ) which maps an element x ∈ M into the homomorphism fx such that 1 7→ x. This is an R-isomorphism if we make homR (R, M ) into a right R-module by defining f a, a ∈ R, by (f a)(b) = f (ab). Also homZ (R, M ) is a right R-module using this definition of f a. Clearly homR (R, M ) is a submodule of homZ (R, M ). Since M is isomorphic to homR (R, M ), we have an embedding of M in homZ (R, M ). Now embed M in an injective Z-module Q, which can be done by the foregoing corollary. Then we have an embedding of homZ (R, Q) as R-modules. This gives an embedding of M in an injective R-module, since we have the following lemma. Lemma 4.3.9. If Q is an injective Z-module, then homZ (R, Q) is an injective R-module. f Proof. We must show that if 0 → N ′ → N is an exact sequence of R-modules, then f∗ homR (N, homZ (R, Q)) → homR (N ′ , homZ (R, Q)) → 0 is exact, where f ∗ = homR (f, homZ (R, Q)). We have an isomorphism ϕN : homZ (N ⊗R R, Q) → homR (N, homZ (R, Q)) and the definition shows that this is “natural” in N . Since the isomorphism of N ⊗R R onto N such that y ⊗ 1 7→ y is natural in N , we have an isomorphism ψN : homZ (N, Q) → homR (N, homZ (R, Q)) 118 4. Modules and Noetherian Rings which is natural in N , that is we have the commutativity of homZ (N, Q) ψN / homR (N, homZ (R, Q)) f¯ f∗ homZ (N ′ , Q) / homR (N ′ , homZ (R, Q)) ψN ′ where f¯ = hom(f, Q). Now f¯ is surjective since Q is Z-injective. Since ψN and ψN ′ are isomorphisms, this implies that f ∗ is surjective. The foregoing lemma completes the proof of the embedding theorem. Theorem 4.3.10. Any module can be embedded in an injective module. The proof we have given is due to B. Eckmann and A. Schöpf. We can apply the theorem to complete the following characterization of injectives, which we idicated earlier. Theorem 4.3.11. The following properties of a module Q are equivalent: (i) Q is injective. (ii) Any short exact sequence 0 → Q → M → N → 0 splits. (iii) Q is a direct summand of every module containing it as a submodule. Proof. We leave the proof of (i) ⇒ (ii) as an exercise. Conversely, suppose any short exact sequence 0 → Q → M → N → 0 splits. By the embedding theorem we have an exact sequence p i i 0 → Q → M where M is injective. Then we have the short exact sequence 0 → Q → M → M/Q → 0 where p is the canonical homomorphism of M onto M/Q. By hypothesis, we can find a p′ : M → Q such that p′ i = id Q . Now suppose we have a diagram 0 / N′ j /N f Q Since M is injective, we can enlarge this to a commutative diagram 0 / N′ j /N p′ g f x QO g p′ i M This means that by the injectivity of M we have g : N → M such that if = gj. Then f = id Q f = p′ if = (p′ g)j. Hence, Q is injective. Exercises 4.3. 1. Let R = Z6 . Define Z6 × Z2 → Z2 by [r]6 [x]2 := [rx]2 and Z6 × Z3 → Z3 by [r]6 [x]3 := [rx]3 . Prove that Z2 and Z3 are Z6 -modules and Z6 ∼ = Z2 ⊕ Z3 as Z6 -modules. Z2 and Z3 are not free Z6 -modules. Z6 is a free Z6 -module. Since Z2 and Z3 are direct summands of Z6 -module, they are projective. 119 4.4. Modules over a PID 2. Show that if e is an idempotent (e2 = e) in a ring R, the eR is a projective right module and Re is an projective left module. 3. Show Mthat (a) Pα is projective if and only if every Pα is projective. α M (b) Qα is injective if and only if every Qα is injective. α 4. Prove that (a) A direct sum of abelian groups is divisible if and only if each summand is divisible. (b) A homomorphic image of a divisible module is divisible. 5. Let R be an integral domain that is not a field. If M is an R-module such that M is both injective and projective, prove that M = {0}. 6. Prove (i) ⇒ (ii) in Theorem 4.3.11 by dualizing the proof of Theorem 4.3.3. 7. Consider the polynomial ring Z[x] as a Z-module. (a) Is Z[x] free? (b) Is Z[x] projective? (c) Is Z[x] injective? (d) Is Z[x] divisible? Project 22 (Injective hull). It is possible to prove a sharper result than Theorem 4.3.10, namely that there is a minimal injective R-module H containing M in the sense that any injective map of M into an injective R-module Q factor through H. More precisely, show that if M ⊆ Q for an injective R-module Q then there is an injection i : H → Q that restricts to the identity map on M ; using i to identify H as a subset of Q we have M ⊆ H ⊆ Q. This module H is called the injective hull or injective envelope of M . For example, the injective hull of Z is Q, and the injective hull of any field is itself. Furthermore, prove that: (a) The injective hull of an injective module is itself. (b) The injective hull of an integral domain is its field of fractions. 4.4 Modules over a PID Our main goal of this section is to prove the structure theorem for modules over a PID. Theorem 4.4.1. Let R be a PID and suppose that M is a finitely generated R-module. Then there is an integer r ≥ 0 and nonzero elements d1 , . . . , dk ∈ R with d1 |d2 , . . . , dk−1 |dk such that · · ⊕ R} ⊕R/Rd1 ⊕ · · · ⊕ R/Rdk . M∼ =R | ⊕ ·{z r copies Moreover, if N is another finitely generated R-module and · · ⊕ R} ⊕R/Rd¯1 ⊕ · · · ⊕ R/Rd¯k̄ , N∼ =R | ⊕ ·{z r̄ copies where d¯i |d¯i+1 , then M and N are isomorphic as R-modules if and only if r = r̄, k = k̄ and di and d¯i are associates for i = 1, . . . , k. Note that we cannot assert more than that di and d¯i are associates, for if d and d¯ are associates, ¯ then R/Rd ∼ = R/Rd. Since abelian groups are equivalent to Z-modules, this theorem can be stated as “A finitely generated Z-module is a direct sum of cyclic modules”. Actually, the theorem was more precise in that it actually classified all finitely generated Z-modules up to isomorphism. That is, one has 120 4. Modules and Noetherian Rings Theorem 4.4.2. Let M be a finitely generated Z-module. Then there are nonnegative integers r ≥ 0, d1 , . . . , dk > 0 where d1 |d2 , . . . , dk−1 |dk such that M∼ · · ⊕ Z} ⊕Z/d1 Z ⊕ · · · ⊕ Z/dk Z. =Z | ⊕ ·{z r copies Moreover, if N is another finitely generated Z-module and N∼ · · ⊕ Z} ⊕Z/d¯1 Z ⊕ · · · ⊕ Z/d¯k̄ Z, =Z | ⊕ ·{z r̄ copies where d¯i |d¯i+1 , then M and N are isomorphic if and only if r = r̄, k = k̄ and di = d¯i for i = 1, . . . , k. r is called the rank or torsion-free rank of M and d1 , . . . , dk are called the invariant factors of M . Therefore, we have a major theorem on abelian groups: Corollary 4.4.3. A finitely generated abelian group is a direct product of cyclic groups. The strategy of our proof is the following. First we observe that even with no hypothesis on the ring R, the following statements are equivalent: (i) M is a finitely generated R-module and can be generated by s elements. (ii) Let F = Rx1 + · · · + Rxs be a free R-module with s free generators. Then there is an exact φ /M /F / 0 where K = ker φ. /K sequence of R-modules 0 ∼ (iii) M = F/K where F is a free R-module on s free generators and K is an R-submodule of F . Now let us suppose that R is commutative and we have a free R-module F and a submodule K where F = Rx1 + · · · + Rxr + Ry1 + · · · + Ryk and K = d1 Ry1 + · · · + dk Ryk . Then it is easy to see that · · ⊕ R} ⊕R/Rd1 ⊕ · · · ⊕ R/Rdk F/K ∼ =R | ⊕ ·{z r copies since Rxi ∼ = R and Ryi /di Ryi ∼ = R/Rdi . If we have an arbitrary commutative ring R, K may not have an appropriate form. Moreover, no change of basis may be possible which changes K to the appropriate form. However, in case R is a PID, it is always possible to choose a basis for F and a basis for K (which will also be free) so that the above situation exists. In addition, it will be possible to choose d1 , . . . , dk so that d1 |d2 , . . . , dk−1 |dk . This will yield the desired structure theorem for finitely generated modules over R. The proof will consist of two stages. Stage I. We prove an appropriate theorem about m × n matrices over a PID R. Stage II. We show that theorem about m × n matrices proved in Stage I can be translated into a theorem about modules—namely the structure theorem for modules over a PID. We shall now prove a theorem which says that any m × n matrix [A] over a PID R can be transformed to a diagonal matrix by a transformation [A] → [P ][A][Q] where [P ] and [Q] are appropriate invertible matrices over R. Let R be a ring and GLn (R) the group of invertible n × n matrices over R. It is called the general linear group over R. Moreover, if R is commutative, then GLn (R) = {A ∈ Mn (R) : det A is a unit in R}. 121 4.4. Modules over a PID Theorem 4.4.4. Let R be a commutative ring. a b . ∗ ∗ 2. If A = [aij ] is an m × n matrix over R, P ∈ GLm (R) and P A = [bij ], then X X Raij = Rbij . 1. If Ra + Rb = R, then GL2 (R) contains a matrix of the form i,j i,j In other words, the entries of A and of P A generate the same ideal in R. 3. If E is the elementary matrix obtained by interchanging the i-th and j-th rows of the identity matrix Im and A is an m × n matrix, then E ∈ GLm (R) and EA is the matrix obtained by interchanging the i-th and j-th rows of A. 4. If E is the elementary matrix obtained by multiplying the i-th row of the identity matrix Im by a unit c ∈ R and A is an m × n matrix, then E ∈ GLm (R) and EA is the matrix obtained by multiplying the i-th row of A by c. 5. If E is the elementary matrix obtained by adding c times the j-th row of Im to the i-th row of Im and A is an m × n matrix, then E ∈ GLm (R) and EA is the matrix obtained by adding c times the j-th row of A to the i-th row of A. 6. The analogues of (1)–(5) hold for right multiplications and column transformations. Proof. (1) If Ra + Rb = R, let ra + sb = 1. Then 1 0 a b r −b . = 0 1 −s r s a P P (2) The entries bij of P A are R-linear combinations of the aij , so Rbij ⊆ P Raij . But since P P −1 (P A) = A the entries aij of A are R-linear combinations of bij , so Raij ⊆ Rbij . (3), (4), (5) and (6) are clear. Remark. Passing from A to EA as in (3), (4) and (5) of the above theorem are called elementary row transformations of A. Elementary column transformations of A are defined similarly. Recall that if R is a PID and a, b ∈ R, then Ra + Rb = Rc where c = gcd(a, b). Moreover, if we let a = cx and b = cy, then gcd(x, y) = 1, or Rx + Ry = R. More generally, Ra1 + · · · + Ran = Rd where d = gcd(a1 , . . . , an ) and if ai = dbi , then Rb1 + · · · + Rbn = R. In UFD, gcd(a, b) = 1 does not imply Ra + Rb = R. For example, R = F [x, y], where F is a field, and gcd(x, y) = 1. that a b ∈ M2 (R). Then there exist P, Q ∈ GL2 (R) such c d e 0 P AQ = 0 f Theorem 4.4.5. Let R be a PID and A = where e = gcd(a, b, c, d) and e | f . Proof.We first claim p q (∗) if ∈ M2 (R), either p = gcd(p, q, r, s), or there exist P1 , Q1 ∈ GL2 (R) such that Rp1 ⊃ r s Rp and p q p1 q1 P1 . Q = r s 1 r1 s 1 Case I. q, r, s ∈ Rp. Then p = gcd(p, q, r, s) and we are done. Case II. q ∈ / Rp. Then Rp + Rq = Rp1 where p1 = gcd(p, q) and so Rp1 ⊃ Rp. Let p = p1 x 122 4. Modules and Noetherian Rings and q = p1 y. Then Rx + Ry = R, so we can choose u, v with xu + yv = 1. Then pu + qv = (p1 x)u + (p1 y)v = p1 and p q u −y p ∗ = 1 r s v x ∗ ∗ u −y where Q1 = ∈ GL2 (R). v x Case III. r ∈ / Rp . Then we do the analogue of Case II with a transformation on the first column. Case IV. q, r ∈ Rp and s ∈ / Rp. Then we perform a succession of elementary row and column transformations followed by the manoeuvre of Case II: Let q = αp and r = βp: p q p q − αp p 0 ∼ = r s r s − αr r s − αβp p 0 p 0 ∼ = r − βp s − αβp 0 s − αβp p s − αβp p s gcd(p, s) ∗ ∼ ∼ ∼ (by Case II). 0 s − αβp 0 s − αβp ∗ ∗ p q p q Since this succession of operators corresponds to a transformation ∼ P1 Q where r s r s 1 P1 , Q1 ∈ GL2 (R), we are done in Case IV also. This proves (∗). Next we claim e ∗ (∗∗) there exist P̄ , Q̄ ∈ GL2 (R) such that P̄ AQ̄ = where e = gcd(a, b, c, d). ∗ ∗ If a = gcd(a, b, c, d) we are done. If not, use (∗) to choose P1 , Q1 ∈ GL2 (R) such that Ra1 ⊃ Ra and a 1 b1 a b P1 Q = . c 1 d1 c d 1 If a1 = gcd(a1 , b1 , c1 , d1 ), take P̄ = P, Q̄ = Q and end the process. If not, use (∗) again to choose P2 , Q2 ∈ GL2 (R) such that Ra2 ⊃ Ra1 and a 2 b2 a1 b1 . Q2 = P2 c 2 d2 c 1 d1 Continue inductively to get Pm+1 am+1 bm+1 a m bm Qm+1 = cm+1 dm+1 c m dm where Ram+1 ⊃ Ram as long as am 6= gcd(am , bm , cm , dm ). Since Ra ⊂ Ra1 ⊂ Ra2 ⊂ . . . is a strictly increasing chain of ideals in a PID R, the process must terminate. That is, am = gcd(am , bm , cm , dm) for some let P̄ = Pm Pm−1 . . . P1 and Q̄ = Q1 . . . Qm−1 Qm . Then m. Now a b a m bm P̄ AQ̄ = P̄ where am = gcd(am , bm , cm , dm ) = gcd(a, b, c, d) because Q̄ = c d c m dm P̄ , Q̄ ∈ GL2 (R), as required. This proves (∗∗). Finally, to prove the theorem we follow the transformation A ∼ P̄ AQ̄ by two elementary transformations: a b e αe e 0 e 0 e 0 A= ∼ P̄ AQ̄ = ∼ ∼ = . c d βe γe βe (γ − αβ)e 0 (γ − αβ)e 0 f Hence, we have the desired theorem. 123 4.4. Modules over a PID Theorem 4.4.6. Let R be a PID and A an m × n matrix over R. Then there exist P ∈ GLm (R) and Q ∈ GLn (R) such that d1 d2 0 . . . dr P AQ = 0 .. . 0 0 where the di 6= 0 and d1 | d2 , . . . , dr−1 | dr . Proof. We shall prove this theorem by induction on (m, n). Case I. A = a1 . . . an is a 1 × n matrix. The proof of Theorem 4.4.5 shows that for any a, b ∈ R there is a 2 × 2 matrix Q ∈ GL2 (R) such that a b Q = gcd(a, b) 0 . Hence, for an appropriate Q1 ∈ GL2 (R) Q1 1 1 a1 . . . an = gcd(a1 , a2 ) 0 a3 . . . an . .. . 1 n×n Then a succession of such right multiplications together with elementary transformations (or a suitable induction) show that we can obtain AQ = gcd(a1 , . . . , an ) 0 . . . 0 for some Q ∈ GLn (R) as follows: a1 a2 . . . an ∼ gcd(a1 , a2 ) 0 a3 . . . an ∼ gcd(a1 , a2 ) a3 0 a4 . . . an ∼ gcd(a1 , a2 , a3 ) 0 0 a4 . . . an ∼ · · · ∼ gcd(a1 , . . . , an ) 0 0 . . . 0 . Case II. A is an m × 1 matrix. This is similar to Case I. Case III. The general case. We already know the result for m × 1 and 1 × n matrices and to proceed the induction, we shall assume that m, n ≥ 2 and that we know the result for an (m − 1) × (n − a11 . . . a1n Q1 = 1) matrix over R. Let A = [a ] and let Q ∈ GL (R) be such that ij m×n 1 n a 0 . . . 0 where a = gcd(a11 , . . . , a1n ) by Case I. Then a 0 ... 0 AQ1 = X Y m×n where X is an (m − 1) × 1 columns matrix and Y is some (m − 1) × (n − 1) matrix. By the inductive hypothesis there are P2 ∈ GLm−1 (R) and Q2 ∈ GLn−1 (R) such that e1 .. . P2 Y Q2 = es 124 4. Modules and Noetherian Rings where e1 | e2 , . . . , es−1 | es . Then 1 1 1 a 0 1 a 0 AQ1 = = P2 Q2 P2 X Y Q2 P2 X P2 Y Q2 a 0 ··· 0 a 2 e1 .. . . = a3 .. . es am b a a 3 0 We next use Case II to find a P2′ ∈ GLm−1 (R) such that P2′ . = . where b = gcd(a, a3 , . . . am ). .. .. am 0 We now perform a pair of transformations a 0 0 ··· ··· 0 a 2 e1 0 · · · · · · 0 a 2 e1 0 · · · a 2 e1 a 0 0 · · · · · · 0 b 0 ∗ · · · a3 a 3 0 e2 0 0 e2 . ∼ . ∼. .. .. .. .. . .. .. . . . . Z . .. .. .. .. .. . . . es . es . am am 0 0 0 e2 .. . where Z = is a matrix all of whose entries are divisible by e1 . es ··· ··· 0 ∗ Now by Theorem 4.4.5, there are P3 , Q3 ∈ GL2 (R) such that a 2 e1 e 0 Q3 = P3 b 0 0 f where e = gcd(a2 , e1 , b) = gcd(a2 , e1 , b, entries of Z) = gcd(entries of A). Then P3 a 2 e1 0 · · · 0 Q3 b 0 ∗ · · · ∗ 1 1 0 . . . . . . . . . Z 1 e 0 ∗ ··· 0 f ∗ · · · = 0 .. . Z 0 0 ∗ e ∗ ··· ∗ 0 = .. . W 0 1 ∗ e 0 ··· 0 0 = .. ′ . W 0 where W ′ is an (m − 1) × (n − 1) matrix over R and e divides every entry of W ′ . Now we use the inductive hypothesis again to choose P4 ∈ GLm−1 (R), Q4 ∈ GLn−1 (R) such that d2 .. . P4 W ′ Q4 = dr 125 4.4. Modules over a PID where d2 | d3 , . . . , dr−1 | dr . We note that since e divides all the entries of W ′ , e | d2 . Hence, setting e = d1 , we have e d1 d2 d2 1 1 e . . . . = = . . Q4 P4 W′ dr dr where d1 | d2 , . . . , dr−1 | dr . The whole series of transformations on A amount to a transformation A ∼ P AQ where P ∈ GLm (R) and Q ∈ GLn (R). Therefore, the theorem is proved. Before we can use our result on matrices to show that a finitely generated module over a PID R is a direct sum of cyclic modules, we need to show that every submodule of Rn = R · · ⊕ R} | ⊕ ·{z is finitely generated. In fact we shall show that every submodule of Rn is free of rank ≤ n. n Theorem 4.4.7. Let R be a ring and let M = Rx1 + · · · + Rxm and N = Ry1 + · · · + Ryn be free j π R-modules of rank m and n, respectively. Suppose 0 → M → P → N → 0 is an exact sequence of R-modules. Then P is a free R-module of rank m + n. Proof. It follows from Theorem 4.3.3. Theorem 4.4.8. If R is a PID and P is a submodule of Rn = R · · ⊕ R}, then P is free of | ⊕ ·{z n rank ≤ n. Proof. We shall use induction on n. For n = 1, we have P is a submodule of R, i.e. P is an ideal of R, so P = Rx for some x ∈ R. If x = 0, P = 0, so P is free of rank 0; if x 6= 0, Rx ∼ = R as a left R-module, so P is free of rank 1. Next suppose n > 1 and the theorem is true for free R-submodules of rank < n. Let Rn = Rx1 ⊕ · · · ⊕ Rxn = (Rx1 ⊕ · · · ⊕ Rxn−1 ) ⊕ Rxn . Then we have an exact sequence i π 0 → Rx1 ⊕ · · · ⊕ Rxn−1 → Rn → Rxn → 0 where i is the inclusion map and π is the projection onto the last factor. Let M = (Rx1 ⊕ · · · ⊕ Rxn−1 ) ∩ P ⊆ Rx1 ⊕ · · · ⊕ Rxn−1 and N = π(P ) ⊆ Rxn . Then i π|P 0→M →P → N →0 is an exact sequence of R-modules. M is a submodule of Rn−1 and N is a submodule of Rxn ∼ = R, so both are free of ranks ≤ n − 1 and 1, respectively. Hence, P is free of rank ≤ n = (n − 1) + 1 by Theorem 4.4.7. Theorem 4.4.9. Let R be a PID and A a finitely generated R-module. Then A is a direct sum of cyclic R-modules. More precisely, · · ⊕ R} ⊕R/Rd1 ⊕ · · · ⊕ R/Rdk A∼ =R | ⊕ ·{z r where r ≥ 0 and d1 , . . . , dk are nonzero elements of R and d1 | d2 , . . . , dk−1 | dk . 126 4. Modules and Noetherian Rings Proof. Since A is finitely generated, there is an exact sequence 0 → N → M → A → 0 where M = Rx1 + . . . Rxn is free of finite rank n and N is a submodule of M . By Theorem 4.4.8, N is finitely generated, say by y1 = a11 x1 + a21 x2 + · · · + an1 xn y2 = a12 x1 + a22 x2 + · · · + an2 xn .. . ym = a1m x1 + a2m x2 + · · · + anm xn . Let M̄ = Rn be the space of n × 1 column vector over R and let N̄ be the R-submodule of M̄ generated by the columns of the n × m matrix a11 a21 [N̄ ] = . .. a12 a22 .. . ... ... an1 an2 . . . a1m a2m .. . . anm There is an obvious R-module isomorphism α : M → M̄ defined by r1 .. α(r1 x1 + · · · + rn xn ) = . rn and it is clear that α(N ) = N̄ . Hence, we have R-module isomorphisms A∼ = M/N ∼ = α(M )/α(N ) = M̄ /N̄ . By Theorem 4.4.6, there are matrices [P ] ∈ GLn (R) and [Q] ∈ GLm (R) such that [P ][N̄ ][Q] = d1 .. . 0 dk 0 0 .. . 0 n×m where the di 6= 0 and d1 | d2 , . . . , dk−1 | dk . [N̄ ][Q] is an n × m matrix whose columns generate N̄ as an R-module. Further, left multiplication by [P ] is an isomorphism of M̄ which carries N̄ to the R-submodule Ū of M̄ generated by the columns of [P ][N̄ ][Q]. Then A∼ = M̄ /N̄ ∼ = [P ]M̄ /[P ]N̄ = M̄ /Ū . However, we can see by inspection that M̄ /Ū ∼ · · ⊕ R} . = R/Rd1 ⊕ R/Rd2 ⊕ · · · ⊕ R/Rdk ⊕ R | ⊕ ·{z n−k Hence, A ∼ · · ⊕ R} ⊕R/Rd1 ⊕ R/Rd2 ⊕ · · · ⊕ R/Rdk where d1 | d2 , . . . , dk−1 | dk . =R | ⊕ ·{z n−k 127 4.4. Modules over a PID Example 4.4.1. Let A be the Z-module generated by x, y and z with the relations x + y = 0 and x − y + 2z = 0. Express A as a direct sum of cyclic modules. Solution. Observe that A = {(x, y, z) ∈ Z3 : x + y = 0 and x − y + 2z = 0}. Consider the exact sequence 0 / N = Z(x + y) + Z(x − y + 2z) / Zx ⊕ Zy ⊕ Zz /A / 0. Then y1 = x + y and y2 = x − y + 2z, so 1 0 1 1 [N̄ ] = 1 −1 ∼ 0 2 = [Ū ]. 0 0 0 2 Thus, U = Zx ⊕ Z(2y). Hence, A∼ = Z2 ⊕ Z = {0} ⊕ Z2 ⊕ Zz ∼ = M/U = (Zx ⊕ Zy ⊕ Zz)/(Zx ⊕ Z(2y)) ∼ as desired. Our next goal is to show that the direct summands which occur in the above decomposition are unique up to isomorphism. This does NOT mean that the actual summands which occur are unique. For example, suppose R = Z and A = Z ⊕ Z ⊕ Z3 ⊕ Z3 ∼ = Zx1 ⊕ Zx2 ⊕ Zy1 ⊕ Zy2 where Zx1 and Zx2 are free summands and 3y1 = 3y2 = 0. Then we can also write A = Z(x1 + 2x2 + y2 ) ⊕ Zx2 ⊕ Z(2y1 + y2 ) ⊕ Z(y1 + y2 ) ∼ = Z ⊕ Z ⊕ Z3 ⊕ Z3 In the first case, the direct summands are Zx1 ∼ = Z3 . = Z3 , Zy2 ∼ = Z, Zy1 ∼ = Z, Zx2 ∼ In the second case, the direct summands are Z(x1 + 2x2 + y2 ) ∼ = Z3 . = Z3 , Z(y1 + y2 ) ∼ = Z, Z(2y1 + y2 ) ∼ = Z, Zx2 ∼ Then the summands which occurs are distinct submodules in the two cases, but the isomorphism classes of summands are the same, namely Z, Z, Z3 and Z3 . As a preparation for proving uniqueness, we shall need the concept of a torsion element. Let R be an integral domain and M an R-module. An m ∈ M is called a torsion element of M if there is a nonzero r ∈ R such that rm = 0. Let τ (M ) denote the set of torsion elements of R, called the torsion submodule of M . If τ (M ) = 0, M is said to be a torsion free R-module. Theorem 4.4.10. Let R be an integral domain and M an R-module. Then 1. τ (M ) is a submodule of M . 2. τ (M/τ (M )) = 0. 128 4. Modules and Noetherian Rings Proof. (1) The only problem in showing that τ (M ) is a submodule of M is in showing that τ (M ) is closed under addition. Suppose x, y ∈ τ (M ). Then there exist nonzero elements r, s ∈ R such that rx = 0 and sy = 0. Since R is an integral domain, rs 6= 0. But rs(x + y) = s(rx) + r(sy) = 0 + 0 = 0. Hence, x + y ∈ τ (M ). (2) Suppose x + τ (M ) ∈ τ (M/τ (M )). Then there is a nonzero r ∈ R such that r(x + τ (M )) = rx + τ (M ) = 0 + τ (M ), i.e., rx ∈ τ (M ). Thus, there is a nonzero s ∈ R such that s(rx) = 0, so sr 6= 0 and (sr)x = 0. Hence, x ∈ τ (M ), so τ (M/τ (M )) = 0. Remarks. 1. If R is not an integral domain, then the torsion elements of an R-module M may not form a submodule, even if R is commutative. For example, let R = F × F where F is a field and let M = R = F × F . Then the torsion elements of M are all elements of the form (a, 0) or (0, b). But if a, b 6= 0, (a, b) = (a, 0) + (0, b) is not a torsion element. 2. If R is not commutative, then the torsion elements of an R-module M may not form a submodule, even if R has no zero divisors. For example, there exists a non-commutative domain R (such as the polynomial rings over the quaternion ring) such that for some nonzero x, y ∈ R, Rx ∩ Ry = 0. In other words, x and y have no common left multiple except 0. For such an R, x and y, let M = R/Rx as a left R-module. Then (a) y +Rx is not a torsion element of M , for 0 = r(y +Rx) = ry +Rx, so ry ∈ Rx∩Ry = 0. Thus, ry = 0, so r = 0. (b) 1 + Rx is a torsion element of M since x(1 + Rx) = x + Rx = 0. Since 1 + Rx generates M = R/Rx as a left R-module, it follows that the torsion elements of M do not form a submodule. Theorem 4.4.11. Let R be a PID and let p be an irreducible element of R. 1. Rp is a maximal ideal of R, i.e., R/Rp is a field. 2. If d ∈ R and p ∤ d, then p(R/Rd) = R/Rd. 3. If d ∈ R and p | d, then p(R/Rd) = Rp/Rd ∼ = R/R(d/p). Proof. (1) Suppose Rp is not a maximal ideal and let Rp ⊂ Rx ⊂ R. Then p = rx where neither r nor x is a unit of R, which contradicts the hypothesis that p is irreducible. (2) Since Rp is a maximal ideal and p ∤ d, Rp + Rd = R. Thus, we can choose r, s ∈ R with rp + sd = 1. Then for any x ∈ R, x + Rd = (rp + sd)x + Rd = prx + Rd = p(rx + Rd). Hence, p(R/Rd) = R/Rd. (3) Since p | d, Rd ⊂ Rp. The multiplication by p defines an onto R-module homomorphism ϕp : R → Rp/Rd where ϕp (x) = xp + Rd. It is easy to verify that ker ϕp = R(d/p). Hence, we have the theorem. Theorem 4.4.12. Let R be a PID and suppose d1 , . . . , dk , e1 , . . . , em are nonzero nonunits of R, where d1 | d2 , . . . , dk−1 | dk , e1 | e2 , . . . , em−1 | em . Suppose A = R/Rd1 ⊕ · · · ⊕ R/Rdk ∼ = R/Re1 ⊕ · · · ⊕ R/Rem = B. Then k = m and R/Rdi ∼ = R/Rei for i = 1, . . . , m. In particular, di and ei are associate for i = 1, . . . , m. Proof. Let p be a prime of R which divides d1 . Then p | di for all i = 1, . . . , k, so (R/Rdi )/p(R/Rdi ) = (R/Rdi )/(Rp/Rdi ) ∼ = R/Rp 129 4.4. Modules over a PID for all i = 1, . . . , k. Thus, A/pA ∼ = R/Rp ⊕ · · · ⊕ R/Rp. In other words, A/pA is a vector space {z } | k over the field R/Rp of dimension k. Note that since p(M/pM ) = pM/pM = 0 for any Rmodule M , M/pM may be considered as an R/Rp module, i.e., as a vector space over R/Rp. Since A ∼ = B, A/pA ∼ = B/pB since any isomorphism φ : A → B carries pA onto pB. But since B is generated as an R/Rp-module by ≤ m elements. Thus, m ≥ dimR/Rp (B/pB) = dimR/Rp (A/pA) = k. By symmetry, k ≥ m. Hence, m = k. We now show that R/Rdi ∼ = R/Rei by induction on the number n of prime divisors of d1 · · · dk . E.g., for d1 . . . dk = pα1 1 . . . pαr r , we have n = α1 + · · · + αr . If n = 1, then k = 1 and A = R/Rd1 ∼ = R/Re1 = B. For inductive step, let p be a prime divisor of d1 and hence of d2 , . . . , dk . Then A/pA ∼ = R/Rp ⊕ · · · ⊕ R/Rp as above. Suppose p ∤ e1 . Then p(R/Re1 ) = R/Re1 , by Theorem {z } | k 4.4.11, so (R/Re1 )/p(R/Re1 ) = 0. Thus, B/pB = (R/Re1 )/p(R/Re1 ) ⊕ · · · ⊕ (R/Rek )/p(R/Rek ) ∼ = (R/Re2 )/p(R/Re2 ) ⊕ · · · ⊕ (R/Rek )/p(R/Rek ) is generated by ≤ k − 1 elements. Hence, dimR/Rp (B/pB) ≤ k − 1 and dimR/Rp (A/pA) = k, a contradiction, since A/pA ∼ = B/pB as above. Then p | e1 , so p | ei for all i = 1, . . . , k. By Theorem 4.4.11, we have isomorphisms R/R(d1 /p) ⊕ · · · ⊕ R/R(dk /p) ∼ = p(R/Rd1 ) ⊕ · · · ⊕ p(R/Rdk ) ∼ pB ∼ = pA = = p(R/Re1 ) ⊕ · · · ⊕ p(R/Rek ) ∼ = R/R(e1 /p) ⊕ · · · ⊕ R/R(ek /p). Now the number of prime factors of (d1 /p) · · · (dk /p) is strictly less than the number of prime factors of d1 · · · dk . Hence, the inductive hypothesis applies to the isomorphism R/R(d1 /p) ⊕ · · · ⊕ R/R(dk /p) ∼ = R/R(e1 /p) ⊕ · · · ⊕ R/R(ek /p). Thus, we may conclude that R/R(di /p) ∼ = R/R(ei /p) for i = 1, . . . , k. Note that for any ideal I of R, I = ann(R/I) = {r ∈ R : r(R/I) = 0}. Hence, R/I ∼ = R/J if and only if I = ann(R/I) ∼ = ann(R/J) = J (as submodules of R), and so R/R(di /p) ∼ = R/Rei . = Rei ⇔ R/Rdi ∼ = R(ei /p) ⇔ Rdi ∼ = R/R(ei /p) ⇔ R(di /p) ∼ Therefore, R/Rdi ∼ = R/Rei for i = 1, . . . , k and the theorem is proved. Theorem 4.4.13. Let R be a PID. Suppose that · · ⊕ R} ⊕R/Rd1 ⊕ · · · ⊕ R/Rdk A∼ =R | ⊕ ·{z r and · · ⊕ R} ⊕R/Re1 ⊕ · · · ⊕ R/Rem B∼ =R | ⊕ ·{z s are isomorphic R-modules where the di and ei are nonzero nonunits, d1 | d2 , . . . , dk−1 | dk and e1 | e2 , . . . , em−1 | em . Then r = s, k = m and R/Rdi ∼ = R/Rei for all i = 1, . . . , k. 130 4. Modules and Noetherian Rings Proof. We first observe that the torsion submodules of A and B are τ (A) = R/Rd1 ⊕ · · · ⊕ R/Rdk and τ (B) = R/Re1 ⊕ · · · ⊕ R/Rem . Also, · · ⊕ R} . A/τ (A) = R · · ⊕ R} and B/τ (B) = R | ⊕ ·{z | ⊕ ·{z r s Now if φ : A → B is an isomorphism, it is easy to see that φ(τ (A)) = τ (B), so φ induces isomorphisms φ|τ (A) : τ (A) → τ (B) and φ̂ : A/τ (A) → B/τ (B). In particular, φ̂ is an isomorphism between a free R-module of rank r and one of rank s. Hence, r = s by Theorem 4.2.6. Finally, Theorem 4.4.12 applies to the isomorphism between τ (A) and τ (B) and shows that k = m and R/Rdi ∼ = R/Rei for all i = 1, . . . , k. Exercises 4.4. 1. Let R be a commutative ring such that every submodule of a free R-module is free. Prove that R is a PID. 2. Prove that every finitely generated subgroup of the additive group (Q, +) is cyclic. 3. Let R = Z[x] and let M = (2, x) be the ideal generated by 2 and x, considered as a submodule of R. Show that {2, x} is not a basis of M . Show that the rank of M is 1 but that M is not free of rank 1. 4. Let R be a PID. Prove that (a) For any a, b ∈ R, if gcd(a, b) = 1, then R/Rab ∼ = R/Ra ⊕ R/Rb. (b) If d = pn1 1 . . . pnk k where p1 , . . . , pk are distinct primes and n1 , . . . , nk > 0, then n R/Rd ∼ = R/Rpn1 1 ⊕ · · · ⊕ R/Rpk k . 5. Let M be the Z-module generated by a, b and c with the relations 4a + 3b + 3c = 0 and 2a − b + 3c = 0. Express M as a direct sum of cyclic modules. What are the orders of these modules? 6. Let D be the ring of Gaussian integers Z[i] and M = D3 the free D-module of rank 3. Take K to be the submodule generated by (1, 2, 1), (0, 0, 5) and (1, −i, 6). Prove that M/K is finite and determine its order. 7. Let D be the ring of Gaussian integers Z[i]. Determine the structure of D3 /K where K is generated by f1 = (1, 3, 6), f2 = (2 + 3i, −3i, 12 − 18i) and f3 = (2 − 3i, 6 + 9i, −18i). Show that M = D3 /K is finite (of order 352512). (The order of the ring Z[i]/(a + bi) is a2 + b2 .) 8. Let D = Q[x] be the polynomial ring in one variable over the field Q of rational numbers. Let K be the submodule of D3 generated by (2x − 1, x, x2 + 3) and (x, x, x2 ). Find polynomials g1 , . . . , gr such that D3 /K ∼ = D/(g1 ) ⊕ · · · ⊕ D/(gr ). 4.5 Noetherian Rings In the proof of Theorem 2.4.11 (every PID is a UFD), the fact that “every ideal of R is principal” is used to argue that there is no infinite strictly increasing chain of ideals in R. A ring with this property is called a Noetherian ring, in honor of Emmy Noether, who inaugurated the use of chain condition in algebra. Noetherian rings are of the utmost importance in algebraic geometry and algebraic number theory. One reason for this is that for any field F , F [x1 , . . . , xn ], n ≥ 2, is Noetherian domain but not a PID. We shall study Noetherian rings in this section. A partially ordered set Σ has the ascending chain condition (a.c.c.) if every chain s1 ≤ s2 ≤ . . . 131 4.5. Noetherian Rings eventually breaks off, that is, sk = sk+1 = . . . for some k. This is a finiteness condition in logic that allows arguments by induction, even when the partially ordered set Σ is infinite. It is easy to see that a partially ordered set Σ has the a.c.c. if and only if every nonempty subset S ⊂ Σ has a maximal element: If ∅ = 6 S ⊂ Σ does not have a maximal element, then choose s1 ∈ S, and for each sk , an element sk+1 with sk < sk+1 , thus contradicting the a.c.c.. Theorem 4.5.1. Let R be a ring. The following three conditions are equivalent. (i) The set Σ of left ideals of R has the a.c.c.; in other words, every increasing chain of left ideals I1 ⊂ I2 ⊂ . . . eventually stops, that is Ik = Ik+1 = . . . for some k. (ii) Every nonempty set S of left ideals has a maximal element. (iii) Every left ideal I ⊂ R is finitely generated. If one of these conditions hold, then R is Noetherian (named after E. Noether). Proof. Here (i) ⇔ (ii) is the purely logical statement about partially ordered sets already discussed, whereas (i) or (ii) ⇔ (iii) is directly concerned with rings and ideals. (i) ⇒ (iii). Pick f1 ∈ I, then if possible f2 ∈ I r(f1 ), and so on. At each step, if I 6= (f1 , . . . , fk ), pick fk+1 ∈ I r (f1 , . . . , fk ). Then by the a.c.c. (i), the chain of ideals (f1 ) ⊂ (f1 , f2 ) ⊂ · · · ⊂ (f1 , . . . , fk ) ⊂ · · · must break off at some stage, and this can only happen if (f1 , . . . , fk ) = I for some k. This proof involves an implicit appeal to the axiom of choice. It is perhaps cleaner to do (i) ⇒ (ii) purely in set theory, then argue as follows. (ii) ⇒ (iii). Let I be a left ideal of R and consider the set S of finitely generated left ideals contained in I. Then {0} ∈ S , so that S has a maximal element J by (ii). But then J = I, since any element f ∈ I r J would give rise to a strictly bigger finitely generated left ideal J ⊂ (J, f ) ⊆ I. S (iii) ⇒ (i). Let I1 ⊂ I2 ⊂ . . . be an increasing chain of left ideals. Then J = k Ik is again an ideal. If J is finitely generated then J = (f1 , . . . , fn ) and each fi ∈ Iki , so that setting k = max ki gives J = Ik and the chain stops. Remarks. 1. Every PID is Noetherian. Hence, we may consider a Noetherian ring as a generalization of a PID. 2. Most rings of interest are Noetherian this is a very convenient condition to work with. At first sight, more concrete conditions (such as R finitely generated over k or over Z) might seem more attractive, but as a rule, the Noetherian condition is both more general and more practical to work with. 3. The descending chain condition (d.c.c.) on a partially ordered set is defined in a similar way. A ring whose ideals satisfy the d.c.c. is called an Artinian ring. This is also a very important notion, but is more special: the d.c.c. for rings turns out to be very much stronger than the a.c.c. (and implies it). We shall discuss this kind of rings in the next section. Example 4.5.1. Z is Noetherian but not Artinian since Z ⊃ pZ ⊃ p2 Z ⊃ . . . , p prime, is a decreasing chain which does not stop. Examples 4.5.2. Here are three examples of non-Noetherian rings. Let k be a field. 1. The polynomial ring k[x1 , . . . , xn , . . . ] in an infinite number of indeterminates is obviously non-Noetherian. 2. Consider the ring A1 of polynomials in x, y of the form f (x, y) = a + xg(x, y) with a a constant and g ∈ k[x, y]; that is, f involves no pure power y j of y with j > 0. In other words, n o X A1 = f (x, y) = aij xi y j : i, j ≥ 0 and i > 0 if j 6= 0 = k[x, xy, xy 2 , . . . , xy n , . . . ] ⊂ k[x, y]. 132 4. Modules and Noetherian Rings It is clear that (x, xy, xy 2 , . . . ) is a maximal ideal of A1 , and is not finitely generated. (It looks as if it should be generated by x, but, of course, y, y 2 , . . . are not elements of the ring A1 .) Thus, A1 is not Noetherian. 3. A rather similar example is the ring A2 of polynomials in x, y, y −1 of the form g(x, y) + xh(x, y, y −1 ); that is, n o X A2 = f (x, y) = aij xi y j : i ≥ 0, and j ≥ 0 if i = 0 = k[x, y, x/y, x/y 2 , . . . , x/y n , . . . ]. In this ring x = (x/y) · y, and x/y = (x/y 2 ) · y, etc., so that the element x does not have a factorization into irreducibles and (x) ⊂ (x/y) ⊂ (x/y 2 ) ⊂ · · · is an infinite ascending chain. Let R be a ring. An R-module M is Noetherian if the submodules of M have the a.c.c., that is, any increasing chain M1 ⊂ M2 ⊂ . . . ⊂ Mk ⊂ . . . of submodules eventually stops. Just as before, it is equivalent to say that any nonempty set of submodules of M has a maximal element, or that every submodule of M is finitely generated. /L α /M β /N Theorem 4.5.2. Let 0 M is Noetherian if and only if L and N are. / 0 be an exact sequence of R-modules. Then Proof. Obviously, the condition is necessary. Suppose M1 ⊂ M2 ⊂ . . . is an increasing chain of submodules of M ; then identifying α(L) with L and taking intersection gives a chain L ∩ M1 ⊂ L ∩ M2 ⊂ . . . of submodules of L and applying β gives a chain β(M1 ) ⊂ β(M2 ) ⊂ . . . of submodules of N . Each of these two chains eventually stops, by the assumption on L and N , so that we need to prove the following statement: Lemma 4.5.3. For submodules, M1 ⊂ M2 ⊂ M , if L ∩ M1 = L ∩ M2 and β(M1 ) = β(M2 ), then M1 = M2 . Proof. Indeed, if m ∈ M2 , then β(m) ∈ β(M2 ) = β(M1 ), so that there is an n ∈ M1 such that β(m) = β(n). Then β(m − n) = 0, so that m − n ∈ M2 ∩ ker β = M1 ∩ ker β. Hence, m ∈ M1 . We record consequences of this theorem in: Corollary 4.5.4. Let M be an R-modules and N an R-submodule of M . Then M is Noetherian if and only if N and M/N are. Corollary 4.5.5. 1. If Mi are Noetherian modules, i = 1, . . . , r, then r M i=1 Mi is Noetherian. 2. If R is a Noetherian ring, then an R-module M is Noetherian if and only if it is finitely generated over R. 3. If R is a Noetherian ring and M is a finitely generated R-module, then any submodule N ⊂ M is again finitely generated. 4. If R is a Noetherian ring and ϕ : R → B is a ring homomorphism such that B is a finitely generated R-module, then B is a Noetherian ring. In particular, a homomorphic image of a Noetherian ring is a Noetherian ring. 133 4.5. Noetherian Rings Proof. (1) A direct sum M1 ⊕ M2 is a particular case of an exact sequence, so that the previous proves (1) when r = 2. The case r > 2 follows by an easy induction. (2) If M is finitely generated then there is a surjective homomorphism Rr → M → 0 for some r, so that M is a quotient M ∼ = Rr /N for some submodule N ⊂ Rr ; now Rr is a Noetherian module by (1), so M Noetherian follows by the implication ⇒ of the above theorem. Conversely, M Noetherian obviously implies M is finitely generated. (3) This just uses the previous implication: M finitely generated and R Noetherian implies that M is Noetherian, so that N is Noetherian, which implies that N is a finitely generated R-module. (4) B is Noetherian as an R-module; but left ideals of B are submodules of B as an R-submodule, so that B is a Noetherian ring. The following result provides many examples of Noetherian rings, and is the main motivation behind the use of the a.c.c. in commutative algebra. Note that in Hilbert’s day, a “basis” of a module meant simply a family of generators. Theorem 4.5.6. [Hilbert Basis Theorem] If R is a commutative Noetherian ring, then so is the polynomial ring R[x]. Proof. We shall prove that any ideal I ⊂ R[x] is finitely generated. For this, define auxiliary sets Jn ⊂ R by Jn = {a ∈ R : there exists f ∈ I such that f (x) = axn + bn−1 xn−1 + · · · + b1 x + b0 }. In other words, Jn is the set of leading coefficients of elements of I of degree n ≥ 0. Then it is easy to check that Jn is an ideal (using the fact that I is an ideal), and that Jn ⊂ Jn+1 (because for f ∈ I, also xf ∈ I), and therefore J0 ⊆ J1 ⊆ J2 ⊆ . . . is an increasing chain of ideals. Using the assumption that R is Noetherian, we deduce that Jn = Jn+1 = . . . for some n. For each m ≤ n, the ideal Jm ⊂ R is finitely generated, say Jm = (am,1 , . . . , am,rm ); and by definition of Jm , for each am,j with 1 ≤ j ≤ rm there is a polynomial fm,j ∈ I of degree m having the leading coefficient am,j . This allows us to write down a finite set S = {fm,j }0≤m≤n,1≤j≤rm of elements of I. We now claim that S generates I. Indeed, for any polynomial f (x) ∈ I, if f (x) has degree m P then its leading coefficient bi an,i P a is in Jm , hence if m ≥ n, then a ∈ Jm = Jn , so that a = withP bi ∈ R and f (x) − bi xm−n fn,i (x) has degree < m; similarly, if m ≤ n, then a ∈ J , so that m P a= bi am,i with bi ∈ R and f (x) − bi fm,i (x) has degree < m. By induction on m, it follows that f can be written as a linear combination of the finitely many elements in S. This proves that any ideal of R[x] is finitely generated. Corollary 4.5.7. If R is a commutative Noetherian ring and ϕ : R → B is a ring homomorphism such that B is a commutative finitely generated extension ring of ϕ(R), then B is Noetherian. Proof. The assumption is that B is a quotient of a polynomial ring, B ∼ = R[x1 , . . . , xn ]/I for some ideal I. Now by Hilbert Basis Theorem and an obvious induction, R Noetherian implies that so is R[x1 , . . . , xn ], and by Corollary 4.5.5, (4), R[x1 , . . . , xn ] is Noetherian implies that so is R[x1 , . . . , xn ]/I. 134 4. Modules and Noetherian Rings Exercises 4.5. 1. Let M be a finitely generated R-module where R is Noetherian. Suppose I is an ideal of R such that for each element a ∈ I, there exists a nonzero element m ∈ M such that am = 0M . Show that Ix = {0M } for some nonzero element x ∈ M . 2. Let M be a Noetherian R-module. Prove that I = {r ∈ R : rm = 0M for all m ∈ M } is an ideal of R and R/I is Noetherian. 3. Let M be a Noetherian R-module and ϕ : M → M be a surjective module homomorphism. Prove that ϕ is an isomorphism. [Hint. consider the chain of submodules ker ϕ ⊂ ker ϕ2 ⊂ · · · .] 4.6 Artinian Rings In this section, we study deeper commutative ring theory. Our main goal is to show that any finite commutative ring is a direct product of a finite number of local rings. However, we present results on more a general ring, called an “Artinian ring”. The Jacobson radical of a ring R is the intersection of all maximal ideals of R and is denoted by Jac R. Note that if R is a local ring with unique maximal ideal M , then Jac R = M . Let R be a ring. An element a ∈ R is nilpotent if an = 0 for some n ∈ N. The set of all nilpotent elements in a commutative ring R is an ideal, called the nilradical of R. It is also clear that every prime ideal in a commutative ring contains the nilradical. Theorem 4.6.1. Let J be the Jacobson radical of a commutative ring R. 1. If I is a proper ideal of R, then so is the ideal generated by I and J. 2. The Jacobson radical contains the nilradical of R. 3. For x ∈ R, x ∈ J if and only if 1 − rx is a unit for all r ∈ R. In partucular, if R is a local ring with unique maximal ideal M , then 1 − m is a unit in R for all m ∈ M . 4. [Nakayama’s lemma] If M is any finitely generated R-module and JM = M , then M = {0}. 5. If M is finitely generated and M = N + IM for some ideal I ⊆ J and submodule N of M , then M = N . 6. Let I be an ideal in the Jacobson radical of R, and suppose that M is finitely generated. If m1 , . . . , mn have images in M/IM that generate it as an R-module, then m1 , . . . , mn also generate M as an R-module. Proof. (1) If I is a proper ideal of R, then I is contained in some maximal ideal M of R. Since J ⊆ M, I ∪ J ⊆ M. (2) Let a ∈ R be nilpotent. Then an = 0 for some n ∈ N. Since maximal ideals are prime and an ∈ M , so a ∈ M . (3) Suppose 1 − rx is not a unit for some r ∈ R and let M be a maximal ideal containing 1 − rx. Since 1 ∈ / M , rx ∈ / M , so x ∈ / M . But J ⊆ M , it follows that x ∈ / J. Conversely, assume that x∈ / J. Then there is a maximal ideal M such that x ∈ / M . Thus, R = (x, M ), so 1 = rx + m for some r ∈ R and m ∈ M . Hence, 1 − rx = m ∈ M which implies that 1 − rx is not a unit in R. (4) Assume that M 6= {0} and let n be the smallest positive integer such that M is generated by n elements, say m1 , . . . , mn . Since M = JM , we have mn = r1 m1 + · · · + rn mn for some r1 , . . . , rn ∈ J. Thus, (1 − rn )mn = r1 m1 + · · · + rn−1 mn−1 . By (3), 1 − rn is a unit, so mn lies in the module generated by m1 , . . . , mn−1 which contradicts the minimality of n. Hence, M = {0}. (5) Apply (4) to M/NP . (6) Apply (5) to N = i Rmi . 135 4.6. Artinian Rings Remark. In the special case of a finitely generated module M over a local ring R with unique maximal ideal J, the quotient M/JM is a vector space over the field R/J. Statement (6) implies that a basis of M/JM lifts to a minimal set of generators of M . Conversely, every minimal set of generators of M is obtained in this way, and any two such sets of generators are related by an invertible matrix with entries in the ring. A ring whose ideals satisfy the descending chain condition (d.c.c.), i.e., whenever I1 ⊇ I2 ⊇ . . . is a decreasing chain of ideals of R, then there is a positive integer m such that Ik = Im for all k ≥ m, is called an Artinian ring (named after E. Artin). Clearly, every finite ring is Artinian. Also, it is immediate that every quotient ring of an Artinian ring is Artinian. Similar to Theorem 4.5.1, we have the following theorem. Theorem 4.6.2. R is an Artinian ring if and only if every nonempty set S of ideals has a minimal element. An R-module M is said to be Artinian if it satisfies d.c.c. on submodules. Similar to Theorem 4.5.2, we have: /L /M /N / 0 be an exact sequence of R-modules. Then Theorem 4.6.3. Let 0 M is an Artinian R-module if and only if L and N are. Lemma 4.6.4. Let M be a maximal ideal of the commutative ring R and suppose that M m = {0} for some m ∈ N. Then R is Noetherian if and only if R is Artinian. Proof. Observe that each successive quotient M i /M i+1 , i = 0, 1, . . . , m − 1, in the filtration R ⊇ M ⊇ M 2 ⊇ · · · ⊇ M m−1 ⊇ M m = {0} is a module over the field F = R/M . Consider the exact /R / R/M / 0 of R-modules. Assume that R is Noetherian. By /M sequence 0 Theorem 4.5.1, M and R/M is Noetherian. Thus, R/M and M are Artinian by Exercise 4.6 (2). Hence, it follows from Theorem 4.6.3 that R is Artinian. The converse is proved in the same way. Lemma 4.6.5. Let R be a commutative ring and P a prime ideal of R. If I and J are ideals of R such that P ⊇ I ∩ J, then I ⊆ P or J ⊆ P . Proof. Assume that I * P and J * P . Let x ∈ I, ∈ / P and y ∈ J, ∈ / P . Then xy ∈ I ∩ J. Since x and y are not in P and P is a prime ideal, xy ∈ / P which contradicts P ⊇ I ∩ J. Now, we are ready to prove our main results. Theorem 4.6.6. Let R be a commutative Artinian ring. 1. There are only finitely many maximal ideals in R. 2. The quotient R/(Jac R) is a direct product of a finite number of fields. More precisely, if M1 , . . . , Mn are finitely many maximal ideals in R, then R/(Jac R) ∼ = k1 × · · · × kn , where ki is the field R/Mi for all i ∈ {1, . . . , n}. 3. Every prime ideal of R is maximal. The Jacobson radical of R equals the nilradical of R and (Jac R)m = {0} for some m ∈ N. 4. The ring R is isomorphic to the direct product of a finite number of Artinian local rings. 5. Every Artinian ring is Noetherian. 136 4. Modules and Noetherian Rings Proof. (1) Let S be the set of all ideals of R that are the intersection of a finite number of maximal ideals. By Theorem 4.6.2, S has a minimal element, say M1 ∩ · · · ∩ Mn . Then for any maximal ideal M , we have M ∩ M 1 ∩ · · · ∩ Mn = M1 ∩ · · · ∩ Mn , so M ⊇ M1 ∩ · · · ∩ Mn . By Lemma 4.6.5, Mi ⊆ M for some i. Since Mi and M are maximal, Mi = M and hence M1 , . . . , Mn are all maximal ideals of R. (2) Since Mi + Mj = R for all i 6= j and Jac R = M1 ∩ · · · ∩ Mn , the statement follows from the Chinese remainder theorem applied to M1 , . . . , Mn . (3) We first show that J = Jac R is nilpotent. By d.c.c., there is some m ∈ N such that J m = J m+i for all i ∈ N. Assume that J m 6= {0}. Let S be the set of proper ideals I such that IJ m 6= {0}. Then J ∈ S . Let I0 be a minimal element of S . Thus, there is some x ∈ I0 such that xJ m 6= {0}. By minimality of I0 , we have I0 = (x). Since ((x)J)J m = xJ m+1 = xJ m , it follows that (x) = (x)J by minimality of (x). By Nakayama’s lemma, (x) = {0}, a contradiction. Hence, J m = {0}. Since am ∈ J m = {0} for all a ∈ J, every element of J is nilpotent. But J contains the nilradical of R, so these two ideals are equal. Let P be a prime ideal P of R. Then P contains the nilradical of R, so it contains J. Thus, P/J is a prime ideal of R/J. By (2), R/J ∼ = k1 × · · · × kn and thus a prime ideal of R/J consists of the elements that are 0 in one of the components. In particular, such a prime ideal is also a maximal ideal. Hence, P is maximal as desired. (4) Let M1 , M2 , . . . , Mn be all the distinct maximal ideals of R and let J = Jac R and J m = {0} as in (3). Then !m n n \ \ Mim ⊆ Mi ⊆ J m = {0}. i=1 i=1 It follows from the Chinese remainder theorem that R∼ = R/M1m × R/M2m × · · · × R/Mnm , and each R/Mim is an Artinian ring (because R is) with unique maximal ideal Mi /Mim . (5) From (4), it suffices to prove that an Artinian local ring is Noetherian. Assume that R is an Artinian with unique maximal ideal M . Then Jac R = M and M m = {0} for some m ∈ N. Thus, the desired result follows from Lemma 4.6.4. Corollary 4.6.7. Every finite commutative ring is a direct product of a finite number of local rings. Example 4.6.1. Let n > 1. If n = pa11 pa22 . . . par r , then Z/nZ ∼ = Z/pa11 Z × Z/pa22 Z × · · · × Z/par r Z. Each Z/pai i Z is a local ring with unique maximal ideal pi Z/pai i Z for all i ∈ {1, 2, . . . , r}. Exercises 4.6. 1. Prove that an Artinain integral domain is a field. Hence, Z is not Artinian. 2. Suppose R = F is a field. Prove that an R-module M is Artinian if and only if it is Noetherian if and only if M is a finite dimensional vector space over F . 3. Let F be a field and let f (x) be a polynomial in F [x] of degree at least one. Decompose the quotient ring F [x]/(f (x)) as a direct product of a finite number of local rings. 4. Let R and S be commutative rings. Prove that (R × S)× = R× × S × . 4.7. Symplectic Geometry 4.7 137 Symplectic Geometry In this section, we see some applications of module theory, especially a free R-module over commutative rings (§4.2), to obtain a structure theorem for finite dimensional symplectic spaces over a local ring. In addition, we study the symplectic graphs over a commutative ring. This is the work of the author published in Discrete Mathematics [32] and European Journal of Combinatorics [33, 34]. However, we present a new combinatorial approach for obtaining the main theorem (Theorem 4.7.5). In addition, we suggest a nice parallel project on studying the orthogonal graphs over a finite commutative ring. 4.7.1 Symplectic Spaces A bilinear form and a symplectic form for an R-module are defined analogously as for a vector space. Let R be a commutative ring and V a left R-module. A bilinear form on V is a map β : V × V → R which is R-linear in both variables. That is, β(a~x + b~y , ~z) = aβ(~x, ~z) + bβ(~y , ~z) and β(~x, a~y + b~z) = aβ(~x, ~y ) + bβ(~x, ~z) for all ~x, ~y , ~z ∈ V and a, b ∈ R. A bilinear form β is called symplectic if ∀~x ∈ V, β(~x, ~x) = 0. Remark. If a bilinear form β is symplectic then 0 = β(~x + ~y , ~x + ~y ) = β(~x, ~x) + β(~y , ~x) + β(~x, ~y ) + β(~y , ~y ) = β(~y , ~x) + β(~x, ~y ) for all ~x, ~y ∈ V . That is, any symplectic bilinear form is also skew-symmetric. For an R-submodule W of V , we write W ⊥ (read “W perp”) for the submodule {~x ∈ V : ∀w ~∈ W, β(~x, w) ~ = 0}, called the orthogonal complement of W . Example 4.7.1. Consider V = Z2 as a vector space over Z2 . The bilinear form β(x, y) = xy is skew-symmetric but not symplectic because β(1, 1) = 1 6= 0. A bilinear form β : V × V → R is called regular or non-degenerate or non-singular if 1. ∀f ∈ HomR (V, R), ∃~x0 , ~y0 ∈ V, f (~x) = β(~x0 , ~x) and f (~x) = β(~x, ~y0 ). 2. If ~y ∈ V and ∀~x ∈ V, β(~x, ~y ) = 0, then ~y = ~0. Similarly, if ~y ∈ V and ∀~x ∈ V, β(~y , ~x) = 0, then ~y = ~0. Example 4.7.2. Let p be a prime number and let R be the ring of integers modulo pn , Zpn , or the field of pn elements, Fpn , where n ∈ N. For ν ≥ 1, let V denote the set of 2ν-tuples (a1 , . . . , a2ν ) of elements in R. Define β : V × V → R by the product β (a1 , . . . , a2ν ), (b1 , . . . , b2ν ) = (a1 , . . . , a2ν ) K (b1 , . . . , b2ν )t , 0 Iν and Iν is the ν × ν identity matrix, for all vectors (a1 , . . . , a2ν ), where K = −Iν 0 2ν×2ν (b1 , . . . , b2ν ) ∈ V . Then β is a non-degenerate symplectic bilinear form. Let R be a commutative ring and V a free R-module of rank n where n ≥ 2. Let β be a nondegenerate symplectic bilinear form. We call the pair (V, β) a symplectic space. An R-module automorphism σ on V is an isometry on V if β(σ(~x), σ(~y )) = β (~x, ~y ) for all ~x, ~y ∈ V . The group of isometries on V is called the symplectic group of (V, β) over R and denoted by SpR (V ). Let R be a commutative ring and (V, β) a symplectic space, where V is a free R-module of rank n ≥ 2. A vector ~x in V is said to be unimodular if there is an f in HomR (V, R) with f (~x) = 1; equivalently, if ~x = α1~b1 + · · · + αn~bn , where {~b1 , . . . , ~bn } is a basis for V , then the ideal (α1 , . . . , αn ) = R. If ~x is unimodular, then the line Rx is a free R-direct summand of rank one. 138 4. Modules and Noetherian Rings A hyperbolic pair {~x, ~y } is a pair of unimodular vectors in V with the property that β (~x, ~y ) = 1. The module H = R~x ⊕ R~y is called a hyperbolic plane. Note that when R is a field, unimodular vectors coincide with nonzero vectors. When R is a local ring, we have a criterion to determine whether a vector in V is unimodular as follows. Theorem 4.7.1. Let R be a local ring and (V, β) a symplectic space, where V is a free R-module of rank n ≥ 2 with basis {~e1 , . . . , ~en }. A vector ~x = a1~e1 + · · · + an~en in V is unimodular if and only if ai is a unit of R for some i ∈ {1, . . . , n}. Proof. If some ai is a unit in R, then (a1 , . . . , an ) = R, so ~x is unimodular. Conversely, assume that ~x is unimodular. Then there exists an f ∈ V ∗ such that 1 = f (~x) = a1 f (~e1 ) + · · · + an f (~en ). Suppose that ai is not a unit in R for all i. Since R is a local ring, ai ∈ M for all i, and thus a1 f (~e1 ) + · · · + an f (~en ) ∈ M . By Theorem 4.6.1 (3), 0 = 1 − (a1 f (~e1 ) + · · · + an f (~en )) is a unit in R, which is a contradiction. Therefore, ai is a unit of R for some i ∈ {1, . . . , n}. In addition, if R is a local ring, we show that the rank of symplectic space (V, β) must be even. Let {~x, ~y } be a hyperbolic pair of unimodular vectors in V and H = R~x ⊕ R~y the corresponding hyperbolic plane. Then for ~z ∈ V , it is easy to see that the vector w ~ = ~z − β(~z, ~y )~x + β(~z, ~x)~y is in H ⊥ , and so ~z can be decomposed as the sum ~z = w ~ + (β(~z, ~y )~x − β(~z, ~x)~y ) of vectors in H ⊥ and H, respectively. Since β is non-degenerate, H ∩ H ⊥ = {~0}. Thus, V = H ⊕ H ⊥. Notation. If W1 and W2 are R-submodule of an R-module V and β(w ~ 1, w ~ 2 ) = 0 for all w ~ 1 ∈ W1 and w ~ 2 ∈ W2 , we write W1 ⊥W2 . Moreover, any unimodular vector ~u may be complemented to a hyperbolic pair as follows. First note that there is an f ∈ HomR (V, R) such that f (~u) = 1. Since β is non-degenerate, there is a ~v ∈ V with 1 = f (~u) = β(~u, ~v ). Then {~u, ~v } is a hyperbolic pair. Combining this with the previous observation, we have the first part of the following results. Theorem 4.7.2. Let R be a local ring. Let (V, β) be a symplectic space over R of rank ≥ 2. Then V splits as an orthogonal direct sum V = H⊥H ⊥ for some hyperbolic plane H. Moreover, H ⊥ is a free R-module. Therefore, V is an orthogonal direct sum V = H1 ⊥H2 ⊥ . . . ⊥Hm of hyperbolic planes H1 , H2 , . . . , Hm . In particular, the rank of V is even. Note that H ⊥ is a direct summand of the free module V . By Corollary 4.3.4, it is finitely generated and projective. Then this theorem follows directly from the next lemma. Lemma 4.7.3. A finitely generated projective module V over a local ring R is free. Proof. Let M be the unique maximal ideal in R. Choose ~v1 , . . . , ~vt ∈ V so that the cosets {~v1 + M V, . . . , ~vt +M V } is a basis for the vector space V /M V over the field R/M . Here, the scalar P action is given by (c + M )(~x + M V ) = c~x + M V . Let ϕ : Rt → V be defined by ϕ(r1 , . . . , rt ) = ti=1 ri~vi . By Remark after Theorem 4.6.1, ϕ is onto. Since V is projective, Rt = K ⊕ L where K = ker ϕ and L ∼ = M . Then K is finitely generated. Since ϕ induces an isomorphism from Rt /M Rt to V /M V , it follows that K/M K ⊕ L/M L ∼ = V /M V . These are finite dimensional vector spaces over the field R/M . Comparing dimensions yields K/M K = 0. Thus, K = {~0} by Nakayama’s lemma. Hence, ϕ is an isomorphism. 139 4.7. Symplectic Geometry 4.7.2 Symplectic Graphs The general symplectic graph associated with nonsingular alternate matrices over a field is studied by Tang and Wan [42] as a new family of strongly regular graphs. Meemark and Prinyasart [32] introduced the symplectic graph GSpR (V ) for a symplectic space V over a commutative ring R. They showed that their symplectic graph is vertex transitive and arc transitive when R = Zpn , p is an odd prime and n ≥ 1. There are many articles influenced by this definition such as [30], [31], [24]. Mostly, the work was on strong regularity, automorphism groups, vertex and arc transitivities, chromatic numbers and subconstituents of symplectic graphs over a finite field, modulo pn , and modulo pq, where p and q are primes and n ≥ 1. Recently, Meemark and Puirod [33] studied those topics over finite local rings and obtained results parallel to [42], [32], [30], [31] and [24]. Following [32], we recall the definition of the symplectic graph. Let R be a commutative ring and (V, β) a symplectic space, where V is a free R-module of rank 2ν, ν ≥ 1. Define the graph GSpR (V ) with vertex set is the set of lines {R~x : ~x is a unimodular vector in V } and with adjacency given by R~x is adjacent to R~y ⇔ β(~x, ~y ) ∈ R× . We call GSpR (V ) , the symplectic graph of (V, β) over R. A strongly regular graph with parameters (v, k, λ, µ) is a k-regular graph on v vertices such that for every pair of adjacent vertices there are λ vertices adjacent to both, and for every pair of non-adjacent vertices there are µ vertices adjacent to both. Let k be a finite field of odd characteristic and let (V ′ , β) be a symplectic space of dimension 2ν where ν ≥ 1. Tang and Wan [42] showed that the symplectic graph GSpk (V ′ ) is a strongly regular graph with parameters |k|2ν − 1 2ν−1 2ν−2 2ν−2 , |k| , |k| (|k| − 1), |k| (|k| − 1) . |k| − 1 Their proof used orthogonal complements and matrix theory over finite fields. Let R be a finite local ring with unique maximal ideal M and residue k = R/M . Let V be a free R-module of rank 2ν, ν ≥ 1, and let V ′ be the 2ν-dimensional vector space over k induced from V via the canonical map π : R → k given by π : r 7→ r+M . Moreover, if (V, β) is a symplectic space, then (V ′ , β ′ ) is a symplectic space, where β ′ is given by β ′ (π(~a), π(~b)) = π(β(~a, ~b)) for all ~a, ~b ∈ V . Here, we write π(~a) = (π(a1 ), π(a2 ), . . . , π(a2ν )) for all ~a = (a1 , a2 , . . . , a2ν ) ∈ V . Note that the relation R~x ∼ R~y ⇔ kπ(~x) = kπ(~y ) (4.7.1) is an equivalence relation on the vertex set of the graph GSpR (V ) . Since R is a local ring, it follows that β(~a, ~b) ∈ R× ⇔ π(β(~a, ~b)) 6= M ⇔ β ′ (π(~a), π(~b)) ∈ k × . This gives (3) of the next theorem. 140 4. Modules and Noetherian Rings Theorem 4.7.4. [Lifting Theorem] Let R be a finite local ring with unique maximal ideal M and 2ν −1 and ~x1 , ~x2 , . . . , ~xκ be unimodular vectors in V such that the residue k = R/M . Let κ = |k||k|−1 vertex set V(GSpk (V ′ ) ) = {kπ(~xi ) : i = 1, 2, . . . , κ}. 1. The set Π = {R(~x1 + M 2ν ), R(~x2 + M 2ν ), . . . , R(~xκ + M 2ν )} is a partition of V(GSpR (V ) ), where R(~xi + M 2ν ) = {R(~xi + m) ~ : m ~ ∈ M 2ν } for all ∈ {1, 2, . . . , κ}. Moreover, for each i ∈ {1, 2, . . . , κ}, any two distinct vertices in R(~xi + M 2ν ) are non-adjacent vertices. For each i, the lifting of the vertices corresponding with elements in kπ(~xi ) to vertices in R(~xi + M 2ν ) is demonstrated below. • kπ(~x1 ) • kπ(~x2 ) .. . • kπ(~xκ ) −→ • −→ • .. . −→ • • ··· R(~x1 + m), ~ m ~ ∈ M 2ν • ··· R(~x2 + m), ~ m ~ ∈ M 2ν .. . • • ··· R(~xκ + m), ~ m ~ ∈ M 2ν • • 2. |R(~xi + M 2ν )| = |M |2ν−1 for all i ∈ {1, . . . , κ}. 3. For unimodular vectors ~a, ~b ∈ V , we have R~a and R~b are adjacent vertices in V(GSpR (V ) ) if and only if kπ(~a) and kπ(~b) are adjacent vertices in V(GSpk (V ′ ) ). 4. For i, j ∈ {1, 2, . . . , κ}, if kπ(~xi ) and kπ(~xj ) are adjacent vertices, then R(~xi + m ~ 1 ) and R(~xj + m ~ 2 ) are adjacent vertices in V(GSpR (V ) ) for all m ~ 1, m ~ 2 ∈ M 2ν . Proof. The first part of (1) follows from the relation (4.7.1) and (4) is an immediate consequence of (3). Note that β(~xi + m ~ 1 , ~xi + m ~ 2 ) = β(~xi , m ~ 1 ) + β(m ~ 2 , ~xi ) + β(m ~ 1, m ~ 2) ∈ M for all i ∈ {1, 2, . . . , κ} and m ~ 1, m ~ 2 ∈ M 2ν . This proves the second part of (1). Next, let m ~ 1, m ~ 2 ∈ M and assume that R(~xi + m ~ 1 ) = R(~xi + m ~ 2 ). Then ~xi + m ~ 1 = λ(~xi + m ~ 2) for some λ ∈ R× . Thus, (1 − λ)~xi = λm ~2−m ~ 1 ∈ M 2ν . Since ~xi is unimodular, 1 − λ ∈ M , so λ = 1 + µ for some µ ∈ M . Hence, ~xi + m ~ 1 = (1 + µ)(~xi + m ~ 2 ). Finally, we show that R(1 + µ)(~x + m) ~ = R(~x + m) ~ for all µ ∈ M , ~x ∈ V unimodular, and m ~ ∈ M 2ν and we therefore have (2). Clearly, R(1 + µ)(~x + m) ~ ⊆ R(~x + m). ~ Since µ ∈ M , 1 + µ ∈ R× . Then r(~x + m) ~ = −1 (r(1 + µ) )(1 + µ)(~x + m) ~ for all r ∈ R which gives another inclusion. The results for a symplectic graph over a finite local ring are presented in the next theorem. Our proof here is an application of the lifting theorem with some combinatorial arguments. This approach is clean and much difference from the one given in [33]. It explains the reason why we do not have strong regularity clearer and does not involve counting the number of solutions of messy equations like in [32, 33]. Theorem 4.7.5. Let R be a finite local ring and let (V, β) be a symplectic space of dimension 2ν, where ν ≥ 1. |R|2ν − |M |2ν 1. The symplectic graph GSpR (V ) is |R|2ν−1 -regular on many vertices. |R× | 2. Every two adjacent vertices of GSpR (V ) has |R|2ν−2 R× common neighbors. 3. Every two non-adjacent vertices of GSpR (V ) has |R|2ν−2 |R× | or |R|2ν−1 common neighbors. 141 4.7. Symplectic Geometry Proof. By Theorem 4.7.4 (1) and (2), the number of vertices of GSpR (V ) is given by κ|M |2ν−1 = (|k|2ν − 1)|M |2ν |R|2ν − |M |2ν |k|2ν − 1 |M |2ν−1 = = |k| − 1 (|k| − 1)|M | |R× | Since the graph GSpk (V ′ ) is |k|2ν−1 -regular, Theorem 4.7.4 (3) implies that the graph GSpR (V ) is also regular of degree |k|2ν−1 |M |2ν−1 = |R|2ν−1 . For each pair of adjacent vertices R(~xi + m ~ 1 ) and R(~xj + m ~ 2 ) in the graph GSpR (V ) , the number of common neighbors is given by the product of the common neighbors of vertices kπ(~xi ) and kπ(~xj ) and |M |2ν−1 by Theorem 4.7.4 (4). Thus, λ = |k|2ν−2 (|k| − 1)|M |2ν−1 = |R|2ν−2 |R× |. Assume that R(~xi + m ~ 1 ) and R(~xj + m ~ 1 ) are non-adjacent vertices in GSpR (V ) . If i 6= j, then kπ(~xi ) and kπ(~xj ) are non-adjacent vertices in GSpk (V ′ ) , so the number of common neighbors of R(~xi + m ~ 1 ) and R(~xj + m ~ 2 ) is the product of common neighbors of kπ(~xi ) and kπ(~xj ) and 2ν−1 2ν−2 |M | which equals |R| |R× | by Theorem 4.7.4 (3) and (4). For i = j, it is easy to see that the number of common neighbors is the degree of regularity of GSpR (V ) . This proves the theorem. Let R be a local ring with unique maximal ideal M and let (V, β) be a symplectic space of rank 2ν, where ν ≥ 1. By Theorem 4.7.2, V possesses a canonical basis {~e1 , . . . , ~e2ν } such that {~ej , ~eν+j } is a hyperbolic pair for all 1 ≤ j ≤ ν and V is an orthogonal direct sum V = H1 ⊥H2 ⊥ . . . ⊥Hν , where Hj = R~ej ⊕ R~eν+j is a hyperbolic plane for all 1 ≤ j ≤ ν. Write unimodular vectors ~a = a1~e1 +· · ·+a2ν ~e2ν and ~b = b1~e1 +· · ·+b2ν ~e2ν for some ai , bi ∈ R. Then β(~a, ~b) = β(a1~e1 + . . . + a2ν ~e2ν , b1~e1 + . . . + b2ν ~e2ν ) = = 2ν 2ν X X ai bj β(~ei , ~ej ) i=1 j=1 ν X i=1 (ai bν+i − aν+i bi ) because β(~ei , ~ei ) = 0, β(~ei , ~eν+i ) = 1 and β(~ei , ~ej ) = −β(~ej , ~ei ) for all i, j ∈ {1, . . . , 2ν}. Hence, the adjacency condition becomes R~a is adjacent to R~b if and only if ν X i=1 (ai bν+i − aν+i bi ) ∈ R× . Next, let R be a finite commutative ring. By Corollary 4.6.7, R is a product of finite local rings and we have completely studied our graphs over a finite local ring. Write R = R1 × R2 × · · · × Rt as a direct product of finite local rings Ri , i = 1, 2, . . . , t. Consider V = R2ν , a free R-module of rank 2ν, where ν ≥ 1. We have the canonical 1-1 correspondence ϕ (j) (j) (j) ~x = (x1 , x2 , . . . , x2ν ) 7→ ((x1 )tj=1 , (x2 )tj=1 , . . . , (x2ν )tj=1 ). 142 4. Modules and Noetherian Rings Note that if ~x, ~y ∈ V , then this correspondence induces the symplectic map β on V by (j) (j) (j) (j) (j) (j) β(~x, ~y ) = β ((x1 )tj=1 , (x2 )tj=1 , . . . , (x2ν )tj=1 ), ((y1 )tj=1 , (y2 )tj=1 , . . . , (y2ν )tj=1 ) = (β1 (~x(1) , ~y (1) ), β2 (~x(2) , ~y (2) ), . . . , βt (~x(t) , ~y (t) )) ν ν ν X X X (1) (1) (1) (1) (2) (2) (2) (2) (t) (t) (t) (t) (xi yν+i − xν+i yi ), = (xi yν+i − xν+i yi ), . . . , (xi yν+i − xν+i yi ) , i=1 (j) i=1 (j) i=1 (j) where ~x(j) = (x1 , ~x2 , . . . , ~x2ν ) ∈ V (j) := Rj2ν and (V (j) , βj ) is a symplectic space of Rj of rank 2ν, for all j = 1, 2, . . . , t. Since R× = R1× × R2× × · · · × Rt× , we have β(~x, ~y ) ∈ R× ⇔ ν X i=1 (j) (j) (j) (j) (xi yν+i − xν+i yi ) ∈ Rj× for all j ∈ {1, 2, . . . , t}. (4.7.2) This shows that the adjacency condition does not depend on the bilinear map β. Recall from the previous paragraph that when Rj is a local ring, the adjacency condition becomes Rj~a is adjacent to Rj~b if and only if ν X i=1 (ai bν+i − aν+i bi ) ∈ Rj× . (4.7.3) for all j ∈ {1, 2, . . . , t}. Therefore, it follows from Eq. (4.7.3) that GSpR (V ) ∼ = GSpR 1 (V (1) ) ⊗ GSpanR 2 (V (2) ) ⊗ · · · ⊗ GSpR t (V (t) ) , (4.7.4) as a graph isomorphism. Here, for two graphs G and H, we define their tensor product G ⊗ H to be the graph with vertex set V(G) × V(H), where (u, v) is adjacent to (u′ , v ′ ) if and only if u is adjacent to u′ and v is adjacent to v ′ . From Theorem 4.7.5 (1) and the above discussion, we have the number of vertices of GSpR (V ) is equal to t t Y Y |Rj |2ν − |Mj |2ν |V(GSpR (V ) )| = |V(GSpR (V (j) ) )| = j |Rj× | j=1 j=1 and GSpR (V ) is regular of degree |R1 |2ν−1 |R2 |2ν−1 . . . |Rt |2ν−1 = |R|2ν−1 . Moreover, every two adjacent vertices of GSpR (V ) has |R|2ν−2 |R× | common neighbors by Theorem 4.7.5 (2). We record these results in the next theorem. Theorem 4.7.6. Let R be a finite commutative ring and (V, β) be the induced symplectic space of rank 2ν, ν ≥ 1, discussed above. 1. The symplectic graph GSpR (V ) is a |R|2ν−1 -regular and isomorphic to the graph GSpR 1 (V (1) ) ⊗ GSpR 2 (V (2) ) ⊗ · · · ⊗ GSpR t (V (t) ) . 2. Every two adjacent vertices of GSpR (V ) has |R|2ν−2 |R× | common neighbors. Remark. Other topics for symplectic graphs over a finite commutative ring such as vertex and arc transitivity, automorphism groups and the chromatic number can be found in the following exercises and [32, 33]. Exercises 4.7. 1. Let R be a local ring with unique maximal ideal M and let (V, β) be a symplectic space of R-dimension 2ν, where ν ≥ 1. Let ~a = a1~e1 + · · · + a2ν ~e2ν and ~b = b1~e1 + · · · + b2ν ~e2ν be unimodular vectors in V and assume that ai ∈ R× for some i ∈ {1, . . . , 2ν}. If R~a is adjacent to R~b, prove that ai bl − al bi is a unit for some l ∈ {1, . . . , 2ν} and l 6= i. 143 4.7. Symplectic Geometry 2. The chromatic number of a graph G is the smallest number of colors needed to color the vertices of G so that no two adjacent vertices share the same color. Tang and Wan [42] showed that if k is the field of q elements and V ′ is the symplectic space of dimension 2ν, ν ≥ 1, then the chromatic number of the symplectic graph GSpk (V ′ ) ) is q ν + 1. Let R be a local ring with unique maximal ideal M and residue field k and let (V, β) be a symplectic space of R-dimension 2ν, where ν ≥ 1. Determine the chromatic number of the symplectic graph GSpR (V ) . (Hint. Use the lifting theorem.) 3. Let G and H be graphs. A function σ from V(G) to V(H) is a homomorphism from G to H if σ(g1 ) and σ(g2 ) are adjacent in H whenever g1 and g2 are adjacent in G. It is called an isomorphism if it is a bijection and σ −1 is a homomorphism from H onto G. Moreover, an isomorphism on G is called an automorphism. The set of all automorphisms of a graph G is denoted by Aut (G). It is a group under composition, called the automorphism group of G. Prove that for graphs G and H, Aut (G) × Aut (H) ⊆ Aut (G ⊗ H). 4. A graph G is vertex transitive if its automorphism group acts transitively on the vertex set. That is, for any two vertices of G, there is an automorphism carrying one to the other. An arc in G is an ordered pair of adjacent vertices, and G is arc transitive if its automorphism group acts transitively on its arcs. Klingenberg [28] showed that for a local ring R, if {~x, ~a} and {~x, ~b} are hyperbolic pairs of unimodular vectors in V , then there exists an isometry σ in SpR (V ) which leaves ~x invariant and carries ~a to ~b. Let R be a finite local ring and let (V, β) be a symplectic space of dimension 2ν. Show that (a) SpR (V ) acts transitively on unimodular vectors and on hyperbolic planes. (b) The symplectic graph GSpR (V ) is vertex transitive and arc transitive. Show further that (b) holds for any finite commutative ring R. (Hint. Use 3.) Project 23 (Orthogonal graphs). Similar to symplectic graphs, we may study orthogonal graphs over a finite commutative ring defined as follows. Let R be a commutative ring and let V be a free R-module of rank n, where n ≥ 2. Assume that we have a function β : V × V → R which is R-bilinear, symmetric and the R− module morphism from V to V ∗ = homR (V, R) given by ~x 7→ β(·, ~x) is an isomorphism. For ~x ∈ V , we call β(~x, ~x) the norm of ~x. The pair (V, β) is called an orthogonal space. Let R be a commutative ring and let (V, β) be an orthogonal space, where V is a free R-module of rank n ≥ 2. A vector ~x in V is said to be unimodular if there is an f in homR (V, R) with f (~x) = 1; equivalently, if ~x = α1~b1 + . . . + αn~bn , where {~b1 , . . . , ~bn } is a basis for V , then the ideal (α1 , . . . , αn ) = R. If ~x is unimodular, then the line R~x is free R-direct summand of rank one. Moreover, it is easy to see that if ~x and ~y are unimodular vectors in V , then R~x = R~y if and only if ~x = λ~y for some λ ∈ R× . Define the graph GOR (V ) whose vertex set V(GOR (V ) ) is the set of lines {R~x : ~x is a unimodular vector in V and β(~x, ~x) = 0} and its adjacency condition is given by R~x is adjacent to R~y ⇐⇒ β(~x, ~y ) ∈ R× (or equivalently, β(~x, ~y ) = 1). We call GOR (V ) the orthogonal graph of (V, β) over R. (a) Show that the above adjacency condition is well defined. (b) If k is a finite field of odd characteristic and Vδ′ is an orthogonal space over k of dimension 2ν + δ, where ν ≥ 1 and δ ∈ {0, 1, 2}, then Gu and Wan [23] showed that GOk (Vδ′ ) is a |k|ν+δ−1 + 1-partite ν −1 ν+δ−1 graph with partite sets X1 , X2 , . . . , X|k|ν+δ−1 +1 such that |Xi | = |k| + |k|−1 for all i ∈ {1, 2, . . . , |k| 2ν+δ−2 1}. They also had that (Theorem 2.1 of [23]) the orthogonal graph GOk (Vδ′ ) is |k| -regular on (|k|ν − 1)(|k|ν+δ−1 + 1) many vertices. Moreover, if ν = 1, then it is a complete graph, and if ν ≥ 2, |k| − 1 then the graph is a strongly regular graph with parameters λ = |k|2ν+δ−2 − |k|2ν+δ−3 − |k|ν−1 + |k|ν+δ−2 and µ = |k|2ν+δ−2 − |k|2ν+δ−3 , respectively. Construct the lifting theorem for orthogonal graphs and determine the results similar to Theorem 4.7.5 for orthogonal graphs over a finite local ring. 144 4. Modules and Noetherian Rings (c) Let R be a finite commutative ring of odd characteristic and (Vδ , β) be an orthogonal space of rank 2ν + δ, ν ≥ 1 and δ ∈ {0, 1, 2}. Write R = R1 × R2 × · · · × Rt as a direct product of finite local rings Ri , i = 1, 2, . . . , t. Prove that the orthogonal graph GOR (Vδ ) is a |R|2ν−2+δ -regular and isomorphic to the graph GOR (V (1) ) ⊗ GOR (V (2) ) ⊗ · · · ⊗ GOR (V (t) ) . 1 δ 2 δ t δ 5 | Field Theory In Section 2.6, we learn about extensions of a field. Here, we give more details on a construction of extension fields. We prepare the readers to Galois theory which yields a connection between field theory and group theory. Applications of Galois theory are provided in proving fundamental theorem of algebra, finite fields, and cyclotomic fields. We discuss some results on a transcendental extension in the final section. 5.1 Splitting Fields Let F be a field. Given a polynomial f (x) ∈ F [x] we would like to have at hand an extension field E of F which in some sense contains all the roots of the equation f (x) = 0. We recall that f (r) = 0 if and only if f (x) is divisible by x − r. n Y c(x − ri ), that is, it is a product of We say that f (x) splits in an extension field E if f (x) = i=1 linear factors in E[x] and c ∈ F . We shall first study some facts about the roots of f (x) ∈ F [x] as follows. Theorem 5.1.1. If f (x) ∈ F [x] and deg f (x) = n ≥ 1, then f (x) can have at most n roots counting multiplicities in any extension field of F . Proof. We shall prove the theorem by induction on the degree of f (x). If deg f (x) = 1, then f (x) = ax + b for some a, b ∈ F and a 6= 0. Then −b/a is the unique root of f (x) and −b/a ∈ F , so we are done. Let deg f (x) = n > 1 and assume that the result is true for all polynomials of degree < n. Let E be any extension field of F . If f (x) has no roots in E, then we are done. Let r ∈ E be a root of f (x) of multiplicity m ≥ 1. Then there exists q(x) ∈ E[x] such that f (x) = (x − r)m q(x) and q(r) 6= 0. Thus, deg q(x) = n − m. By the inductive hypothesis q(x) has at most n − m roots in E counting multiplicities. Hence, f (x) has at most m+(n−m) roots in E counting multiplicities. Theorem 5.1.2. [Kronocker] If p(t) ∈ F [t] is irreducible over F , then there exists an extension field E of F such that [E : F ] = deg p(t) and p(t) has a root in E. Proof. Let E = F [x]/(p(x)) where x is an indeterminate. Then E is a field containing {a + (p(x)) : a ∈ F } as a subfield. But F ∼ = {a + (p(x)) : a ∈ F } by ϕ : a 7→ a + (p(x)), so E can be considered as an extension field of F by considering a as a + (p(x)) for all a ∈ F . Then E = F [x]/(p(x)) = F (t̄) where t̄ = x + (p(x)) is a root of p(t). Since E = F (t̄) and p(t) is irreducible over F , [E : F ] = [F (t̄) : F ] = deg p(t) by Corollary 2.6.5. Corollary 5.1.3. If p(t) ∈ F [t] is a nonconstant polynomial, then there exists a finite extension field E of F containing a root of p(t) and [E : F ] ≤ deg p(t). 145 146 5. Field Theory Proof. Since F [t] is a UFD, p(t) has an irreducible factor in F [t] say p1 (t). By Theorem 5.1.2, there exists an extension field E of F such that E contains a root of p1 (t) and [E : F ] = deg p1 (t). Hence, [E : F ] ≤ deg p(t) and E contains a root of p(t). Let F be a field and f (x) a monic polynomial in F [x]. An extension field E of F is a splitting field of f (x) over F if f (x) = (x − r1 ) . . . (x − rn ) in E[x] and E = F (r1 , . . . , rn ), that is, E is generated by the roots of f (x). The next results demonstrate the existence of a splitting field for a monic polynomial. Theorem 5.1.4. [Existence of Splitting Fields] Let f (x) be a monic polynomial of degree n ≥ 1. Then there exists an extension field E of F such that [E : F ] ≤ n! and E contains n roots of f (x) counting multiplicities. Hence, in E[t], f (x) = c(x − r1 ) . . . (x − rn ) for some c ∈ F and r1 , . . . , rn ∈ E, so that r1 , . . . , rn are n roots of f (x) in E. Proof. We shall prove the theorem by induction on the degree of f (x). If deg f (x) = 1, then f (x) has exactly one root in F and [F : F ] = 1 = 1!. Let deg f (x) = n > 1 and assume that the theorem is true for the case of polynomials of degree < n. By Corollary 5.1.3, there exists an extension field E0 of F such that f (x) has a root, say r ∈ E0 and [E0 : F ] ≤ n. Since r is a root of f (x), f (x) = (x − r)q(x) for some q(x) ∈ E0 [x], so deg q(x) = n − 1. By the inductive hypothesis, there exists an extension field E of E0 such that [E : E0 ] ≤ (n − 1)! and E contains n − 1 roots of q(x). Then E is an extension field of F , [E : F ] = [E : E0 ][E0 : F ] ≤ n! and E contains n roots of f (x) counting multiplicities. Corollary 5.1.5. Let F be a field and f (x) a nonconstant polynomial over F of degree n. Then there exists a splitting field E of f (x) over F . Moreover, [E : F ] ≤ n!. Proof. We have seen from Theorem 5.1.4 that there exists an extension field E of F such that f (x) = c(x − r1 ) . . . (x − rn ), for some c ∈ F and r1 , . . . , rn ∈ E, is a product of linear factors in E[x] and [E : F ] ≤ n!. Hence, E = F (r1 , . . . , rn ) is a desired field. Examples 5.1.1 (Examples of splitting fields). 1. Let f (x) = x2 + ax + b. If f (x) is reducible in F [x] (F arbitrary) then F is a splitting field. Otherwise, put E = F [x]/(f (x)) = F (r1 ) where r1 = x + (f (x)). Then E is a splitting field since f (r1 ) = 0, so f (x) = (x − r1 )(x − r2 ) in E[x]. Thus, E = F (r1 ) = F (r1 , r2 ). Since f (x) is the minimal polynomial of r1 over F , [E : F ] = 2. 2. Let the base field F be Z/(2), the field of two elements, and let f (x) = x3 + x + 1. Since 1 + 1 + 1 6= 0 and 0 + 0 + 1 6= 0, f (x) has no roots in F ; hence f (x) is irreducible in F [x]. Put r1 = x + (f (x)) in F [x]/(f (x)) so F (r1 ) is a field and x3 + x + 1 = (x + r1 )(x2 + ax + b) in F (r1 )[x]. (Note that we can write + for − since characteristic is two.) Comparison of coefficients shows that a = r1 , b = 1 + r12 . The elements of F (r1 ) can be listed as c+dr1 +er12 , c, d, e ∈ F . There are eight of these: 0, 1, r1 , 1+r1 , r12 , 1+r12 , r1 +r12 and 1+r1 +r12 . Substituting these in x2 + r1 x + 1 + r12 , we reach (r12 )2 + r1 (r12 ) + 1 + r12 = r14 + r13 + 1 + r12 = 0 since r13 = r1 + 1 and r14 = r12 + r1 . Hence, x2 + ax + b factors into linear factors in F (r1 )[x] and E = F (r1 ) is a splitting field of x3 + x + 1 over F . 3. Let F = Q, f (x) = (x2 − 2)(x2 − 3). Since the rational roots of x2 − 2 and x2 − 3 must be integral, it follows that x2 − 2 and x2 − 3 are irreducible in Q[x]. Form K = Q(r1 ), r1 = x + (x2 − 2) in Q[x]/(x2 − 2). The elements of K have the form a + br1 , a, b ∈ Q. We claim that x2 − 3 is irreducible in K[x]. Otherwise, we have rational numbers a, b such that 147 5.1. Splitting Fields 2 2 (a + br1 )2 = 3. Then (a2 + 2b2 ) + 2abr1 = √ 3 so that ab = 0 and a + 2b =23. If b = 0 we 2 obtain a = 3 which√ is impossible since 3 is not rational, and if a = 0, b = 3/2. Then 2 (2b ) = 6 and since 6 is not rational, we again obtain an impossibility. Thus, x2 − 3 is irreducible in K[x]. Now form E = K[x]/(x2 − 3). Then this is a splitting field over Q of (x2 − 2)(x2 − 3) and [E : Q] = [E : K][K : Q] = 2 · 2 = 4. 4. Let F = Q, f (x) = xp − 1, p a prime. We have xp − 1 = (x − 1)(xp−1 + xp−2 + · · · + x + 1) and we know that xp−1 + xp−2 + · · · + x + 1 is irreducible in Q[x]. Let E = Q(z) where z = x + (xp−1 + xp−2 + · · · + x + 1) in Q[x]/(xp−1 + xp−2 + · · · + x + 1). We have 1, z, . . . , z p−1 are distinct. Also (z k )p = (z p )k = 1 so every z k is a root of xp − 1. It follows that xp − 1 = Q p k is a splitting field over Q of xp − 1 and [E : Q] =√p − 1. k=1 (x − z ) in E[x].√Thus, E √ √ 3 2) is not a 5. Since x3 − 2 = (x − 3 2)(x − 3 2ω)(x − 3 2ω 2 ) where ω 6= 1 and ω 3 = 1, Q( √ 3 3 splitting field of f (x) = x − 2 over Q. √A splitting field of f (x) is√E = Q( 2, ω). Since g(x) = √ x2 + x + 1 is irreducible over Q( 3 2) and g(ω) = 0, [E : Q( 3 2)] = 2, so [E : F ] = √ [E : Q( 3 2)][Q( 3 2) : Q] = 2 · 3 = 6. e 6. A splitting field over Z/(p) of xp − 1, e ∈ N, is Z/(p). Theorem 5.1.6. [Uniqueness of Splitting Fields] Let η : F → F1 be an isomorphism of fields and let η : F [x] → F1 [x] be the isomorphism which extends η and satisfies η(x) = x. Suppose f (x) is a monic polynomial in F [x], let f1 (x) = η(f (x)) and suppose that E/F and E1 /F1 are splitting fields of f (x) and f1 (x), respectively. Then there exists an isomorphism η ∗ : E → E1 which extends η. Proof. Let fˆ(x) be an irreducible factor of f (x) and let fˆ1 (x) = η(fˆ(x)). Let r ∈ E be a root of fˆ(x) and let r1 ∈ E1 be a root of fˆ1 (x). Then we have a commutative diagram in which the vertical arrows are isomorphisms and the horizontal arrows are inclusion maps F / F [r] O /E i F [x]/fˆ(x)F [x] η η̂ F1 [x]/fˆ1 (x)F1 [x] j F1 / F1 [r1 ] / E1 . The map j η̂i−1 = η̄ is an isomorphism of fields extending η. Also, η̄(f (x)/(x − r)) = f1 (x)/(x − r) and E/F [r], E1 /F1 [r1 ] are splitting fields of f (x)/(x − r) and f1 (x)/(x − r1 ), respectively. Now, by induction on deg f (x), η̄ : F [r] → F1 [r1 ] has an extension to η ∗ : E → E1 and this is the required extension of η. Theorem 5.1.7. Assume f (x) has no multiple factors as an element of F [x]. Under the hypothesis of Theorem 5.1.6, the number of distinct extensions of η : F → F1 to η : E → E1 is at most [E : F ]. Moreover, the number of distinct extensions is equal to [E : F ] if and only if f (x) has distinct roots in E. Proof. Proceeding as in the proof of Theorem 5.1.6, let fˆ(x) be an irreducible factor of f (x), let d be the degree of fˆ(x), let fˆ1 (x) = η(fˆ(x)), let r1 , . . . , re be the distinct roots of fˆ(x) in E and let r1′ , . . . , re′ be the roots of fˆ1 (x) in E1 . (Note that e ≤ d and e = d if fˆ1 (x) has no multiple roots, but this is not always the case.) 148 5. Field Theory Next fix a root r = r1 of fˆ(x). The argument of Theorem 5.1.6 shows that for each root of fˆ1 (x) there is an isomorphism η̄j : F [r] → F1 [rj′ ] extending η, where η̂j (r) = rj′ . r1′ , . . . , re′ / F [r] F η / F1 [r ′ ] F1 / E1 j On the other hand, any isomorphism of F [r] into E1 must carry r into a root of fˆ1 (x), and so must one of the η̄j . Furthermore, as noted above the number of roots of fˆ(x) = e ≤ d = [F [r] : F ]. By induction, the number of ways each η̂j can be extended to an isomorphism E → E1 is at most [E : F [r]]. Thus, the number of extensions of η : F → F1 to η ∗ : E → E1 ≤ e[E : F [r]] ≤ [F [r] : F ][E : F [r]] = [E : F ]. Now we want to answer the question: When is there equality – that is, the number of extensions = [E : F ]? Looking at the first step above we see that the number of roots of fˆ(x) = e = d = [F [r] : F ] if and only if fˆ(x) has d = deg fˆ(x) roots – that is if and only if fˆ(x) has distinct roots. To continue inductively, we now have the set up F [r] /E η̂j F1 [rj′ ] / E1 The key point is that E is the splitting field over F [r] of the polynomial f (x)/(x − r). This polynomial has no multiple factor so inductively the number of extensions of η̂j to an isomorphism η ∗ : E → E1 is equal to [E : F [r]] if and only if f (x)/(x−r) has distinct roots. Combining this with the result for fˆ(x) we get the number of extensions of η : F → F1 to an isomorphism η : E → E1 is equal to [E : F ] if and only if f (x) has distinct roots in E. Remarks. (1) If f (x) is an irreducible polynomial over a field F and r is a root of f (x) in some extension field of F , then F [x]/f (x)F [x] ∼ = F [r]. However, if f (x) = g(x)h(x) where g(x) and h(x) are irreducible polynomials, then by Chinese remainder theorem F [x]/f (x)F [x] ∼ = F [x]/g(x)F [x] × F [x]/h(x)F [x] a direct product of fields. If f (x) = g(x)2 , then F [x]/f (x)F [x] even has nilpotent elements. In general, E/F arises from a succession of simple extensions F ⊆ F1 ∼ = F [x]/f1 (x)F [x], ∼ F1 ⊆ F2 = F1 [x]/f2 (x)F1 [x], .. . Fr−1 ⊆ Fr ∼ = Fr−1 [x]/fr (x)Fr−1 [x] = E. 149 5.2. Algebraic Closure of a Field We shall see that in some important cases (but not all), the splitting field E/F of the polynomial f (x) can be achieved as a simple extension F ⊆ F [x]/g(x)F [x] = E, but usually g(x) 6= f (x). (2) If f (x) and g(x) have the same roots in some extension field E of F (f (x), g(x) ∈ F [x]), then they have the same splitting field. However, one cannot guarantee that the roots of f (x) are distinct (or simple, or one fold). The basic example is the polynomial f (x) = xp − a ∈ F [a] where F is a field of characteristic p > 0. If r is a root of f (x) in some extension field E of F [a], then rp = a and the factorization of f (x) in E[x] is f (x) = xp − a = xp − rp = (x − r)p . Exercises 5.1. 1. Construct a splitting field over Q of x5 − 2. Find its dimension over Q. 4 2. Let f (x) = x + x2 + 1. Find the splitting field of f (x) over Q and determine its dimension. 3. Let E/F be a splitting field of over F of f (x) and let K be a subfield of E/F . Show that any monomorphism of K/F into E/F can be extended to an automorphism of E. 4. If f (x) ∈ F [x] has degree n and K is a splitting field of f (x) over F , prove that [K : F ] divides n!. 5. Let F be a field of characteristic p > 0 and let b ∈ F . Show that either xp − b is irreducible in F [x] or b = ap and xp − b = (x − a)p for some a ∈ F . 5.2 Algebraic Closure of a Field We know about the prime field which is the smallest field such that every other field is an extension of it. However, we does not know if we can algebraically extend our field F forever to obtain a field that every polynomial in F [x] has a root in it. We shall assure it in this section. A field F is called algebraically closed if every monic polynomial f (x) of positive degree with coefficients in F has a root in F . Theorem 5.2.1. Let F be a field. The following statements are equivalent. (i) F is algebraically closed. (ii) An irreducible polynomial in F [x] is linear, and hence every polynomial of F [x] of positive degree is a product of linear factors. (iii) F has no proper algebraic extension field. Proof. Since r is a root, that is f (r) = 0, if and only if x − r is a factor of f (x) in F [x], we have (i) ⇔ (ii). Next, we show (i) ⇔ (iii). If E is an extension field of F and a ∈ E is algebraic over F , then [F (a) : F ] is the degree of the minimal polynomial f (x) of a over F , and f (x) is monic and irreducible. Then a ∈ F if and only if deg f (x) = 1. Hence, E is algebraic over F and E ⊃ F implies there exist irreducible monic polynomials in F [x] of degree ≥ 2; hence F is not algebraically closed. Conversely, if F is not algebraically closed, then there exists a monic irreducible f (x) ∈ F [x] with deg f (x) ≥ 2, Thus, the field F [x]/(f (x)) is a proper algebraic extension of F . We recall that (Corollary 2.6.7) if E is an extension field of the field F , then the set of elements of E that are algebraic over F constitutes a subfield A of E/F (that is, a subfield of E containing F ). Evidently E = A if and only if E is algebraic over F . At the other extreme, if A = F , then F is said to be algebraically closed in E. In any case A is algebraically closed in E, since any element of E that is algebraic over A is algebraic over F and so is contained in A. This result shows that if a field F has an algebraically closed extension field, then it has one that is algebraic 150 5. Field Theory over F . We call an extension field E/F an algebraic closure of F if E is algebraic over F and E is algebraically closed. For example, assuming the truth of the fundamental theorem of algebra (Theorem 5.5.6), that C is algebraically closed, it follows that the field of A of algebraic numbers is an algebraic closure of Q, and thus A is algebraically closed. We proceed to prove the existence and uniqueness up to isomorphism of an algebraic closure of any field F . For a countable F a straightforward argument is available to establish these results. We begin by enumerating the monic polynomials of positive degree as f1 (x), f2 (x), . . . . Evidently this can be done. We now define inductively a sequence of extension fields beginning with F0 = F and letting Fi be a splitting field over Fi−1 of fi (x). The construction of such splitting fields was given at the end of the previous section. It is clear that every Fi is countable, so we can realize all S of these constructions in some large set S. Then we can take E = Fi in the set. Alternatively we can define E to be a direct limit of the fields Fi . It is easily seen that E is an algebraic closure of F . We showed that (Theorem 5.1.6) there exists an isomorphism of K1 /F onto K2 /F . This can be used to prove the isomorphism theorem for algebraic closures of a countable field by a simple inductive argument. The pattern of the proof sketched above can be carried over to the general case by using “transfinite induction”. This is what was done by E. Steinitz, who first proved these results. There are several alternative proofs available that are based on Zorn’s lemma. We shall give one that makes use of the following lemma. Lemma 5.2.2. If E is an algebraic extension of a field F , then the cardinality of E cannot exceed the cardinality of F [x]. Proof. Let S be the set of all ordered pairs (f, α) where f (x) ∈ F [x] is nonzero and α ∈ E with f (α) = 0. Since for each polynomial f (x), the number of α such that (f, α) lies in S is finite, we have |S| ≤ |F [x]|ℵ0 = |F [x]|. On the other hand, E maps injectively into S via α 7→ (fα , α) where fα is the minimal polynomial of α, and thus |E| ≤ |S|. Recall that |F [x]| = |F |ℵ0 . If F is infinite, then |F [x]| = |F | and it follows that |E| = |F |. When F is finite, F [x] is countable, and hence E is either finite or countably infinite. Corollary 5.2.3. There exist real numbers transcendental over Q. Proof. There are only countably many polynomials in Q[x]. Since R is uncountable, the above lemma guarantees that R is not algebraic over Q. We can now prove the existence of algebraic closures. Theorem 5.2.4. Any field F has an algebraic closure. Proof. We first embed F in a set S in which we have a lot of elbow room. Precisely, we assume that |S| > |F | if F is infinite and that S is uncountable if F is finite. We now define a set Λ whose elements are (E, +, ·) where E is a subset of S containing F and +, · are binary compositions in E such that (E, +, ·) is an algebraic extension field of F . We partially order Λ by declaring that (E, +, ·) > (E ′ , +′ , ·′ ) if E is an extension field of E ′ . By Zorn’s lemma there exists a maximal element (E, +, ·). Then E is an algebraic extension of F . We claim that E is algebraically closed. Otherwise we have a proper algebraic extension E ′ = E(a) of E. Then |E ′ | < |S|, so we can define an injective map of E ′ into S that is the identity on E and then we can transfer the addition and multiplication on E ′ to its image. This gives an element of Λ bigger than (E, +, ·). This contradiction shows that E is an algebraic closure of F . 5.2. Algebraic Closure of a Field 151 Next we take up the question of uniqueness of algebraic closures. It is useful to generalize the concept of a splitting field of a polynomial to apply to sets of polynomials. If Γ = {fα (x)} is a set of monic polynomials with coefficients in F , then an extension field E/F is called a splitting field over F of the set Γ if 1. every fα (x) ∈ Γ is a product of linear factors in E[x] and 2. E is generated over F by the roots of the fα (x) ∈ Γ. It is clear that if E is a splitting field over F of Γ, then no proper subfield of E/F is a splitting field of Γ and if K is any intermediate field, then E is a splitting field of Γ. Since an algebraic closure E of F is algebraic, it is clear that E is a splitting field over F of the complete set of monic polynomials of positive degree in F [x]. The isomorphism theorem for algebraic closures will therefore be a consequence of a general result on isomorphisms of splitting fields that we shall now prove. Our starting point is the following result, which is Theorem 5.1.6. Let η : a 7→ ã be an isomorphism of a field F onto a field F̃ , f (x) ∈ F [x] be monic of positive degree, f˜(x) the corresponding polynomial in F̃ [x] (under the isomorphism, which is η on F and sends x 7→ x), and let E and Ẽ be splitting fields over F and F̃ of f (x) and f˜(x), respectively. Then η can be extended to an isomorphism of E onto Ẽ. We shall now extend this to a set of polynomials. Theorem 5.2.5. Let η : a 7→ ã be an isomorphism of a field F onto a field F̃ , Γ a set of monic polynomials fα (x) ∈ F [x], Γ̃ the corresponding set of polynomials f˜(x) ∈ F̃ [x], E and Ẽ splitting fields over F and F̃ of Γ and Γ̃, respectively. Then η can be extended to an isomorphism of E onto Ẽ. Proof. The proof is a straightforward application of Zorn’s lemma. We consider the set of extensions of η to monomorphisms of subfields of E/F into Ẽ/F̃ and use Zorn’s lemma to obtain a maximal one. This must be defined on the whole E, since otherwise we could get a larger one by applying the result quoted to one of the polynomials fα (x) ∈ Γ. Now if ζ is a monomorphism of E into Ẽ such that ζ|F = η, then it is clear that ζ(E) is a splitting field over F̃ of Γ̃. Hence, ζ(E) = Ẽ and ζ is an isomorphism of E onto Ẽ. As we have observed, this result applies in particular to algebraic closures. If we take F̃ = F and η = id , we obtain Theorem 5.2.6. Any two algebraic closures of a field F are isomorphic over F . From now on we shall appropriate the notation F̄ for any determination of an algebraic closure of F . If A is any algebraic extension of F , its algebraic closure Ā is an algebraic extension of A, hence of F , and so Ā is an algebraic closure of F . Consequently, we have an isomorphism of Ā/F into F̄ /F . This maps A/F into a subfield of F̄ /F . Thus, we see that every algebraic extension A/F can be realized as a subfield of the algebraic closure F̄ /F . Exercises 5.2. 1. No finite field F is algebraically closed. [Hint. If F = {0, 1, a2 , . . . , an }, consider the polynomial 1 + x(x − 1)(x − a2 ) . . . (x − an ) ∈ F [x].] 2. Let E be an algebraic extension of a field F and A an algebraic closure of F . Show that E/F is isomorphic to a subfield of A/F . [Hint. Consider the algebraic closure Ā of A and note that this is an algebraic closure of F .] 152 5.3 5. Field Theory Multiple Roots and Separability Recall the following facts from Subsection 2.6.2 about the multiple roots. Let R be an integral domain and f (x) ∈ R[x]. If α is a root of f (x), then there exist m ∈ N and g(x) ∈ R[x] such that f (x) = (x − α)m g(x) and g(α) 6= 0. m is called the multiplicity of the root α of f (x) and if m > 1, α is called a multiple root of f (x). If f (x) = a0 + a1 x + · · · + an xn ∈ R[x], we define f ′ (x) ∈ R[x], the derivative of f (x), to be f ′ (x) = a1 + a2 x + · · · + nan xn−1 . We record the straightforward properties of the derivative of polynomials in the next lemma. Lemma 5.3.1. If f (x) and g(x) are polynomials over an integral domain R and c ∈ R, then 1. (cf (x))′ = cf ′ (x), 2. (f (x) + g(x))′ = f ′ (x) + g ′ (x), 3. (f (x)g(x))′ = f (x)g ′ (x) + f ′ (x)g(x), 4. ((f (x))n )′ = n(f (x))n−1 f ′ (x) where n ∈ N. Theorem 5.3.2. Let E be an extension of a field F and f (x) ∈ F [x]. 1. For α ∈ E, α is a multiple root of f (x) if and only if α is a root of both f (x) and f ′ (x). 2. If f (x) and f ′ (x) are relatively prime, then f (x) has no multiple root. 3. If f (x) is irreducible over F having a root in E, then f (x) has no multiple root in E if and only if f ′ (x) 6= 0. Proof. (1) is clear. (2) Since f (x) and f ′ (x) are relatively prime, there exist h(x) and k(x) in F [x] such that 1 = f (x)h(x) + f ′ (x)k(x). If α ∈ E is a multiple root of f (x), by (1), f (α) = 0 = f ′ (α), so 1 = 0, a contradiction. (3) Since f (x) is irreducible, f ′ (x) 6= 0 and deg f ′ (x) < deg f (x), we have f (x) and f ′ (x) are relatively prime, so f (x) has no multiple roots. Conversely, if f ′ (x) = 0, then f (α) = 0 = f ′ (α) for some α ∈ E since f (x) has a root in E. Hence, by (1), α is a multiple root of f (x). Let F be a field. A polynomial f (x) ∈ F [x] is separable if every root (in some splitting field over F ) of its irreducible factor is not a multiple root. If E is an extension of F and α ∈ E is algebraic over F , then α is separable over F if its minimal polynomial over F is separable. Let F ⊂ K ⊂ E be field extensions. Note that if α is separable over F , then α is separable over K since mα,K (x) | mα,F (x). Here mα,− (x) stands for the minimal polynomial of α over the indicated field. Examples 5.3.1. 1. Consider f (x) = x2 +1. Over Q, we have f (x) is irreducible and separable but over Z/(2), we have f (x) = x2 + 1 = (x + 1)2 is not irreducible but is separable since the only irreducible factor is x + 1 which is separable over Z/(2). 2. Let K be a field of characteristic p and F = K(y) be the field of rational functions over K with indeterminate y. Since K[y] is UFD, y is irreducible element in K[y], so the polynomial f (x) = xp − y in F [x] is irreducible over F by Eisenstien criterion. Since f ′ (x) = 0 and f (x) has a root, say α in some splitting field E of F , α is a multiple root of f (x), so f (x) is not separable over F . However, if we consider f (x) = xp − y ∈ E[x], we have f (x) = (x − α)p and its irreducible factor in E[x] is only x − α which is separable over E, so f (x) is separable over E. Suppose that F is a field of characteristic zero and f (x) is a monic irreducible polynomial over F , say f (x) = a0 + a1 x + · · · + an−1 xn−1 + xn . Then f ′ (x) = a1 + 2a2 x + · · · + nxn−1 . The key point is that n 6= 0, so f ′ (x) 6= 0. Since deg f ′ (x) < deg f (x) and f (x) is irreducible, f (x) and f ′ (x) are relatively prime, so all roots of f (x) are simple. Thus, we have shown: 153 5.3. Multiple Roots and Separability Theorem 5.3.3. Let F be a field of characteristic zero. Then every polynomial f (x) ∈ F [x] is separable. We call an algebraic extension field E of a field F a separable extension if the minimal polynomial of every element of E is separable. Hence, if F is of characteristic zero, then every algebraic extension is a separable extension. A field F is perfect if every polynomial f (x) over F is separable. Thus, all fields of characteristic zero are perfect. Remark. Suppose F is a field (or even a commutative ring) of characteristic p > 0. Then the identities (ab)p = ap bp and (a + b)p = ap + bp show that the map ϕ : F → F defined by ϕ(a) = ap is a ring homomorphism. Since F is a field, ϕ has to be one-to-one. But ϕ does not have to be onto - for example ϕ : (Z/pZ)(x) → (Z/pZ)(x) n is not onto; the image is (Z/pZ)(xp ). However, if F is finite of order pn , then ap = a for all a ∈ F , so ϕ is onto and ϕn is the identity map, called the Frobenius’ automorphism. Theorem 5.3.4. Let F be a field of characteristic p > 0, and let a ∈ F . (1) If a ∈ F p and a = rp , then xp − a = (x − r)p . (2) If a ∈ / F p , then xp − a is irreducible. Proof. (1) is trivial. (2) In a splitting field for F , xp − a = (x − r)p (r may not be in F ). Any proper factor of xp − a (after being made monic) has the form (x − r)i where 1 ≤ i ≤ p − 1. Thus, if xp − a has a proper factor over F , then ri ∈ F for some 1 ≤ i ≤ p − 1. But then ri and rp = a ∈ F , so r ∈ F since (i, p) = 1. Hence, a = rp ∈ F p . Theorem 5.3.5. Let F be a field of characteristic p > 0. Then F is perfect if and only if F = F p . Proof. Suppose F 6= F p and choose a ∈ F r F p . By Theorem 5.3.4, xp − a is irreducible. But xp − a does not have distinct roots in a splitting field of F . Hence, F is not perfect. Conversely, assume that F is not perfect. Then there is an irreducible polynomial f (x) over F which does not have simple roots. By Theorem 5.3.2, this means that f (x) and f ′ (x) are not relatively prime. Since f (x) is irreducible and deg f ′ (x) < deg f (x), f ′ (x) = 0. Thus, f (x) is a polynomial in xp , i.e., f (x) = a0 + ap xp + a2p x2p + · · · + a(m−1)p x(m−1)p + xmp . We shall claim that some ajp ∈ / F p . For if each ajp ∈ F p , say ajp = (bj )p , then f (x) = g(x)p where g(x) = b0 + b1 x + · · · + bm−1 xm−1 + xm which contradicts the irreducibility of f (x) over F . This establishes the claim. Hence, ajp ∈ / Fp and F 6= F p . Corollary 5.3.6. Every finite field is perfect. Proof. The characteristic of a finite field F is a prime p. The monomorphism a 7→ ap of F is an isomorphism since F is finite. Hence, F = F p is perfect by Theorem 5.3.5. 154 5. Field Theory We shall end this section by proving the “primitive element theorem” which is a classic of field theory. We first recall that an extension field E of a field F is said to be a simple extension of F if E = F (α) for some α ∈ E. Such an element α is called a primitive element. Theorem 5.3.7. If F is a field and G is a finite subgroup of the multiplicative group of nonzero elements of F , then G is a cyclic group. In particular, the multiplicative group of all nonzero elements of a finite field is cyclic. Proof. If G = {1}, then G is cyclic. Assume that G 6= {1}. Since G is a finite abelian group, G∼ = Z/(m1 ) ⊕ · · · ⊕ Z/(mk ) where k ≥ 1, m1 > 1 and m1 | · · · | mk . Since mk ( k X Z/(mi )) = 0, u is a root of the polynomial i=1 xmk − 1 ∈ F [x] for all u ∈ G. By Theorem 5.1.1, this polynomial has at most mk distinct roots in F , we must have k = 1 and G ∼ = Z/(m1 ) which is a cyclic group. Theorem 5.3.8. [Primitive Element Theorem] Let E be a finite separable extension of a field F . Then there exists α ∈ E such that E = F (α). That is, a finite separable extension of a field is a simple extension. Proof. If F is a finite field, then E is also finite. Let α be a generator for the cyclic group of all nonzero elements of E under multiplication. Clearly, E = F [α], so α is a primitive element in this case. We now assume that F is infinite and prove our theorem in the case that E = F (β, γ). The induction argument from this to the general case is obvious. Let mβ,F (x) and mγ,F (x) be the minimal polynomials over F of β and γ, respectively. Assume that mβ,F (x) has distinct roots β = β1 , . . . , βn and mγ,F (x) has distinct roots γ = γ1 , . . . , γm in F̄ where all roots are of multiplicity 1, since E is a separable extension of F . Since F is infinite, we can find a ∈ F such that a 6= βi − β γ − γj for all i and j, with j 6= 1. That is, a(γ − γj ) 6= βi − β. Letting α = β + aγ, we have α = β + aγ 6= βi + aγj , so α − aγj 6= βi for all i and all j 6= 1. Consider h(x) = mβ,F (α − ax) ∈ F (α)[x]. Now, h(γ) = mβ,F (β) = 0. However, h(γj ) 6= 0 for j 6= 1 by construction, since the βi were the only roots of mβ,F (x). Hence, h(x) and mγ,F (x) have a common factor in F (α)[x], namely the minimal polynomial of γ over F (α), which must be linear, since γ is the only common root of mγ,F (x) and h(x). Thus, γ ∈ F (α), and therefore β = α − aγ is in F (α). Hence, F (β, γ) = F (α). Exercises 5.3. 1. Suppose that F ⊆ K ⊆ E and that E is separable extension of F . Prove that E is separable over K and K is separable over F . 2. Let F be of characteristic p and let a ∈ F . Show that f (x) = xp − x − a has no multiple roots and f (x) is irreducible in F [x] if and √ only if a 6= cp − c for any c ∈ F . 3. Find a primitive element of Q(i, 3 2) over Q. 4. Let K = F25 be the field with 5 elements and let F = Z/(5) be the prime subfield of K. Determine the cardinalities of the following two sets. (a) The set of elements of K which generate K as a field over F . (b) The set of elements of K which generate the group of nonzero elements of K as an abelian group under multiplication. 155 5.4. Automorphisms of Fields and Galois Theory 5. Let F be a field and let F be its algebraic closure. If a monic polynomial p(x) ∈ F [x] is irreducible over F and has distinct roots α1 , α2 , . . . , αk ∈ F , prove that the multiplicities of αj are equal, that is, p(x) = (x − α1 )m (x − α2 )m . . . (x − αk )m for some m ∈ N. 5.4 Automorphisms of Fields and Galois Theory If F is a field, the set of automorphisms of F , Aut F , forms a group under composition of functions. Examples 5.4.1 (Examples of automorphism groups). 1. Any automorphism satisfies ϕ(1) = 1, so ϕ(n) = n for all n ∈ Z and ϕ(n/m) = n/m if n, m ∈ Z and m 6= 0 in F . This implies that the fields Q and Fp = Z/(p) have only the identity map as an automorphism. That is, Aut (F ) = {idF } if F = Q or Fp . Moreover, any field E is an extension of Q or Fp (so called the prime subfield) and any automorphism ϕ : E → E leaves the prime subfield pointwise fixed. 2. The only automorphism ϕ : R → R is the identity map. For, we have known that ϕ(q) = q √ √ for all q ∈ Q. Note that ϕ(a) = ϕ(( a)2 ) = (ϕ( a))2 > 0 for all a > 0. Thus, if a < b, then ϕ(a) < ϕ(b). Let x ∈ R. Suppose ϕ(x) 6= x. Then ϕ(x) < x or ϕ(x) > x. If ϕ(x) < x, then there exists a q ∈ Q such that ϕ(x) < q < x. Thus, q = ϕ(q) < ϕ(x), a contradiction. If x < ϕ(x), then there exists a q ∈ Q such that x < q < ϕ(x), so ϕ(x) < ϕ(q) = q, a contradiction. Hence, ϕ = id R . 3. Complex conjugation: ϕ(z) = z̄ is an automorphism of C of order two. In fact, Aut C is uncountable, but the other automorphisms are “indescribable” and exist only via Zorn’s lemma. However, the group of automorphisms of C which fix all elements of R is a group of order two. 4. Let F be a field and let E = F (t) where t is transcendental over F . As shall be indicated in the Exercise 5.4 below, u ∈ E is a generator of E/F if and only if it has the form u= αt + β , γt + δ αδ − βγ 6= 0. Since an automorphism of E/F sends generators into generators, it follows that every automorphism ϕ : E → E is given by ϕ(a) = a for all a ∈ F and ϕ(t) = αt + β , γt + δ where α, β, γ, δ ∈ F and αδ − βγ 6= 0. Note that if c ∈ F and c 6= 0, then α β cα cβ and γ δ cγ cδ give rise to the same automorphism of F (t). A computation shows composition of functions corresponds to matrix multiplication. The net result is that Aut F (t) ∼ = GL2 (F )/F × = PGL2 (F ), where F × is the set of matrices aI, a 6= 0. 5. If F is a subfield of K, let Aut F K = {ϕ ∈ Aut K : ϕ(a) = a for all a ∈ F }. The group structure of Aut F F (x, y) is known, but very complicated. For n ≥ 3, almost nothing is known about Aut F F (x1 , . . . , xn ). 156 5. Field Theory The above examples show that Aut F is in general very complicated and probably impossible to describe. Galois theory proceeds in a different direction. One takes a subgroup H of Aut F —we shall be almost concerned with finite H—and looks the set F H = {a ∈ F : ϕ(a) = a for all ϕ ∈ H}. It is easy to see that F H is a subfield of F . Moreover, if K is a subgroup of H, then 1⊆K⊆H F ⊇ F K ⊇ F H. The fundamental result of Galois theory is that of F is separable over F H , then there is a oneto-one correspondence between subgroups of H and subfields of F which contain F H . Such correspondences are inclusion reversing and are called “Galois correspondences”. Let E be an extension field of a field F . The Galois group of E over F denoted by Gal(E/F ) is the group {ϕ ∈ Aut E : ϕ(a) = a for all a ∈ F }. Let G be a subgroup of Aut E where E is a field. Then the field of G-invariant of E or the fixed field of G acting on E is the field {a ∈ E : ϕ(a) = a for all ϕ ∈ G}. It is denoted by E G or Inv G. Theorem 5.4.1. (1) If Aut E ⊇ G1 ⊇ G2 , then E G1 ⊆ E G2 . (2) If E ⊇ F1 ⊇ F2 , then Gal(E/F1 ) ⊆ Gal(E/F2 ). (3) If G = Gal(E/F ), then E G ⊇ F . (4) If F = E G , then Gal(E/F ) ⊇ G. Proof. These are immediate consequences of the definitions. We shall now apply these ideas to splitting fields. Using the present terminology, Theorem 5.1.7 can be restated as follows. If E is a splitting field over F of a polynomial f (x), then Gal(E/F ) is finite and we have the inequality |Gal(E/F )| ≤ [E : F ]. Moreover, |Gal(E : F )| = [E : F ] if f (x) has distinct roots. We therefore have the following important preliminary result. Lemma 5.4.2. Let E/F be a splitting field of a separable polynomial contained in F [x]. Then |Gal(E/F )| = [E : F ]. Our next attack will be from the group side. We begin with an arbitrary field E and any finite group of automorphisms G acting in E. Then we have the following Lemma 5.4.3. [Artin] Let G be a finite subgroup of Aut E and let F = E G . Then [E : F ] ≤ |G|. Proof. Let |G| = n and write G = {g1 = 1, g2 , . . . , gn }. We have to show that [E : F ] ≤ n, or equivalently: (∗) If x1 , . . . , xn+1 ∈ E, then there exist u1 , . . . , un+1 ∈ F not all zero, such that u1 x1 + · · · + un+1 xn+1 = 0, 157 5.4. Automorphisms of Fields and Galois Theory that is, x1 , . . . , xn+1 are linearly dependent over F . Consider the following n × (n + 1) matrix with entries in E x1 x2 ... g2 (x1 ) g2 (x2 ) . . . M = . .. gn (x1 ) gn (x2 ) . . . xn+1 g2 (xn+1 ) . gn (xn+1 ) This matrix has rank ≤ n, so there is a nonzero (n + 1) × 1 vector ~v = (v1 , . . . , vn+1 )t with entries in E such that M~v = ~0(n+1)×1 . We wish to find such a vector where entries lie in F . Among all such vectors with entries in E, choose one in which the number of nonzero coordinates, r, is minimal. By renaming the elements x1 , . . . , xn+1 , we may suppose that the nonzero coordinates are the first r of them; by multiplying the vector by vr−1 we may suppose that the last nonzero coordinate is equal to 1. Thus, M~v = ~0(n+1)×1 where ~v = (v1 , . . . , vr−1 , 1, 0, . . . , 0)t . Claim. If h ∈ G and h(~v ) = (h(v1 ), . . . , h(vr−1 ), 1, 0, . . . , 0)t , then M h(~v ) = ~0. Proof of Claim. The inner product of the j-th row of M with h(~v ) is: z = gj (x1 )h(v1 ) + · · · + gj (xr−1 )h(vr−1 ) + gj (xr ) · 1. Apply the automorphism h−1 to z, h−1 z = h−1 gj (x1 )h(v1 ) + · · · + h−1 gj (xr−1 )h(vr−1 ) + h−1 gj (xr ) · 1 = gi (x1 )v1 + · · · + gi (xr−1 )vr−1 + gi (xr ) · 1 = 0, since h−1 gj = gi for some i. This proves the claim. Now we consider, for any h ∈ G ~v − h(~v ) = (v1 , . . . , vr−1 , 1, 0, . . . , 0)t − (h(v1 ), . . . , h(vr−1 ), 1, 0, . . . , 0)t r−1 z }| { = (∗, . . . , ∗, 0, . . . , 0)t . Since M (~v − h(~v )) = ~0 and ~v − h(~v ) has at most r − 1 nonzero entries, ~v − h(~v ) = ~0 by the minimal choice of r. This means that for all h ∈ G and i = 1, . . . , r − 1, we have h(vi ) = vi . Thus, all the vi lie in E G = F and (u1 , . . . , un+1 ) = (v1 , . . . , vr−1 , 0, . . . , 0) is a set of elements of F which satisfies (∗). Recall that an algebraic extension field E of a field F is a separable extension if the minimal polynomial of every element of E is separable. We call an algebraic extension field E of a field F a normal extension if every irreducible polynomial in F [x] which has a root in E splits into linear factors in E. This is equivalent to saying that E contains a splitting field for the minimal polynomial of every element of E. Normality plus separability, called a Galois extension, mean that every irreducible polynomial of F [x] which has a root in E is a product of distinct linear factors in E[x]. Also, by the results of the last section, if E is algebraic over F , then E is necessarily separable over F if the characteristic is zero or if the characteristic is p > 0 and F p = F . We are now ready to derive our main results, the first of which gives two abstract characterizations of splitting fields of separable polynomials and some important additional information. We state this as 158 5. Field Theory Theorem 5.4.4. Let E be an extension field of a field F . Then the following conditions on E/F are equivalent. (i) E is a splitting field over F of a separable polynomial f (x). (ii) F = E G for some finite group G of automorphisms of E. (iii) E is finite dimensional Galois (normal and separable) over F . Moreover, if E and F are as in (i) and G = Gal(E/F ), then F = E G and if G and F are as in (ii), then G = Gal(E/F ). Proof. (i) ⇒ (ii). Let G = Gal(E/F ). Then E G is a subfield of E containing F . Also it is clear that E is a splitting field over E G of f (x) as well as over F and G = Gal(E/E G ). Hence, by Lemma 5.4.2, |G| = [E : F ] and |G| = [E : E G ]. Since E ⊇ E G ⊇ F , we have [E : F ] = [E : E G ][E G : F ]. Hence, [E G : F ] = 1, and so E G = F . We have prove also that F = E G for G = Gal(E/F ), which is the first of the two supplementary statements. (ii) ⇒ (iii). By Artin’s lemma, [E : F ] ≤ |G|, and so E is finite dimensional over F . Let f (x) be an irreducible polynomial in F [x] having a root r in E. Let {r = r1 , r2 , . . . , rm } be the orbit of r under the action of G. Thus, this is the set of distinct elements of the form σ(r), σ ∈ G. Hence, if σ ∈ G, then the set {σ(r1 ), σ(r2 ), . . . , σ(rm )} is a permutation of {r1 , r2 , . . . , rm }. We have f (r) = 0 which implies that f (ri ) = 0. Then Q f (x) is divisible by x − ri , and since the ri , 1 ≤ i ≤ m, are distinct, f (x) is divisible by g(x) = m the automorphism i=1 (x − ri ). We now apply to g(x) Q m of QmE[x], which sends x → x and a → σ(a) for a ∈ E. This gives σg(x) = i=1 (x − σ(ri )) = i=1 (x − ri ) = g(x). Since this holds for every σ ∈ G we see that the coefficients of g(x) are G-invariant.QHence, g(x) ∈ F [x]. Since we assumed f (x) irreducible in F [x] we see that f (x) = g(x) = (x − ri ), a product of distinct linear factors in E[x]. Thus, E is separable and normal over F and (iii) holds. (iii) ⇒ (i). Since we are given that [E : F ] < ∞ we can write E = F (r1 , r2 , . . . , rk ) and each ri is algebraic over F . Let fi (x) be the minimal polynomial of ri over F . Then the hypothesis Q implies that fi (x) is a product of distinct linear factors in E[x]. It follows that f (x) = ki=1 fi (x) is separable and E = F (r1 , r2 , . . . , rk ) is a splitting field over F of f (x). Hence, we have (i). It remains to prove the second supplementary statement. We have seen that under the hypothesis of (ii) we have [E : F ] ≤ |G|, and that since (i) holds, we have |Gal(E/F )| = [E : F ]. Since G ⊆ Gal(E/F ) and |G| ≥ [E : F ] = |Gal(E/F )|, equivalently G = Gal(E/F ). The above proof also yields Corollary 5.4.5. If E/F is the splitting field of f (x) ∈ F [x] and r1 , . . . , rn are distinct roots of f (x) in E, then G = Gal(E/F ) may be identified with a subgroup of Sn , the group of permutations of {r1 , . . . , rn }. However, it is not always the case that Gal(E/F ) is the full group of permutations of the roots of f (x). There are two observations underlying the above corollary. 1. Each σ ∈ G permutes r1 , . . . , rn . 2. σ ∈ G is determined by its action on r1 , . . . , rn because r1 , . . . , rn generate E as a field over F , i.e., E = F [r1 , . . . , rn ] = F (r1 , . . . , rn ). Example 5.4.2 (Elementary symmetric functions). If K is a field, then the polynomial ring K[x1 , . . . , xn ] is an integral domain. The quotient field of K[x1 , . . . , xn ] is denoted by K(x1 , . . . , xn ) and is called the field of rational functions in x1 , . . . , xn over K. In the field extension K ⊂ K(x1 , . . . , xn ) each xi is easily seen to be transcendental over K. In fact, every element of K(x1 , . . . , xn ) not in K itself is transcendental over K (Prove!). 159 5.4. Automorphisms of Fields and Galois Theory Let Sn be the symmetric group on n letters. A rational function ϕ ∈ K(x1 , . . . , xn ) is said to be symmetric in x1 , . . . , xn over K if for every σ ∈ Sn , ϕ(x1 , x2 , . . . , xn ) = ϕ(xσ(1) , xσ(2) , . . . , xσ(n) ). Trivially, every constant polynomial is a symmetric function. More generally, the elementary symmetric functions in x1 , . . . , xn over K are defined to be the polynomials: e1 = x 1 + x 2 + · · · + x n = X e2 = xi xj ; n X xi ; i=1 1≤i<j≤n .. . X ek = x i1 x i2 . . . x ik ; 1≤i1 <···<ik ≤n .. . en = x 1 x 2 . . . x n . The verification that the ei are indeed symmetric follows from the fact that they are simply the coefficients of t in the polynomial p(t) ∈ K[x1 , . . . , xn ][t], where p(t) = (t − x1 )(t − x2 ) . . . (t − xn ) = tn − e1 tn−1 + e2 tn−2 − · · · + (−1)n−1 en−1 t + (−1)n en . If σ ∈ Sn , then the assignments xi 7→ xσ(i) , i = 1, 2, . . . , n and f (x1 , . . . , xn )/g(x1 , . . . , xn ) 7→ f (xσ(1) , . . . , xσ(n) )/g(xσ(1) , . . . , xσ(n) ) define a K-automorphism of the field E = K(x1 , . . . , xn ) which will also be denoted σ. The map Sn → Gal(E/K) given by σ 7→ σ is clearly a monomorphism of groups, whence Sn may be consider as a subgroup of the Galois group Gal(E/K). Clearly, the fixed field F = E Sn consists precisely of symmetric functions; that is, the set of all symmetric functions is a subfield of E containing K. Therefore, by Theorem 5.4.4, E is a Galois extension of F with Galois group Gal(E/F ) = Sn and dimension |Sn | = n!. Example 5.4.3. Let K be a field and x1 , x2 , x3 be indeterminates over K, set e1 = x 1 + x 2 + x 3 , e2 = x 1 x 2 + x 2 x 3 + x 3 x 1 , e3 = x 1 x 2 x 3 and consider the fields F = K(e1 , e2 , e3 ) ⊆ K(x1 , x2 , x3 ) = E. The relevant subfields of E are indicated in the diagram ♦♦ 3♦♦♦♦ ♦♦♦ ♦♦♦ F = K(e1 , e2 , e3 ) 3 PPP PPP3 PPP PPP F (x1 ) P PPP PPP2 PPP PP F (x2 ) 2 K(x1 , x2 , x3 ) = E ♥♥ 2♥♥♥♥ F (x3 ) ♥♥♥ ♥♥♥ The fields F (x1 ), F (x2 ) and F (x3 ) are all isomorphic (over F ), but they are distinct subfields of E. Moreover, E is a splitting field for f (t) = t3 − e1 t2 + e2 t − e3 but F (x1 ), F (x2 ) and F (x3 ) are not. 160 5. Field Theory We know that G = Gal(E/F ) = S3 where S3 is identified with the group of permutations on 3 letters. We next calculate E H when H is a subgroup of G = Gal(E/F ) = S3 . The following is a diagram of the lattice of subgroups of S3 and there indices. ❥ S3 ❙ ❍❍❙❙❙❙ ❥❥❥✈❥✈✈ ❍❍ ❙❙❙ ❍❍ ❙❙2❙ ✈✈ ❥ ❙❙❙ ❥ ❍❍ ✈ ❥ ❥ ✈ 3 3 ❥ ❙❙❙ ❍ ✈ ❥ ✈ ❙❙ ❥❥❥ h(12)i ❚❚ h(13)i h(23)i ❦❦ A3 ❍ ❚❚❚❚ ✈ ❦❦❦ ✈ ❦ ❚❚❚❚ ❍❍❍❍2 ❦ ✈ 2✈ ❦❦❦ ✈ ❚❚❚❚ ❍❍❍ ✈✈ ❦❦❦ 3 ❚❚❚❚❍ 2 ✈✈❦❦❦❦❦ 3❥❥❥❥❥ {(1)} We have already calculated that E S3 = E G = F and of course E {(1)} = E. It is not hard to verify that E h(12)i = F [x3 ], E h(13)i = F [x2 ], E h(23)i = F [x1 ]. It is somewhat more difficult to verify that E A3 = F [∆] where ∆ = (x1 − x2 )(x2 − x3 )(x3 − x1 ). Note that σ(∆) = ∆ if σ ∈ A3 , but σ(∆) = −∆ if σ ∈ S3 r A3 . We already know that [F [x1 ] : F ] = [F [x2 ] : F ] = [F [x3 ] : F ] = 3 and one can verify that [F [∆] : F ] = 2. Thus, we get the following diagrams of all (by Galois Theory) subfields of E containing F F [x3 ] = E h(12)i F = E SP3 ❳❳❳ PPP ❳❳❳❳❳ ❢❢❢❢♥♥♥ ❢ ❢ ❢ ❢ ❳❳ PPP 3 ❢❢❢❢ ♥♥ ♥ ❢ ❢ PPP ❳❳❳2❳❳❳❳❳❳ ♥ ❢ ❢❢❢ ♥♥ 3 3 ❢ ❳❳❳❳❳ ❢ P ♥ ❢ ❢ P ♥ ♥ ❢❢❢ ❳ h(13)i ❳❳❳❳❳ F [x2 ] = E PPP ❳❳❳❳❳ PPP2 ❳❳❳❳❳ P ❳ ❳ ❳ ❳❳❳❳P❳PPP 2 ❳ F [x1 ] = E h(23)i ❣ F [∆] = E ♥♥ ❣❣❣❣❣ ❣ ❣ 2 ♥♥♥ ❣ ❣ ❣ ♥♥♥ ❣❣❣❣❣3❣ ♥♥❣♥❣❣❣❣❣❣ A3 E = E {(1)} The indices are the same as in the lattice diagram for S3 , but inclusions are reversed. Recall that E is the splitting field of a separable polynomial f (t) = (t − x1 )(t − x2 )(t − x3 ) for any field in the above diagram. More generally, it is clear that if M/L is a splitting field for f (t) ∈ L[t] and M ⊇ N ⊇ L, then M/N is a splitting field for f (t), regarded as a polynomial in N [t]. Furthermore, for each field L in the above diagram, we have L = E H for some subgroup H of G = S3 and Gal(E/L) = H. On the other hand, things are not so nice for the extensions L/F . For example, Gal(F [xi ]/F ) = 1 for all i = 1, 2, 3 and Gal(F [∆]/F ) ∼ = Z/(2) = hϕi where the action of ϕ is ϕ(∆) = −∆. Here ∆2 ∈ F and F [∆] is the splitting field of the polynomial t2 − ∆2 over F , so it is Galois. However, we may conclude that the fields F [x1 ], F [x2 ] and F [x3 ] are not the splitting fields of any polynomials over F . The previous example illustrates the fundamental theorem of Galois theory: if E/F is the splitting field of a separable polynomial f (t) ∈ F [t], then the map H ←→ E H = {a ∈ E : ϕ(a) = a for all ϕ ∈ H} 161 5.4. Automorphisms of Fields and Galois Theory is a 1-1 correspondence between subgroups of Gal(E/F ) ←→ subfields of E which reverses inclusions. In addition, H is a normal subgroup of Gal(E/F ) if and only if E H is the splitting field of some separable polynomial over F (i.e., E H is normal over F ), and if H is normal in Gal(E/F ), then Gal(E H /F ) ∼ = Gal(E/F )/H. In our example, the only proper normal subgroup of S3 is A3 , and Gal(E A3 /F ) = Gal(F [∆]/F ) ∼ = S3 /A3 = Gal(E/F )/A3 . = Z2 ∼ We now formally establish Galois’ fundamental group-field pairing as follows. Theorem 5.4.6. [Fundamental Theorem of Galois Theory] Let E be a finite dimensional Galois extension of a field F (i.e., the conditions of Theorem 5.4.4 holds) and let G = Gal(E/F ). Let Γ = {H}, the set of subgroups of G, and Σ, the set of intermediate fields between E and F (the subfields of E/F ). Then the map H 7→ E H and K 7→ Gal(E/K), H ∈ Γ, K ∈ Σ, are inverses to each other. In particular, they are one-to-one correspondences between Γ and Σ. Moreover, the pairing Γ ↔ Σ has the following properties: 1. H1 ⊇ H2 if and only if E H1 ⊆ E H2 . 2. |H| = [E : E H ] and [G : H] = [E H : F ] = [E H : E G ]. 3. H is normal in G if and only if E H is normal over F . In this case, Gal(E H /F ) ∼ = G/H. This is the main theorem. Most of our remaining field theory will be consequences of it. Proof. Let H be a subgroup of G = Gal(E/F ). Since F = E G , F ⊆ E H and E H is thus a subfield of E containing F . Also, E/E H is Galois. Applying the second supplementary result of Theorem 5.4.4 to H in place of G we see that Gal(E/E H ) = H. By Lemma 5.4.2, |H| = |Gal(E/E H )| = [E : E H ]. Now let K be any subfield of E/F . Then Gal(E/K) ⊆ G = Gal(E/F ), so Gal(E/K) is a subgroup of G. It is clear also that E is a splitting field over K of a separable polynomial. Hence, the first supplementary result of Theorem 5.4.4 applied to the pair E and K shows that K = E Gal(E/K) . We have now shown that the specified maps between Γ and Σ are inverses. Also, we know that if H1 ⊇ H2 , then E H1 ⊆ E H2 . Moreover, if E H1 ⊆ E H2 , then we have also that H1 = Gal(E/E H1 ) ⊇ Gal(E/E H2 ) = H2 . Hence, (1) holds. The first part of (2) was noted before. Since |G| = [E : F ] = [E : E H ][E H : F ] = |H|[E H : F ] and |G| = |H|[G : H], evidently [E H : F ] = [G : H]. This proves (2). −1 If H ∈ Γ, then E ηHη = η(E H ) for all η ∈ G. This is clear since the condition σ(x) = x is equivalent to (ηση −1 )(η(x)) = η(x). It now follows that H is normal in G if and only if η(E H ) = E H for every η ∈ G. Suppose H is normal in G. Then every η ∈ G maps E H onto itself and so its restriction η̄ = η|E H is an automorphism of E H /F . Thus, we have the restriction homomorphism η → η̄ of G = Gal(E/F ) into Gal(E H /F ). The image Ḡ is a group of automorphisms in E H and clearly (E H )Ḡ = F . Hence, Ḡ = Gal(E H /F ). The kernel of the homomorphism η → η̄ is the set of η ∈ G such that η|E H = idE H . By the pairing, this is Gal(E/E H ) = H. Hence, the kernel is H and Ḡ = Gal(E H /F ) ∼ = G/H. Since F = (E H )Ḡ , E H is normal over F by Theorem 5.4.4. Conversely, H suppose E is normal over F . Let a ∈ E H and let f (x) be the minimal polynomial of a over F . Then f (x) = (x − a1 ) . . . (x − am ) in E H [x] where a = a1 . If η ∈ G, then f (η(a)) = 0 which implies that η(a) = ai for some i. Thus, η(a) ∈ E H . We have therefore shown that η(E H ) = E H . Hence, H is a normal subgroup of G. This completes the proof of (3). 162 5. Field Theory √ √ 2)(x2 − 3). Then E is Example 5.4.4. Let E = Q[ 2, 3] be a splitting field of f (x) = (x2 − √ Galois over Q. Let G = Gal(E/Q). Then |G| = [E : Q] = 4. Since Q( 2) is a splitting field of x2√− 2, it is √ Galois over Q and its Galois group consists of 2 elements, namely σ1 = id and σ : 2 → 7 − ways; √ √ 2. Each √ automorphism extends to an automorphism of E √in two√different √ √ √2 3 7→ 3 or 3 → 7 − 3. Then the four elements of G are τ = id , τ : 2 → 7 2, 3 → 7 − 3, 1 2 E √ √ √ √ √ √ √ √ τ3 : 2 7→ − 2, 3 7→ 3 and τ4 : 2 7→ − 2, 3 7→ − 3. Each of these elements except τ1 has order 2. Thus, G ∼ = Z2 × Z2 . Hence, the subgroup-intermediate subfield correspondence for the fundamental theorem of Galois theory is shown in the lattice diagrams {idE } hτ2 i ②② ②② ② ② ②② ❊❊ ❊❊ ❊❊ ❊❊ hτ3 i ❋❋ ❋❋ ❋❋ ❋❋ ❋ G ①① ①① ① ①① ①① hτ4 i ♠ E ◗◗◗◗ ♠♠♠ ◗◗◗ ♠ ♠ ◗◗◗ ♠♠ ♠ ♠ ◗◗◗ ♠♠♠ ◗ ♠ √ √ √ Q[ 3] = E hτ3 i Q[ 6] = E hτ4 i Q[ 2] = E hτ2 i ◗◗◗ ♠♠♠ ◗◗◗ ♠♠♠ ◗◗◗ ♠ ♠ ♠ ◗◗◗ ◗◗ ♠♠♠♠♠ Q Exercises 5.4. 1. Let E = F (t) where t is transcendental over F and write any non-zero element of E as u = f (t)/g(t) where (f (t), g(t)) = 1. Call the maximum of degrees of f and g the degree of u. Show that if x and y are indeterminates then f (x) − yg(x) is irreducible in F [x, y] and hence is irreducible in F (y)[x]. Show that t is algebraic over F (u) with minimal polynomial the monic polynomial which is a multiple in F (u) of f (x) − ug(x). Hence, conclude that [F (t) : F (u)] = 1, and F (u) = F (t) if and only if deg u = 1. Note that this implies u= 2. 3. 4. 5. 6. 7. 8. 5.5 at + b ct + d where ad − bc 6= 0. Therefore, deduce that Gal(E/F ) is the set of maps h(t) 7→ h(u) where u is of the form indicated. Let F ⊆ K ⊆ E and E Galois over F . Prove that E is Galois over K. Show that every element of K(x1 , . . . , xn ) which is not in K is transcendental over K. Show that in the subgroup-intermediate subfield correspondence given in the fundamental theorem of Galois theory, the subfield corresponding to the intersection of two subgroups H1 and H2 is the subfield generated by the composite field E H1 E H2 , the smallest subfield of E generated by E H1 and E H2 , and the intersection of two intermediate fields K1 and K2 corresponds to the subgroup generated by Gal(E/K1 ) ∪ Gal(E/K2 ). Use the fact that any finite group G is isomorphic to a subgroup of Sn (Cayley’s theorem) to prove that given any finite group G, there exist fields E and E/F such that Gal(E/F ) = G. Let E = Q(r) where r3 + r2 − 2r − 1 = 0. Verify that r′ = r2 − 2 is also a root of x3 + x2 − 2x − 1 = 0. Determine Show that E is normal over Q. p Gal(E/Q). √ Let α = 2 + 2 in R, f (x) the minimal polynomial of α over Q and E is a splitting field of f (x) over Q. (a) Compute f (x) and [E : Q]. (b) Find G = Gal(E/Q) and draw a lattice diagram for the subgroup-intermediate subfield correspondence for the fundamental theorem of Galois theory. Let (Z/(p))(t) where t is transcendental over Z/(p). Let G be the group of automorphisms generated by the automorphism of E such that t 7→ t + 1. Determine F = E G and [E : F ]. Some Consequences of Galois Theory In this section, we shall derive some consequences of Galois theory including another proof of the fundamental theorem of algebra. 163 5.5. Some Consequences of Galois Theory Theorem 5.5.1. Let K be a finite dimensional separable extension of a field F . Then there are only finitely many fields L such that K ⊇ L ⊇ F . Proof. Since K/F is finite separable, by primitive element theorem, K = F [α] for some α ∈ K. Let E be the splitting field of mα,F (x). Then E is Galois over F and E ⊆ K ⊆ F . By fundamental theorem of Galois theory, the number of intermediate fields between E and F is the number of subgroups of Gal(E/F ). Hence, the number of intermediate fields between K and F is at most the number of subgroups of Gal(E/F ). Remark. If G = Gal(E/F ), then K = E H for some subgroup H of G and the fields L such that K ⊇ L ⊇ F are in 1-1 correspondence with the subgroups J of G such that G ⊇ J ⊇ H. The primitive element theorem and the previous theorem both fail for inseparable extensions as shown in the following example. Example 5.5.1. Let F be an infinite field of prime characteristic p and let u and v be indeterminates over F . Consider F (u, v) ⊇ F (up , v p ) It is easy to see that [F (u, v) : F (up , v p )] = p2 . On the other hand, if z ∈ F (u, v), then z p ∈ F (up , v p ), so [F (up , v p )(z) : F (up , v p )] ≤ p. Hence, there is no z such that F (u, v) = F (up , v p )(z), that is, no primitive element. On the other hand, the nonexistence of a primitive element shows that the fields F (up , v p )(u + αv), for α ∈ F , are all distinct. To see this, assume that F (up , v p )(u + αv) = F (up , v p )(u + βv) = E for some α 6= β in E. Then u + αv and u + βv in E, so α(u + βv) − β(u + αv) = (α − β)u ∈ E. Since α − β 6= 0, u is in E which implies that v is also in E. Thus, E = F (u, v), a contradiction. Hence, there are infinitely many fields L such that F (u, v) ⊃ L ⊃ F (up , v p ). Let us now recall some concepts from group theory. Suppose a group G acts on a set S. The action is transitive if for any s, t ∈ S there is a g ∈ G such that gs = t. Remark. The action of G being transitive simply means that the action of G on S has only one orbit. Assuming G acts transitively on S. let s ∈ S and let H = {g ∈ G : gs = s} be the stabilizer of s. Then S can be identified with the set of left cosets {gH : g ∈ G}, with G acting by left multiplication. Note that the subgroup H depends on the choice of s and choosing a different s will give a conjugate of H. More precisely, if s ∈ S and x ∈ G, and H = stabilizer of s = {g ∈ G : gs = s} then xHx−1 = stabilizer of xs = {g ∈ G : g(xs) = xs}. (If gs = s, then (xgx−1 )(xs) = xs.) A basic example of this phenomenon is the action of Sn on {1, 2, . . . , n}. The stabilizer of i ∈ {1, 2, . . . , n} is Sym{1, . . . , i − 1, i + 1, . . . , n} which may be identified with Sn−1 , but Sn−1 has n conjugates in Sn . 164 5. Field Theory Theorem 5.5.2. Let E be the splitting field over F of a separable polynomial f (x) ∈ F [x] which is irreducible over F . Then Gal(E/F ) acts transitively on the roots of f (x). Hence, Gal(E/F ) may be identified with a subgroup of Sym{r1 , . . . , rn } which acts transitively on {r1 , . . . , rn }, the roots of f (x) in E. Proof. This is implicit in the proof of Theorem 5.1.6. For, if r and s are roots of f (x) in E, then F (r) ∼ = F (s) = F [x]/(f (x)) ∼ with r 7→ x + (f (x)) 7→ s by an isomorphism which fixes F pointwise. Let η : F (r) → F (s) be this isomorphism. By Theorem 5.1.6, η extends to an isomorphism η̂ : E → E. Then η̂ ∈ Gal(E/F ) and η̂(r) = s, which is what we need to prove. Remarks. 1. The hypothesis that f (x) be irreducible over F is essential. For, example, if f (x) = f1 (x) . . . fk (x) where f1 (x), . . . , fk (x) are distinct irreducible polynomials, then all one can say is that Gal(E/F ) permutes the roots of each fi (x) among themselves. It is still true that Gal(E/F ) can be identified with a subgroup of the group of permutations of the roots, but not a transitive one. 2. Assume that f (x) is irreducible and separable over F of degree n, E/F is a splitting field for f (x) over F and r is one root of f (x). Then the fundamental theorem of Galois theory gives the following picture E {id E } F [r] Gal(E/F [r]) = H n n F Gal(E/F ). Thus, F [r] = E H where H is a subgroup of index n in G. The basic Theorems 5.1.4 and 5.1.6 give the existence and uniqueness of splitting fields. That is, if F is a field and f (x) is a monic polynomial in F [x], then 1. A splitting field E for f (x) exists. E is generated over F by the roots of f (x) and f (x) splits into linear factors in E[x]. 2. The splitting field E/F is unique up to isomorphism over F . In other words, if E ′ /F is another splitting field for f (x) over F , then there is an isomorphism ϕ : E → E′ which is identity on F . What does this means if we are searching for the splitting field of some f (x) ∈ Q[x]? It means that we can realize E as a subfield of C. More precisely, f (x) is a product of linear factors in C[x], say f (x) = (x−α1 ) . . . (x−αk ) and we can take E to be the field Q(α1 , . . . , αk ) ⊆ C. This could be very helpful because it allows us to work in a concrete and explicit field. The fundamental theorem of algebra (every f (x) ∈ C[x] is a product of linear factors) is usually proved in complex analysis and there is also a topological proof. Here we present a proof based on Galois theory and the intermediate value theorem from real analysis or calculus. We shall start with some basic results. Theorem 5.5.3. Let f (x) ∈ R[x] be a polynomial of odd degree. Then f (x) has a root in R. 165 5.5. Some Consequences of Galois Theory Proof. It is enough to prove for a monic polynomial f (x) = xn + an−1 xn−1 + · · · + a1 x + a0 with ai ∈ R and n is odd. If a = |a0 | + · · · + |an−1 |, then it is easy to see that f (a) > 0 and f (−a) < 0. By intermediate value theorem (because f (x) is continuous), there exists r ∈ R such that f (r) = 0. p Consider α + βi with α, β ∈ R. If γ = α2 + β 2 , then p p ( (γ + α)/2 + i (γ − α)/2)2 = α + βi. Hence, we have proved Theorem 5.5.4. Every complex number has a square root. Theorem 5.5.5. If K is a field containing C, then [K : C] 6= 2. Proof. Suppose conversely that [K : C] = 2 and let K = C + Cu for some u ∈ K. Then u satisfies a polynomial f (x) = x2 − bx + c of degree two over C, since 1, u, u2 are linearly dependent over C. The roots of f (x) are √ −b ± b2 − 4ac 2 which lie in C, since every element of C has a square root in C. Thus, u ∈ C, a contradiction. Recall that a finite p-group G is nilpotent, so by Exercise 3.3, a maximal subgroup M of G is normal and [G : M ] = p, i.e., if G is a nontrivial finite p-group, then G has a normal subgroup of index p. Theorem 5.5.6. [Fundamental Theorem of Algebra] Let f (x) ∈ C[x]. Then f (x) is a product of linear factors in C[x]. Proof. Let ¯: C → C denote the complex conjugation. Then g(x) = f (x)f (x) ∈ R[x]. Let E be a splitting field for g(x)(x2 + 1) over R and identify C with the subfield of E generated by the roots of x2 + 1. Since the characteristic is zero, all polynomials are separable, so E is the splitting field of a separable polynomial. Hence, E is Galois over R by Theorem 5.4.4. Let G = Gal(E/R), |G| = 2a m, where m is odd, and let P be a Sylow 2-subgroup of G. Consider the diagram of fields E❆ ✈✈ ✈✈ ✈ ✈✈ ✈✈ ❆❆ a ❆❆2 ❆❆ ❆ ❍❍ ❍❍ ❍❍ 2 ❍❍❍ ⑥⑥ ⑥⑥ ⑥ ⑥⑥ m ⑥⑥ EP R[i] = C R Thus, E P = {α ∈ E : ϕ(α) = α for all ϕ ∈ P } is an extension of R of odd degree m, by the fundamental Galois correspondence. If u ∈ E P , the minimal polynomial q(x) of u over R is an irreducible polynomial in R[x] of odd degree, so it has a root in R by Theorem 5.5.3. Since q(x) is irreducible, it has degree one. Hence, E P = R and G = P , so |G| = 2a . By the fundamental 166 5. Field Theory theorem of Galois theory, C = E H where H is a subgroup of G of index 2. If H 6= {1}, it has a subgroup K of index 2, so R 2 C = EH 2 EK E. Thus, [E K : C] = 2 which contradicts Theorem 5.5.5. Hence, |G| = 2, H = {1} and C = E H = E. Therefore, C is a splitting field for g(x)(x2 + 1) = f (x)f (x)(x2 + 1) over R, so g(x)(x2 + 1) (and hence f (x)) splits into linear factors in C[x]. The fundamental theorem of algebra was first rigorously proved by Gauss in 1816 (his doctoral dissertation in 1798 provides a proof using geometric considerations requiring some topological justification). There was a proof due to Laplace in 1795. However, Laplace’s proof was deemed unacceptable because he assumed the existence of a splitting field for polynomials (i.e., that the roots existed somewhere in some field), which had not been established at that time. The elegant above proof was given by Artin. 5.6 Finite Fields Let k be a field of q elements. Then (k, +) is an abelian group, so q · 1 = 0. Thus, F is of characteristic prime p > 0 and p | q, so it contains Z/pZ as a subfield and it is a finite extension of Z/pZ. Its cardinality |k| = q = pd is a power of p, with d = [k : Z/pZ]. This also indicates that the additive group of k is a direct sum of d copies of cyclic group of order p. We shall restate the following fact (Theorem 5.3.7). Theorem 5.6.1. k × is cyclic of order q − 1. Some immediate consequences of the above theorem are as follows. Corollary 5.6.2. The field k consists of the solutions to xq − x = 0 in an algebraic closure of Z/pZ containing k. Corollary 5.6.3. There is an element α ∈ k such that k = (Z/pZ)[α], that is, k is a simple extension of the prime field Z/pZ. Corollary 5.6.4. For each positive divisor r of q − 1(= |k × |) there are exactly φ(r) elements in k × of order r. Corollary 5.6.5. Let p be a prime and d a positive integer. Then, up to isomorphism, there is exactly one field of order q = pd . d Proof. Let E be a splitting field of f (t) = tp − t over Z/pZ in an algebraic closure of Z/pZ. By d Theorem 5.1.6, E is unique up to isomorphism. It consists of the roots of tp = t in the algebraic d closure of Z/pZ. Thus, |E| is the number of roots of tp − t. Since f ′ (t) = −1, f (t) is separable, so |E| = pd . Thus, we have constructed a field of order q = pd , namely E, the splitting field of f (t) over Z/pZ. For q = pd , we may write Fq for the (unique up to isomorphism) field of q elements. Also, we may write Fp for Z/pZ. 167 5.6. Finite Fields Corollary 5.6.6. Given any positive integer d, there exists an irreducible polynomial of degree n over Fp . Proof. By Corollary 5.6.3, Fpd = Fp [α] for some α ∈ Fpd . Let f (t) be the minimal polynomial of α over Fp . Then Fpd = Fp [α] ∼ = Fp [t]/(f (t)) shows deg f (t) = [Fpd : Fp ] = d. Next, we shall study finite extensions of a finite field. For simplicity, k stands for the finite field Fq . Let kn be a degree n field extension of k. If km is an intermediate field of degree m over k, then kn is a vector space over km , so m divides n. Conversely, any degree m extension of k within an algebraic closure of k with m | n is a subfield of kn by Corollary 5.6.2 since m | n implies (q m − 1) | (q n − 1). Consider the map σ on kn which sends x to xq . From σ(x + y) = (x + y)q = xq + y q = σ(x) + σ(y) and σ(xy) = (xy)q = xq y q = σ(x)σ(y), we see that σ is an endomorphism. Furthermore, σ(x) = xq = 0 implies x = 0. So σ is one-to-one. As kn is finite, we have shown that σ is an automorphism of kn . Finally, σ(x) = xq = x for x ∈ k, this shows that σ ∈ Gal(kn /k), called the Frobenius’ automorphism. Let r be the order of σ. Then r σ r (x) = xq = x for all x ∈ kn implies r = n since kn× is cyclic of order q n − 1. Hence, Gal(kn /k) contains the cyclic group hσi of order n. Since |Gal(kn /k)| ≤ [kn : k] = n, Gal(kn /k) = hσi and so the field kn is Galois over k. We record this in Theorem 5.6.7. The field kn is Galois over k with the Galois group Gal(kn /k) cyclic of order n, generated by the Frobenius’ automorphism σ. Note that an element x ∈ kn lies in k if and only if it satisfies xq = x, in other words, if and only if it is fixed by the Frobenius’ automorphism, or equivalently, by the group Gal(kn /k). Using G = Gal(kn /k), we define two important maps, called trace and norm, denoted by Trkn /k and Nkn /k , respectively, from kn to k as follows: Trkn /k : x 7→ Nkn /k : x 7→ X τ (x) = τ (x) = τ ∈G σ i (x), i=1 τ ∈G Y n X n Y σ i (x). i=1 One check easily that the images of trace and norm maps are in k. It is clear that Trkn /k is a homomorphism from the additive group kn to the additive group k and Nkn /k is a homomorphism from kn× to k × . Next we investigate their images. We shall first need Lemma 5.6.8. If E is an extension field of a field F , then the automorphisms in Gal(E/F ) are E-linearly independent F -linear transformations. Proof. Suppose otherwise. Let a1 τ1 + · · · + ar τr = 0 be a shortest nontrivial linear relation with a1 , . . . , ar ∈ E × and P τ1 , . . . , τr ∈ Gal(E/F ). Then r ≥ 2 and τi are distinct. Let y ∈ E be such that τ1 (y) 6= τ2 (y). From ri=1 ai τi = 0 we get r X i=1 ai τi (yx) = r X i=1 ai τi (y)τi (x) = 0 168 5. Field Theory for all x ∈ E, so Pr i=1 ai τi (y)τi r X i=1 = 0. This yields another nontrivial relation ai τi (y)τi − τ1 (y) r X a i τi = i=1 r X i=2 ai (τi (y) − τ1 (y))τi = 0, which is shorter than the relation we started with, a contradiction. Theorem 5.6.9. [Hilbert Theorem 90] 1. The norm map Nkn /k from kn× to k × is surjective with the kernel consisting of x/σ(x), x ∈ kn× . 2. The trace map Trkn /k from kn to k is surjective with the kernel consisting of x − σ(x), x ∈ kn . Q Q Proof. (1) Since Nkn /k (σ(x)) = ni=1 σ i+1 (x) = ni=1 σ i (x) = Nkn /k (x), so x/σ(x) lies in the kernel of the norm map for all x ∈ kn× . Further, x/σ(x) = y/σ(y) if and only if xy −1 ∈ k × , hence the elements x/σ(x) with x ∈ kn× form a subgroup of kn× of order (q n − 1)/(q − 1). Thus, it is equal to the whole kernel if and only if the norm map is surjective. To see Nkn /k is onto, observe that Nkn /k (x) = n Y i=1 2 σ i (x) = x · xq · xq · · · · · xq n−1 = x1+q+q 2 +···+q n−1 = x(q n −1)/(q−1) for all x ∈ kn× . Hence, any generator x of kn× has Nkn /k (x) of order q − 1. (2) Since elements in Gal(kn /k) are k-linear maps, the image P of Trkn /k (kn ) is a vector space over k, hence Trkn /k (kn ) = 0 or k. If Trkn /k = 0, then ni=1 σi = 0, which is a nontrivial linear relation among elements of Gal(kn /k), so impossible by Lemma 5.6.8. Therefore, Trkn /k is surjective. Then its kernel has order q n−1 . Clearly, Trkn /k (σ(x)) = Trkn /k (x) so that kernel contains x − σ(x) for all x ∈ kn . Further, x − σ(x) = y − σ(y) if and only if x − y ∈ k, so the group {x − σ(x) : x ∈ kn } has order q n /q, thus is equal to the kernel. Remark. The Hilbert Theorem 90 for norm and trace maps is usually proved using first cohomology group of the Galois group (à la Noether). When the base field is finite, we may use counting argument, as shown above. Given z ∈ kn , it defines a k-linear transformation Lz on kn by x 7→ zx, that is, multiplication by z. The trace and determinant of Lz are defined as the trace and determinant of any n × n matrix representing Lz . They are in fact given by Trkn /k and Nkn /k of z. More precisely, we have Theorem 5.6.10. Let z ∈ kn and define Lz as above. Then 1. Tr Lz = Trkn /k (z) and det Lz = Nkn /k (z). 2. Suppose k(z) = kn . Let f (t) = tn + a1 tn−1 + · · · + an−1 t + an be the minimal polynomial of z over k. Then a1 = −Trkn /k (z) and an = (−1)n Nkn /k (z). Proof. We shall prove (1) and (2) under the assumption (2) and leave (1) for the case k(z) being a proper subfield kn as an exercise. For each τ ∈ Gal(kn /k), 0 = τ (f (z)) = f (τ (z)), hence τ (z) is also a root of f (x). Further, if τ and τ ′ are two different elements in Gal(kn /k), then τ (z) 6= τ ′ (z) (otherwise they would agree on k(z) = kn ). This shows that z has n distinct images under Gal(kn /k) and they are the roots of f (t). Therefore, −a1 = the sum of roots of f (t) = Trkn /k (z) and (−1)n an = the product of roots of f (t) = Nkn /k (z). 169 5.6. Finite Fields This proves (2). For (1), we know that Lz satisfies f (t) = 0. As f (t) is irreducible over k and [kn : k] = n, f (t) is the characteristic polynomial of Lz . The companion matrix attached to Lz is 0 −an 1 0 −an−1 1 0 −an−2 . . .. , .. .. 0 . 1 −a1 which has trace = −a1 and determinant = (−1)n an . This proves (1). Exercises 5.6. 1. Let k6 = F56 be the field with 15625 elements and let k = F5 be its prime subfield. (a) Determine the cardinality of the set of elements of k6 which generate k6 as a field over k. (b) Draw a lattice diagram for the subgroup-intermediate subfield correspondence for the fundamental theorem of Galois theory of k6 /k. 2. Let k be a finite field with finite extensions km and kmn of degrees m and mn, respectively. Show that Trkmn /k = Trkm /k ◦ Trkmn /km and Nkmn /k = Nkm /k ◦ Nkmn /km . 3. Let z ∈ kn . Suppose k(z) = km is a proper subfield of kn . Prove that Tr Lz = Trkn /k (z) = (n/m)Trkm /k (z) and det Lz = Nkm /k (z)n/m . 4. (a) (Normal Basis Theorem) There exists an element z ∈ kn such that the set {τ (z) : τ ∈ Gal(kn /k)} is a basis of kn over k. [Hint: Consider the minimal polynomial of the Frobenius’ automorphism σ.] (b) For z in (a), we have Trkn /k (z) 6= 0. [Hint: Express an element in kn as a k-linear combination of {τ (z)}. Then show Trkn /k (kn ) = kTrkn /k (z).] Project 24 (Primitive elements in a finite field). The polynomial p(x) = x2 − 2 is irreducible in Z5 [x]. Then Z[x]/(x2 − 2) is a field with 25 elements and we denote it by F25 . We shall investigate a way to find a primitive element for F25 in this project. Let α = x + (p(x)). (a) Prove that the order of α is 8 and the order of α + 1 is 12. (b) Use (a) to obtain a primitive element (the element of order 24) in F25 . (Hint. Theorem 1.4.12 is useful.) (c) Find a primitive element for the fields Z2 [x]/(x4 + x + 1), Z3 [x]/(x3 + 2x + 1) and Z[x]/(x2 − 2). (d) Write an algorithm to obtain a primitive element for finite fields. Project 25 (Paley Graph). Let q be an odd prime power and consider the finite field Fq of q elements. For the additive group (Fq , +), let P (q) be the Cayley graph Cay(Fq , (F∗q )2 ), called the Paley graph (named after Raymond Paley). Here, (F∗q )2 = {x2 : x ∈ F∗q } (a) Prove that −1 is in (F∗q )2 if and only if q ≡ 1 (mod 4). (Hint. Use Theorem 5.6.1.) (b) Deduce that the Paley graph is undirected if and only if q ≡ 1 (mod 4). (c) Assume that q ≡ 1 (mod 4). Then the Paley graph P (q) is a regular graph. What is its degree? Prove also that if x and y are adjacent the number of common neighbors of x and y is (q − 5)/4 and if x and y are adjacent the number of common neighbors of x and y is (q − 1)/4. (d) Can we generalize the definition of the Paley graph to a finite local ring and a finite commutative ring? (Hint. See Example 3.5.11 and Meemark and Suntornpoch [35].) 170 5.7 5. Field Theory Cyclotomic Extensions In this section, we shall study other important examples of Galois extension, called cyclotomic fields, and compute their Galois groups. Note that “Cyclotomy” is Greek for the art of dividing a circle into equal parts. Theorem 5.7.1. Let K be a field of characteristic 0 and let E be a splitting field of xn − 1 over K. Then Gal(E/K) is isomorphic to a subgroup of Aut Z/(n) ∼ = (Z/(n))× . In particular, Gal(E/K) is abelian. Proof. Since (xn − 1)′ = nxn−1 6= 0, the roots of xn − 1 (in E) are distinct, say xn − 1 = (x − 1)(x − α2 ) . . . (x − αn ). Then A = {z ∈ E : z n = 1} = {1, α2 , . . . , αn } is a finite subgroup of E × , so it is cyclic of order n by Theorem 5.3.7. Any automorphism of E, θ : E → E induces an automorphism θ : A → A, so there is a group homomorphism from Gal(E/K) to Aut A defined by θ 7→ θ|A . This homomorphism is 1-1 since any automorphism of E/K is completely determined by its action on the roots of xn − 1. Hence, Gal(E/K) is isomorphic to a subgroup of Aut A = Aut Z/(n). We call a Galois extension field E/F abelian [cyclic] over F if Gal(E/F ) is abelian [cyclic]. Hence, the above theorem provides an example of abelian extension. Our next objective is to show that if E is a splitting field of xn − 1 over Q, then Gal(E/Q) ∼ = Aut Z/(n) ∼ = (Z/(n))× . We first recall some properties of the cyclic group of order n. Let Z/(n) = hai. Then 1. For each divisor d of n, Z/(n) has a unique subgroup of order d, generated by an/d . 2. All subgroups of Z/(n) are as in (1). Thus, the number of subgroups of Z/(n) is equal to the number of divisors of n. 3. If x, y ∈ Z/(n), then hxi = hyi ⇐⇒ o(x) = o(y) ⇐⇒ θ(x) = y for some θ ∈ Aut Z/(n) ⇐⇒ x and y lie in the same orbit under the action of Aut Z/(n). An element ω in a field K is an nth root of unity if ω n = 1, it is a primitive nth root of unity if o(ω) = n in K × , that is, ω n = 1 and ω m 6= 1 if 1 ≤ m < n. In the complex numbers C, the nth roots of unity are the powers of ω = e2πi/n = cos(2π/n) + i sin(2π/n) and ω t = e2πit/n = cos(2πt/n) + i sin(2πt/n). Thus, Q[ω] is the splitting field of xn − 1 over Q, so [Q[ω] : Q] is the degree of the minimal polynomial of ω over Q. We know that the set U of the nth roots of unity is a cyclic group of order n under multiplication. Hence, the number of primitive nth roots of 1, that is, the number of generators of U , is φ(n). For a positive integer d and x an indeterminate, the dth cyclotomic polynomial, Φd (x) is the product Y Φd (x) = {(x − ε) : ε is a primitive dth root of unity}. If η ∈ Gal(Q[ω]/Q) and z is primitive nth root of unity, then η(z) is primitive. Hence, η(Φn (x)) = Φn (x) and so Φn (x) ∈ Q[x]. It is clear that Φn (x) | (xn − 1) and, in fact, since any nth root of unity has an order d | n we see that Y Φd (x). (5.7.1) xn − 1 = d|n 171 5.7. Cyclotomic Extensions Remark. The formula (5.7.1) provides us with an algorithm for calculating the polynomial Φn (x). To begin with we have Φ1 (x) = x − 1 and assuming we already know the Φd (x) for proper divisors d of n then (5.7.1) gives us Φn (x). For example, for a prime p, Φ1 (x)Φp (x) = xp − 1, so we get Φp (x) = xp−1 + xp−2 + · · · + x + 1. Then Φ2 (x) = x + 1 and Φ3 (x) = x2 + x + 1, so x4 − 1 = x2 + 1 Φ1 (x)Φ2 (x) x6 − 1 Φ6 (x) = = x2 − x + 1 Φ1 (x)Φ2 (x)Φ3 (x) x12 − 1 = x4 − x2 + 1. Φ12 (x) = Φ1 (x)Φ2 (x)Φ3 (x)Φ4 (x)Φ6 (x) Φ4 (x) = Next, we observe that Φn (x) has integer coefficients. This holds for Qn = 1 and assuming it holds for every Φd (x), d < n, we have xn −1 = Φn (x)g(x) where g(x) = d|n;d<n Φd (x) is a monic polynomial with integer coefficients. The division algorithm gives integral polynomials q(x) and r(x) with deg r(x) < deg g(x) such that xn − 1 = q(x)g(x) + r(x). Since q(x) and r(x) are unique in Z[x] and xn − 1 = Φn (x)g(x) in Q[x], we see that Φn (x) = q(x) ∈ Z[x]. We shall now prove Theorem 5.7.2. The nth cyclotomic polynomial Φn (x) has integer coefficients and is an irreducible polynomial in Q[x]. Proof. Suppose that Φn (x) = h(x)k(x), where h(x), k(x) ∈ Z[x] and h(x) is irreducible in Z[x], hence, in Q[x] (Gauss’ lemma). We may also assume that h(x) and k(x) are monic and so deg h(x) ≥ 1. Let p be a prime integer not dividing n and let δ be a root of h(x). Since (p, n) = 1, δ p is a primitive nth root of unity. Assume that δ p is not a root of h(x). Then δ p is a root of k(x); consequently δ is a root of k(xp ). Since h(x) is irreducible and has δ as a root also, (h(x), k(xp )) 6= 1 and thus h(x) | k(xp ). It follows (as mentioned earlier) that k(xp ) = h(x)l(x), where l(x) is monic with integral coefficients. Since xn − 1 = Φn (x)g(x), we have xn − 1 = h(x)k(x)g(x). We now pass to congruences modulo p or, which is the same thing, to equations in (Z/(p))[x]. This gives xn − 1̄ = h̄(x)k̄(x)ḡ(x) (5.7.2) where, in general, if f (x) = a0 xm + a1 xm−1 + · · · + am ∈ Z[x], then f¯(x) = ā0 xm + ā1 xm−1 + · · · + ām , āi = ai + (p) in Z/(p). Similarly, we have k̄(xp ) = h̄(x)¯l(x). Now, using āp = ā for any a ∈ Z, we see that f¯(x)p = (ā0 xm + ā1 xm−1 + · · · + ām )p = āp0 xpm + āp1 xp(m−1) + · · · + āpm = ā0 xpm + ā1 xp(m−1) + · · · + ām = f¯(xp ) for any f (x) ∈ Z[x]. Thus, k̄(x)p = k̄(xp ) = h̄(x)¯l(x) which implies that (h̄(x), k̄(x)) 6= 1. Then (5.7.2) shows that xn − 1̄ has multiple roots in its splitting field over Z/(p). Since the derivative (xn − 1̄)′ = n̄xn−1 and n̄ 6= 0, we have (xn − 1̄, (xn − 1̄)′ ) = 1̄, contrary to the derivative criterion for multiple roots. This contradiction shows that δ p is a root of h(x) for every prime p 6 |n. A repetition of this shows that δ r is a root of h(x) for every integer r prime to n. Since every primitive nth root of 1 has the form δ r , (r, n) = 1, we see that h(x) is divisible by every x − δ ′ , δ ′ primitive. Hence, h(x) = Φn (x) and Φn (x) is irreducible in Q[x]. 172 5. Field Theory As an immediate consequence of Theorem 5.7.2, we get Theorem 5.7.3. Let ω be a primitive nth root of unity. Then 1. Φn (x) is the minimal polynomial of ω over Q. 2. [Q[ω] : Q] = deg Φn (x) = φ(n), the Euler’s φ-function. 3. Q[ω] is the splitting field of Φn (x) over Q. 4. Gal(Q[ω]/Q) ∼ = (Z/(n))× . Proof. (1), (2) and (3) are obvious. To prove (4), recall that by Theorem 5.7.1, Gal(Q[ω]/Q) is isomorphic to a subgroup of (Z/(n))× . Since [Q[ω] : Q] = φ(n) = |(Z/(n))× |, it must be isomorphic to all of (Z/(n))× . Theorem 5.7.3 implies that Gal(Q[ω]/Q) is isomorphic to the multiplicative group Un of units of the ring Z/(n). If n is a prime then we know that this is a cyclic group of order p−1. Moreover, if n = pe11 pe22 . . . pess , pi distinct primes, then Un is isomorphic to the direct product of the groups Upei . In addition, we know the structures of Upe from the knowledge of primitive roots in number theory as follows. Theorem 5.7.4. [Structure of Upe ] 1. U2 and U4 are cyclic and if e > 3, then U2e is a direct product of a cyclic group of order 2 and one of order 2e−2 . 2. If p is an odd prime, the multiplicative group Upe of units of Z/(pe ) is cyclic. Example 5.7.1. If ω = e2πi/72 is a primitive 72nd root of unity, then Gal(Q[ω]/Q) ∼ = U72 ∼ = Z/(2) × Z/(2) × Z/(6). A finite-dimensional field extension of Q is called a cyclotomic field if it is a subfield of Q[ω] for some root of unity ω. Theorem 5.7.5. Let K be a cyclotomic field. Then K is Galois over Q and Gal(K/Q) is abelian. Proof. Consider Q ⊂ K ⊂ Q[ω] for some nth root of unity ω. By the fundamental theorem of Galois theory K = Q[ω]H for some subgroup H of G = Gal(Q[ω]/Q) ∼ = (Z/(n))× . Since G is abelian, H is normal in G, so the fundamental theorem says that K is Galois over Q with Galois group G/H, an abelian group. Remark. A deep theorem of Kronecker and Weber says that the converse of Theorem 5.7.5 is true, namely, “if K is Galois over Q and Gal(K/Q) is abelian, then K is a cyclotomic field, that is, K ⊂ Q[ω] for some root of unity ω.” Example 5.7.2. Let ω = e2πi/71 be a primitive 71st root of unity. Then G = Gal(Q[ω]/Q) ∼ = Z/(2) × Z/(5) × Z/(7). = Z/(70) ∼ = U71 ∼ Let H = Z/(2) × Z/(5) be the subgroup of G of order 10. Then H is normal in G and consequently we have Q[ω]H is a Galois extension over Q of degree [Q[ω]H : Q] = [G : H] = 7 and Gal(Q[ω]H /Q) ∼ = Z/(7). = G/H ∼ We now have enough tools to find the Galois groups of splitting fields of irreducible separable polynomials xn − a. Note that (xn − a)′ = nxn−1 , so xn − a is separable over a field F if and only if char F 6 |n. In particular, if F contains a primitive nth root of unity, then char F 6 |n. 173 5.7. Cyclotomic Extensions Theorem 5.7.6. Let F be a field which contains a primitive nth root of unity ω, i.e., char F not divide n. Let a ∈ F , f (x) = xn − a, E the splitting field for E over F and r a root of f (x) in E. Then (1) The factorization of f (x) in E[x] is xn − a = (x − r)(x − ωr) . . . (x − ω n−1 r) and E = F [r]. (2) Let d be the least positive integer such that rd = b ∈ F . Then d divides n and xd − b = (x − r)(x − εr) . . . (x − εd−1 r) is the minimal polynomial of r over F where ε = ω n/d , a primitive dth root of unity. In addition, [E : F ] = d and Gal(E/F ) ∼ = Z/(d). The automorphism α : E → E defined by α(r) = εr generates Gal(E/F ). Proof. (1) Since r, ωr, . . . , ω n−1 r are all roots of xn − a, (x − r)(x − ωr) . . . (x − ω n−1 r) must divide xn − a. Since both polynomials are monic of degree n, they must be equal. Also, ω ∈ F by hypothesis, so F [r] contains all the roots of xn − a and is generated over F by them. Hence, E = F [r] by the definition of splitting field. (2) Since d is the generator of the group {m ∈ Z : rm ∈ F } and n is in this group, d divides n. Certainly, r is a root of xd − b ∈ F [x]. If xd − b had a proper factor of degree c, 0 < c < d, looking at its constant term would show that rc ∈ F , contradicting the minimality of d. Thus, xd − b is irreducible. Hence, [E : F ] = [F [r] : F ] = d, so |Gal(E/F )| = d. On the other hand, one sees that αi (r) = εi r, so α is an element of Gal(E/F ) of order d. Therefore, Gal(E/F ) = hαi ∼ = Z/(d). For the sake of clarity, we reformulate Theorem 5.7.6 slightly to emphasize the case where f (x) is irreducible, which is the important one. Theorem 5.7.7. Let F be a field which contains a primitive nth root of unity ω and let a ∈ F . Then xn − a is irreducible if and only if no divisor d of n, d 6= 1, such that a = bd for some b ∈ F . If xn − a is irreducible and E/F is its splitting field, then [E : F ] = n and Gal(E/F ) ∼ = Z/(n). Example 5.7.3. Let f (x) = xn − p ∈ Q[x] where p is prime. (The essential point is not that p is prime, but that it is not a proper power.) By Eisenstein’s criterion f (x) is irreducible over Q. If √ we let r = n p denote the positive real nth root of p and ω = e2πi/n , a primitive nth root of unity, then the factorization of f (x) in C[x] is xn − p = (x − r)(x − ωr) . . . (x − ω n−1 r). Now let E = Q[r, ωr, . . . , ω n−1 r] be a splitting field for f (x), and let ϕ ∈ Gal(E/Q). Then ϕ permutes {r, ωr, . . . , wn−1 r} and ϕ is completely defined by its action on the set {r, ωr, . . . , ω n−1 r}. This gives rise to an embedding Gal(E/Q) ֒→ Sn = Sym{r, ωr, . . . , ω n−1 r}. Note that ω = (ωr)r−1 , so ω ∈ E. This makes it clear that E = Q[r, ωr, . . . , ω n−1 r] = Q[ω, r] = Q[ω][r]. Thus, E is generated over Q by two elements ω and r. We also know that E can be generated over Q by a primitive element. However, using such an element would not simplify the description of Gal(E/Q). 174 5. Field Theory Now consider ϕ ∈ Gal(E/Q). Then ϕ(ω) = ω i and ϕ(r) = ω j r for some 1 ≤ i ≤ n − 1 such that gcd(i, n) = 1 and 0 ≤ j ≤ n − 1. The choice of i and j completely determines ϕ and it turns out that all of the above choices do determine automorphisms of E. Thus, |Gal(E/Q)| = n · φ(n). To describe Gal(E/Q) more precisely, let Q[ω] = E H , and for 0 ≤ j ≤ n − 1, let Q[ω j r] = E Kj . Since Q[ω] is Galois over Q, H is normal in Gal(E/Q). Moreover, by Theorem 5.7.7, H = Gal(E/Q[ω]) = hτ i ∼ = Z/(n) is cyclic of order n with generator τ defined by τ (ω) = ω and τ (r) = ωr. The group Kj are more difficult to describe explicitly, but they are all conjugate in Gal(E/Q) and isomorphic as abstract groups to Gal(Q[ω]/Q) ∼ = (Z/(n))× . We have the following diagram of subgroups of Gal(E/Q) which does not include all subgroups. Gal(E/Q) ❡❡❡❣❡❣ ❡❣❡❣❡❣❡❣❡❣❡❣❣❣ rrrr ❡ ❡ ❡ n ❡ ❡ ❣ r ❡❡ ❣❣ ❡❡❡❣❡❣❡❣❡❣❣❣❣❣ rrr ❡ ❡ r ❡ ❡ ❡ r ❡❡❡❡❡ ❣❣❣❣❣ K0 ❨❡❨❨❨❨❨K Kn−1 ▼ ❨❨1❨❨❲❨❲❨❲❲❲❲.❲.❲.❲ ▼▼▼ ❨❨❨❨❨❨❲❲❲❲❲ ▼ ❨❨❨❨❨❨❲❲❲❲❲ ❨ ❲ ❨❨❨❨❨❲❨❲❲❲ ▼▼▼▼ φ(n) ❨❨❨❲❨❲❨❲❨❲❲▼▼ ❨❨❲ ■■ ■■ φ(n) ■■ ■■ ■■ 1 tt tt t tt tt n tt H The corresponding invariant fields are ❣❡❣❡❣ Q ❇❇ ❡❡❣❡❡ ❡❡❡❣❡❣❡❣❡❣❣ ✈✈✈✈ ❇❇ φ(n) ❇❇ ✈ ✈ ✈ ❇❇ ✈✈ ❡❣❡❣❡❣❣❣ n❡❡❡❡❡❡❣ ❣❣ ❡❡❡ ❡❡❡❡❣❡❣❡❣❣❣❣❣❣❣ ❡ ❡ ❡ ❡ ❡ ❣❣❣ ❡❡❡❡❡ Q[r] ❨❡❨❨❨❨❨Q[ωr] Q[ω n−1 r] ❨❨❨❨❨❨❲❲❲❲❲❲.❲.❲. ❍❍ ❨❨❨❨❨❨ ❲❲❲❲ ❍❍ ❨❨❨❨❨❨❲❲❲❲❲ ❍❍ ❨❨❨❨❨❲❨❲❲❲ ❍ ❲ ❨ φ(n) ❨❨❨❲❨❲❨❲❨❲❲❍❍❍ ❨❨❨❲❨❲ Q[ω] E ④④ ④④ ④ ④ n ④④ As a group, Gal(E/Q) is a semi-direct product H ⋊ Ki for any i. We conclude this section with the statement of the following theorem on the Galois group of splitting fields of irreducible separable polynomials xn − a without proof. Theorem 5.7.8. Let F [ω] be a splitting field for xn − 1 over F where ω is a primitive nth root of unity. Suppose that a ∈ F and f (x) = xn − a is irreducible over F and let E be a splitting field for f (x) over F . Let d be the largest divisor of n such that bd = a for some b ∈ F [ω] (possibly d = 1). Let G = Gal(E/F ) and H = Gal(E/F [ω]). Then H is cyclic of order d and normal in G, Gal(F [ω]/F ) ∼ = G/H is isomorphic to a subgroup of (Z/(n))× and G is isomorphic to a semi-direct product of H by G/H. Using the cyclotomic polynomials, we now present the proof of Wedderburn’s theorem as follows. Theorem 5.7.9. [Wedderburn, 1909] A finite division ring is a field. 175 5.7. Cyclotomic Extensions Proof. Let D be a finite division ring. Then the center of D, denoted by F , is a finite field (see Exercises 2.1). Assume that |F | = q. Since D is a vector space over F , |D| = q n for some n ∈ N. Also, for an element d ∈ D, the set C(d) = {r ∈ D : rd = dr} is a division ring containing F and |C(d)| = q m for some m ≤ n, which is strictly less than if d ∈ / F . Thus, the class equation (Corollary 1.3.8) for the multiplicative group D r {0} is q n − 1 = |F r {0}| + s X i=1 [D r {0} : C(di ) r {0}] = q − 1 + s X qn − 1 , q mi − 1 i=1 where d1 , d2 , . . . , ds represent the conjugacy classes of D r {0} which contains more than one element and |C(di )| = q mi for some mi < n for all i. Because each (q n − 1)/(q mi − 1) = [D r {0} : C(di ) r {0}] is an integer, mi is a proper divisor of n. Thus, the quotient xn − 1 Φn (x)(xmi − 1) is a polynomial in Z[x]. Substitute q for x, we see that Φn (q) divides (q n − 1)/(q mi − 1). It follows from the class equation that Φn (q) divides q − 1 because it divides all the other terms. Then |Φn (q)| ≤ q − 1. On the other hand, since 1 is the closest point, on the unit circle {z ∈ C : |z| = 1}, to the positive integer q, we have that for every primitive nth root of unity ω j , |q − ω j | ≥ q − 1 ≥ 1, and the first inequality is strict unless ω j = 1, that is, unless 1 is a primitive nth root of unity which means n = 1. So the product |Φn (q)| of the |q − ω j |’s is greater than or equal to q − 1, with equality only if n = 1. Because |Φn (q)| is both at most q − 1 and at least q − 1, we get |Φn (q)| = q − 1 and hence n = 1. Therefore, |D| = q = |C(D)|, so D = C(D) which implies D is commutative as desired. Given a field F and a polynomial p(x) ∈ F [x], we say that p(x) is solvable by radicals over F if we can find a finite sequence of fields F1 = F (ω1 ), F2 = F1 (ω2 ), . . . , Fk = Fk−1 (ωk ) such that ω1r1 ∈ F , ω2r2 ∈ F1 , . . . , ωkrk ∈ Fk−1 and all roots of p(x) lie in Fk . If K is the splitting field of p(x) over F , then p(x) is solvable by radicals over F if we can find a finite sequence of fields as above such that K ⊆ Fk . An important remark, and one we shall use later, in the proof of Theorem 5.7.10, is that if such an Fk can be found, we can, without loss of generality, assume it to be a normal extension of F . We leave its proof as an exercise. Theorem 5.7.10. [Galois] Let F be a field which contains a primitive nth root of unity for every positive integer n. If a polynomial p(x) ∈ F [x] is solvable by radical over F , then the Galois group over F of p(x) is solvable. Proof. Let K be the splitting field of p(x) over F . Since p(x) is solvable by radicals, there exists a finite sequence of fields F = F0 ⊂ F1 = F (ω1 ) ⊂ F2 = F1 (ω2 ) ⊂ . . . ⊂ Fk = Fk−1 (ωk ), where ω1r1 ∈ F , ω2r2 ∈ F1 , . . . , ωkrk ∈ Fk−1 and K ⊆ Fk such that Fk is normal over F . As a normal extension of F , Fk is also a normal of any intermediate fields, hence Fk is a normal extension of each Fi . Theorem 5.7.6 implies that Fi is a normal extension of Fi−1 and Gal(Fi /Fi−1 ) is abelian for all i. Thus, by the Galois correspondence, Gal(Fk /Fi ) is a normal subgroup in Gal(Fk /Fi−1 ). Consider the normal series Gal(Fk /F0 ) ⊃ Gal(Fk /F1 ) ⊃ Gal(Fk /F2 ) ⊃ . . . ⊃ Gal(Fk /Fk−1 ) ⊃ {1}. Since Gal(Fk /Fi−1 )/Gal(Fk /Fi ) ∼ = Gal(Fi /Fi−1 ) is abelian for all i, Gal(Fk /F ) is solvable. It ∼ follows that Gal(K/F ) = Gal(Fk /F )/Gal(Fk /K) is solvable by Theorem 3.2.4 (2). 176 5. Field Theory We make two remarks without proof. 1. The converse of Theorem 5.7.10 is also true; that is, if the Galois group of p(x) over F is solvable, then p(x) is solvable by radicals over F . 2. Theorem 5.7.10 and its converse are true even if F does not contain roots of unity. Recall that for n ≥ 5, Sn is not solvable. Thus we have Corollary 5.7.11. The general polynomial of degree n ≥ 5 over Q is not solvable by radical. k−1 Exercises 5.7. 1. Prove the following statements. (a) If p is a prime number, then Φpk (x) = Φp (xp ). (b) If n > 1 is odd, then Φ2n (x) = Φn (−x). Φn (xp ) , if p 6 |n, (c) If p is a prime number, then Φpn (x) = Φn (x) p Φn (x ), if p|n. 2πi/18 2. Let ω = e be a primitive 18th root of unity. (a) Find the minimal polynomial of ω over Q. (b) Draw a lattice diagram for the subgroup-intermediate subfield correspondence for the fundamental theorem of Galois theory of Q[ω]/Q. 3. Give an example of field E containing the field of rational numbers Q such that E is Galois over Q and Gal(E/Q) is a cyclic group of order five. 4. Let K be a finite separable extension over F and E its normal closure (smallest normal extension over F containing K). (a) Prove that [E : F ] is finite. (b) If Gal(E/F ) is abelian, show that K is normal over F . 5. If p(x) is solvable by radicals over F , prove that we can find a finite sequence of fields F ⊂ F1 = F (ω1 ) ⊂ F2 = F1 (ω2 ) ⊂ . . . ⊂ Fk = Fk−1 (ωk ), where ω1r1 ∈ F , ω2r2 ∈ F1 , . . . , ωkrk ∈ Fk−1 containing all the roots of p(x) such that Fk is normal over F . 6. Assume that xp − a, a ∈ Q, is irreducible in Q[x]. Show that the Galois group of xp − a over Q is isomorphic to the group of transformations of Z/(p) of the form y 7→ ky + l where k, l ∈ Z/(p) and k 6= 0. Project 26 (Insolvability of a quintic). Consider g(x) = 3x5 − 15x + 5. By Eisenstein’s criterion, g(x) is irreducible over Q. (a) Use the intermediate value theorem to show that g(x) has a real root between −2 and −1 and also has a real root between 0 and 1 and between 1 and 2. (b) Use Rolle’s theorem to assure that there is no other real roots. Hence, the other two roots of g(x) are non real complex numbers, say a + bi and a − bi. (c) Let K be the splitting field of g(x) in C. Show that Gal(K/Q) is isomorphic to S5 . (d) Since S5 is not solvable, deduce that g(x) is not solvable by radical by Galois (Theorem 5.7.10). (e) Give another example of irreducible polynomial in Z[x] of degree five that is solvable by radical, compute the Galois group of its splitting field over Q and show that this group is solvable. 5.8 Normal Bases Let E be an extension field of a field F . We have known from Lemma 5.6.8 that the automorphism in Gal(E/F ) are E-linearly independent F -linear transformations. 177 5.8. Normal Bases Theorem 5.8.1. If E/F is a finite Galois extension with Galois group G = {1, σ2 , . . . , σn }. Then {u1 , u2 , . . . , un } is a basis for E/F if and only if u1 u2 ... un σ2 (u1 ) σ2 (u2 ) . . . σ2 (un ) det . 6= 0. . . σn (u1 ) σn (u2 ) . . . σn (un ) Proof. Call the above matrix M and suppose that det M = 0. Since M ∈ Mn (E), there are α1 , α2 , . . . , αn ∈ E, not all zero, such that α1 α2 . . . αn M = ~0. This translates to θ(u1 ) = θ(u2 ) = · · · = θ(un ) = 0 where θ = α1 1 + α2 σ2 + · · · + αn σn : E → E. But θ : E → E is a F -linear map, so θ is the zero map, since it vanishes on u1 , u2 , . . . , un . Since 1, σ2 , . . . , σn are linearly independent over K, Lemma 5.6.8 says that θ 6= 0, so we have a contradiction. Conversely, if u1 , u2 , . . . , un are not a basis for E/F , then there are β1 , β2 , . . . , βn ∈ F , not all zero, such that u1 β1 + u2 β2 + · · · + un βn = 0. Then for any σi ∈ G, so M β1 σi (u1 )β1 + σi (u2 )β2 + · · · + σi (un )βn = σi (u1 β1 + u2 β2 + · · · + un βn ) = 0, T β2 . . . βn = ~0. Hence, det M = 0. Note that if |K| = q, then αq − α = 0 for all α ∈ K, so f (x) = xq − x is a nonzero polynomial but it is a zero function. The next theorem says that such a polynomial cannot exist if K is infinite. Theorem 5.8.2. Let F be an infinite field and F ⊆ E. If f (x1 , . . . , xn ) is a nonzero polynomial in E[x1 , . . . , xn ], then there exist α1 , . . . , αn ∈ F such that f (α1 , . . . , αn ) 6= 0. Proof. We shall use induction on n. For n = 1, since f (x1 ) has only finitely many roots and F is infinite, there is αi ∈ F such that f (α1 ) 6= 0. Assume that the statement holds for n, and let f (x1 , . . . , xn+1 ) = f0 (x1 , . . . , xn ) + f1 (x1 , . . . , xn )xn+1 + · · · + ft (x1 , . . . , xn )xtn+1 . Since f 6= 0, at least one of f0 (x1 , . . . , xn ), . . . , ft (x1 , . . . , xn ) is nonzero, so there are α1 , . . . , αn ∈ F such that f (α1 , . . . , αn , xn+1 ) 6= 0 in E[xn+1 ]. By the one variable case, there is αn+1 ∈ F such that f (α1 , . . . , αn , αn+1 ) 6= 0. Theorem 5.8.3. Let F be an infinite field and E/F Galois with Galois group G = Gal(E/F ) = {1, σ2 , . . . , σn }. Suppose that 0 6= f (x1 , . . . , xn ) ∈ F [x1 , . . . , xn ] where x1 , . . . , xn are indeterminates over F . Then there exists u ∈ E such that f (u, σ2 (u), . . . , σn (u)) 6= 0. Proof. Let {u1 , . . . , un } be a basis for E/F . By Theorem 5.8.1, the matrix u1 u2 ... un σ2 (u1 ) σ2 (u2 ) . . . σ2 (un ) M = . ∈ Mn (E) .. σn (u1 ) σn (u2 ) . . . σn (un ) 178 5. Field Theory is invertible. This means that the map on E[x1 , . . . , xn ] defined by g(x1 , . . . , xn ) 7→ g(u1 x1 + · · · + un xn , . . . , σn (u1 )x1 + · · · + σn (un )xn ) is an isomorphism. Thus, h(x1 , . . . , xn ) = f (u1 x1 + · · · + un xn , . . . , σn (u1 )x1 + · · · + σn (un )xn ) is a nonzero polynomial in E[x1 , . . . , xn ]. By Theorem 5.8.2, there are a1 , . . . , an in F such that h(a1 , . . . , an ) 6= 0. Let u = u1 a1 + · · · + un an , this translates to 0 6= h(a1 , . . . , an ) = f (u1 a1 + · · · + un an , . . . , σn (u1 )a1 + · · · + σn (un )an ) = f (u, σ2 (u), . . . , σn (u)), since σi (u1 )a1 + · · · + σi (un )an = σi (u1 a1 + · · · + un an ) = σi (u). Consider E = Q[i] is a Galois extension over Q. Its Galois group is of order two and consists of the identity map and the complex conjugation. A basis over Q for it is {1, i}. This basis is not invariant under the Galois action, namely after acting by the complex conjugation, we obtain {1, −i}. We are showing the existence of a basis for a finite Galois extension which forms a single orbit under the action of the Galois group. For example, for Q[i], we may use {1 + i, 1 − i}. In the case of finite fields, this means that each of the basis elements is related to any one of them by applying the Frobenius’ automorphism repeatedly. Let E/F be Galois with Galois group G = Gal(E/F ) = {σ1 , . . . , σn }. A normal basis for E/F is a basis of the form {σ1 (u), . . . , σn (u)} for some u ∈ E. Eisenstein conjectured the existence of a normal basis in 1850 for finite extensions of finite fields and Hensel gave a proof for finite fields in 1888. Dedekind used such bases in number fields in his work on the discriminant in 1880, but he had no general proof. (See the quote by Dedekind on the bottom of page 51 of Curtis’s “Pioneers of Representation Theory: Frobenius, Burnside, Schur, and Brauer”.) In 1932 Noether gave a proof for some infinite fields while Deuring gave a uniform proof for all fields (also in 1932). This basis is frequently used in cryptographic applications that are based on the discrete logarithm problem such as elliptic curve cryptography. Hardware implementations of normal basis arithmetic typically have far less power consumption than other bases. Theorem 5.8.4. [Normal Basis Theorem] Let E/F be a Galois extension with Galois group G = Gal(E/F ) = {σ1 , . . . , σn }. Then E/F has a normal basis. Proof. We shall assume that F is infinite and leave the finite case as an exercise (see Exercise 5.6). Let u ∈ E. By Theorem 5.8.1, {σ1 (u), σ2 (u), . . . , σn (u)} is a basis for E/F if and only if 2 σ1 (u) σ1 σ2 (u) . . . σ1 σn (u) σ2 σ1 (u) σ 2 (u) . . . σ2 σn (u) 2 det 6= 0. .. . σn σ1 (u) σn σ2 (u) . . . σn2 (u) Note that the entries in each row or column of the above matrix, call M , are a permutation of the elements σ1 (u), . . . , σn (u). In other words, each σi (u) occurs exactly once in each row and column of M . Thus, M = σ1 (u)A1 + · · · + σn (u)An where each Ai is a permutation matrix (a matrix with a single entry 1 in each row and column and the remaining entries zero). Since det Ai = ±1, we see by inspection that if x1 , . . . , xn are indeterminates over E f (x1 , . . . , xn ) = det(x1 A1 + · · · + xn An ) = ±xn1 ± · · · ± xnn + other terms 179 5.9. Transcendental Extensions In particular, f (x1 , . . . , xn ) is a nonzero polynomial in E[x]. By Theorem 5.8.3, there is a ū ∈ E such that f (σ1 (ū), . . . , σn (ū)) 6= 0. This translates to 0 6= f (σ1 (ū), . . . , σn (ū)) = det(σ1 (ū)A1 + · · · + σn (ū)An ) = det M. Hence, σ1 (ū), . . . , σn (ū) is a desired normal basis for E/F . √ √ Exercises 5.8. 1. Determine a normal basis for the field Q( 2, 3) over Q by using the Galois group in Example 5.4.4. 2. Determine a normal basis for the cyclotomic field Q(e2πi/p ) over Q where p is a prime number. 5.9 Transcendental Extensions Most of extension fields seen in the previous section are algebraic. In this section, we shall present some results on transcendental extension. The final theorem, namely Lüroth’s theorem, has many applications in algebraic geometry and function field theory. Let F be a subfield of a field E and let x1 , x2 , . . . be independent indeterminates over E. An element z ∈ E is transcendental over F if the homomorphism F [x1 ] → E defined by f (x1 ) 7→ f (z) is one-to-one. We call z ∈ E algebraic over F if it is not transcendental over F . A finite set {z1 , . . . , zn } ⊂ E is algebraically independent over F if the homomorphism F [x1 , . . . , xn ] → E defined by f (x1 , . . . , xn ) 7→ f (z1 , . . . , zn ) is one-to-one. (Note that the empty set is algebraically independent since F ֒→ E is one-to-one.) An arbitrary subset Z of E is algebraically independent over F if all of its finite subsets are algebraically independent. A subset Z of E is algebraically dependent if it is not algebraically independent. Remarks. 1. If z is transcendental over F , then F [z] ∼ = F [x1 ], so F [z] is not a field and F [z] is infinite dimensional over F . 2. If z is algebraic over F , then F [z] ∼ = F [x1 ]/(f (x1 )) where f (x1 ) is the minimal polynomial of z over F . Thus, F [z] = F (z) is a field and F [z] is finite dimensional over F . Example 5.9.1. Let F ⊂ F (y, z) ⊂ E where y and z are independent indeterminates over F . Then {y 2 , z 2 } is an algebraically independent set but {y 2 , yz, z 2 } is not (for, if f (x1 , x2 , x3 ) = x1 x3 − x22 , then f (y 2 , yz, z 2 ) = 0). Let F be a subfield of a field E. E is algebraic over F if each element of E is algebraic over F . E is purely transcendental over F if it is isomorphic (by an isomorphism which is the identity on F ) to F ({xα }) where {xα } is a (possibly infinite) set of independent indeterminates. Theorem 5.9.1. Let F be a subfield of a field E. 1. There exists a subset X of E (possibly X is empty) such that (a) X is algebraically independent over F . (b) X is maximal among algebraically independent sets, in the sense: If X ⊆ Y ⊆ E and X 6= Y , then Y is not algebraically independent. 2. F (X) is purely transcendental over F and E is algebraic over F (X). E algebraic F (X) purely transcendental F 180 5. Field Theory Proof. (1) Let S = {X ⊆ E : X is algebraically independent}. Since the empty set S is algebraically independent, S is nonempty. Let {Xα }α∈Λ be a chain in S . Let {z1 , . . . , zn } ⊆ α∈Λ Xα . Then ∀i, ∃αi ∈ Λ, zi ∈ Xαi . Since {Xα }α∈Λ is a chain, we may rearrange αi so that there exists j ∈ Λ such that zi ∈ Xαj for all i. Since Xαi is algebraically independent, so is {z1 , . . . , zn }. Thus, S α∈Λ Xα is an upper bound of this chain in S . By Zorn’s Lemma, S has a maximal element, say X. Hence, F (X) is purely transcendental over F . The maximality of X implies that E must be algebraic over F . (2) The definition of algebraically independent means that F (X) is purely transcendental over F . Consider z ∈ E. If z ∈ X ⊂ F (X), then z is algebraic over F (X). If z ∈ / X, the set X ∪ {z} is algebraically dependent, so for some n there is a nonzero polynomial f (x1 , . . . , xn , xn+1 ) (x1 , . . . , xn+1 are indeterminates over F ) and a1 , . . . , an ∈ X such that f (a1 , . . . , an , z) = 0. The polynomial f (x1 , . . . , xn , xn+1 ) cannot be a polynomial in only x1 , . . . , xn , since {a1 , . . . , an } is an algebraically independent set. Write f (x1 , . . . , xn , xn+1 ) = f0 (x1 , . . . , xn ) + f (x1 , . . . , xn )xn+1 + · · · + fr (x1 , . . . , xn )xrn+1 . Thus, f (a1 , . . . , an , xn+1 ) ∈ F (X)[xn+1 ] is a nonzero polynomial having z as a root, so z is algebraic over F (X). Hence, E is algebraic over F (X). Remark. There is no uniqueness for the field F (X). For example, if E = F (t) where t is an indeterminate, then we can take X = {p(t)/q(t)} where p(t)/q(t) is any element of E which is not in F . In this case [E : F (p(t)/q(t))] = n where n = max{deg p(t), deg q(t)} (Theorem 5.9.3). However, we shall see shortly that the number of elements in the set X is independent of particular set X. Let F be a subfield of E. A maximal algebraically independent (over F ) subset of E is called a transcendence basis for E/F . Remark. By Theorem 5.9.1, a transcendence basis for E/F exists. It may be empty, which happens precisely when E is algebraic over F . Also, E is purely transcendental over F if it has a transcendence base B such that E = F (B). Theorem 5.9.2. Let F be a subfield of E. Then any two transcendence bases for E/F have the same cardinality. We call the number of elements of transcendence bases of E the transcendence degree of E/F . For example, an algebraic extension has transcendence degree zero; F (x) has transcendence degree one over F ; in general, F ((xα )α∈Λ ) has transcendence degree |Λ| over K. The purely transcendental extension fields E/F , especially those having a finite transcendence degree, appear to be the simplest type of extension fields. It is clear that such a field is isomorphic to the field of fractions F (x1 , . . . , xn ) of the polynomial ring F [x1 , . . . , xn ] in indeterminates x1 , . . . , xn . Even though these fields look quite innocent, there are difficult and unsolved problems particularly on the nature of the subfields of F (x1 , . . . , xn )/F . The one case where the situation is quite simple is that in which E has transcendence degree one. We shall consider this case and close this chapter. Let E = F (t), t transcendental, and let u ∈ E, ∈ / F . We can write u = f (t)/g(t) where f (t), g(t) ∈ F [t] and (f (t), g(t)) = 1. If n is the larger of the degrees of f (t) and g(t), then we can write f (t) = a0 + a1 t + · · · + an tn and g(t) = b0 + b1 t + · · · + bn tn , ai , bi ∈ F , and either an or bn 6= 0. We have f (t) − ug(t) = 0, so (an − ubn )tn + (an−1 − ubn−1 )tn−1 + · · · + (a0 − ub0 ) = 0 (5.9.1) and an − ubn 6= 0 since either an 6= 0 or bn 6= 0 and u ∈ / F . Thus, (5.9.1) shows that t is algebraic over F (u) and [F (t) : F (u)] ≤ n. We prove the following more precise result. 181 5.9. Transcendental Extensions Theorem 5.9.3. Let E = F (t), t transcendental over F , and let u ∈ F (t), ∈ / F . Write u = f (t)/g(t) where (f (t), g(t)) = 1, and let n = max{deg f (t), deg g(t)}. Then u is transcendental over F , t is algebraic over F (u), and [F (t) : F (u)] = n. Moreover, the minimal polynomial of t over F (u) is a multiple in F (u) of f (x, u) = f (x) − ug(x). Proof. Put f (x, y) = f (x) − yg(x) ∈ F [x, y], x, y indeterminates. This polynomial in x and y is of first degree in y and it has no factor h(x) of positive degree since (f (x), g(x)) = 1. Thus, it is irreducible in F [x, y]. Now t is algebraic over F (u) so if u were algebraic over F , then t would be algebraic over F , contrary to the hypothesis. Hence, u is transcendental over F . Then F [x, u] ∼ = F [x, y] under the isomorphism over F fixing x and mapping u into y and hence f (x, u) is irreducible in F [x, u]. It turns out that f (x, u) is irreducible in F (u)[x]. Since f (t, u) = f (t) − ug(t) = 0, it follows that f (x, u) is a multiple in F (u)[x] of the minimal polynomial of t over F (u). Therefore, [F (t) : F (u)] is the degree in x of f (x, u). This degree is n, so the proof is complete. We can determine all of the subfields E/F for E = F (t), t transcendental: These have the form F (u) for some u. This important result is called the Lüroth’s Theorem. Lüroth proved it in case K = C in 1876. It was first proved for general fields K by Steinitz in 1910, by the following argument. Theorem 5.9.4. [Lüroth] If E = F (t), t transcendental over F , then any subfield K of E/F , K 6= F , has the form F (u), u transcendental over F . Proof. Let v ∈ K, ∈ / F . Then we have seen that t is algebraic over F (v). Thus, t is algebraic over K. Let f (x) = xn + k1 xn−1 + · · · + kn be the minimal polynomial of t over K, so the ki ∈ K and n = [F (t) : K]. Since t is not algebraic over F , some kj ∈ / F . We shall show that K = F (u), u = kj . We can write u = g(t)/h(t) where g(t), h(t) ∈ F [t], (g(t), h(t)) = 1 and m = max{deg g(t), deg h(t)} > 0. Then, by Theorem 5.9.3, [E : F (u)] = m. Since K ⊃ F (u) and [E : K] = n, we evidently have m ≥ n and equality holds if and only if K = F (u). Now t is a root of the polynomial g(x) − uh(x) ∈ K[x]. Hence, we have a q(x) ∈ K[x] such that g(x) − uh(x) = q(x)f (x). (5.9.2) The coefficient ki of f (x) is in F (t), so there exists a nonzero polynomial c0 (t) of least degree such that c0 (t)ki = ci (t) ∈ F [t] for 1 ≤ i ≤ n. Then c0 (t)f (x) = f (x, t) = c0 (t)xn + c1 (t)xn−1 + · · · + cn (t) ∈ F [x, t], and f (x, t) is primitive as a polynomial in x, that is, the ci (t) are relatively prime. The x-degree of f (x, t) is n. Since kj = g(t)/h(t) with (g(t), h(t)) = 1, the t-degree of f (x, t) is ≥ m. Now replace u in (5.9.2) by g(t)/h(t) and the coefficients of q(x) by their expressions in t. There exist, therefore, ϕ(t) and q(x, t) ∈ F [x, t] such that ϕ(t)[g(x)h(t) − g(t)h(x)] = f (x, t)q(x, t). Since the coefficients c0 (t), c1 (t), . . . , cn (t) of f (x, t) have no common factor, we know that ϕ(t) divides q(x, t). Hence, we may assume ϕ(t) = 1. It turns out that there exists a polynomial q ′ (x, t) ∈ F [x, t] such that g(x)h(t) − g(t)h(x) = f (x, t)q ′ (x, t). Since the t-degree of the left-hand side is ≤ m and that of f (x, t) is ≥ m, it follows that this degree is m and q ′ (x, t) = q ′ (x) ∈ F [x]. Then the right-hand side is primitive as a polynomial in x and so is the left-hand side. By symmetry the left-hand side is primitive as a polynomial in t also. Hence, q ′ (x) = q ′ ∈ F . Thus, f (x, t) has the same x-degree and t-degree so m = n, which implies that K = F (u). 182 5. Field Theory Exercises 5.9. 1. Prove that there is no intermediate field K with Q ⊆ K $ C with C purely transcendental over K. 2. Prove that a purely transcendental proper extension of a field is never algebraically closed. 3. Let E = F (t, v), where t is transcendental over F and v 2 + t2 = 1. Show that E is purely transcendental over F . Project 27 (More on Lüroth’s theorem). Prove more general fact that if F ⊆ L ⊆ E and E is finitely generated over F (finite transcendence degree), then L is also finitely generated over F . We can ask more generally about minimal numbers of generators of finitely-generated extensions. For instance, suppose K ( L ⊆ K(x1 , . . . , xn ) where the xi are algebraically independent over K. If L/K has transcendence degree 1, then L = K(α), This was proved for K = C by Gordan in 1887, and for arbitrary K by Igusa in 1951. If C ( L ⊆ C(x1 , . . . , xn ) where L/C has transcendence degree 2, then L = C(α, β). This was proved by Castelnuovo in 1894. All known proofs are difficult. The result is not true in general for other types of fields K, such as Q or R. Finally, there are fields L with C ( L ( C(x1 , x2 , x3 ) such that L/C has transcendence degree 3 but cannot be generated by three elements. Bibliography General References [1] D. S. Dummit and R. M. Foote, Abstract Algebra, 2nd edn, Prentice-Hall Inc., London, 1999. [2] J. B. Fraleigh, A First Course in Abstract Algebra, 7th edn, Addison Wesley, New York, 2002. [3] P. Grillet, Algebra, 2nd edn, Springer, New York, 2007. [4] T. W. Hungerford, Algebra, Springer, New York, 1974. [5] I. M. Isaacs, Algebra, a graduate course, Brooks/Cole Publishing Company, Pacific Grove, 1993. [6] N. Jacobson, Basic Algebra I and II, W. H. Freeman & Co, 1996. [7] A. W. Knapp, Basic Algebra, Birkhäuser, Boston, 2006. [8] A. W. Knapp, Advanced Algebra, Birkhäuser, Boston, 2007. [9] S. Lang, Algebra, 3rd edn, Springer, New York, 2002. [10] W. K. Nicholson, Introduction to Abstract Algebra, John Wiley & Sons, Inc., New Jersey, 2007. [11] J. J. Rotman, Advanced Modern Algebra, Prentice Hall, 2002. Technical References [12] M. F. Atiyah and I. G. MacDonald, Introduction to Commutative Algebra, Addison-Wesley Publishing Co, Reading, Mass.-London-Don Mills, Ontario, 1969. [13] D. M. Burton, Elementary Number Theory, 7th edn, McGraw-Hill Higher Education, Dubuque, 2010. [14] C. Godsil and G. Royle, Algebraic Graph Theory, Spinger, New York, 2001. [15] T. Head, Modules: A Primer of Structure Theorems, Brooks/Cole Publishing Company, Monterey, 1974. [16] W.-C. Winnie Li, Number Theory with Applications, World Scientific, Singapore, 1996. [17] B. R. McDonald, Finite Rings with Identity, Marcel Dekker, New York, 1974 [18] M. Reid, Undergraduate Commutative Algebra, Cambridge University Press, New York, 1995. [19] I. Stewart, Galois Theory, 3rd edn, Chapman & Hall/CRC mathematics, Boca Raton, 2004. [20] D. J. Winter, The Structure of Fields, Springer, New York, 1974. Research Articles [21] R. A. Brualdi, Energy of a Graph, http://www.public.iastate.edu/lhogben/energyB.pdf. [22] R. D. Carmichael, Note on a new number theory function, Bull. Amer. Math. Soc. 16 (1910) 232–238. 183 184 BIBLIOGRAPHY [23] Z. Gu and Z. Wan, Orthogonal graphs of odd characteristic and their automorphisms, Finite Fields Appls. 14 (2008) 291–313. [24] Z. Gu, Subconstituents of symplectic graphs modulo pn , Linear Algebra Appl. 439 (2013) 1321–1329. [25] I. Gutman, The Energy of a Graph: Old and New Results, Algebraic Combinatorics and Applications, Springer, Berlin, 2001. [26] I. M. Isaacs and M. R. Pournaki, Generalizations of FermatâĂŹs little theorem using group theory, Amer. Math. Monthly 112 (2005), 734–740. [27] D. Kiani, M.M.H. Aghaei, Y. Meemark and Suntornpoch B., Energy of unitary Cayley graphs and gcd-graphs, Linear Algebra Appl. 435 (2011) 1336–1343. [28] W. Klingenberg, Symplectic groups over local rings, Amer. J. Math. 85 (1963) 232–240. [29] W. Klotz and T. Sander, Some properties of unitary Cayley graphs, The Electronic J. Comb., 14 (2007), #R45. [30] F. Li, K. Wang and J. Guo, More on symplectic graphs modulo pn , Linear Algebra Appl. 438 (2012) 2651–2660. [31] F. Li, K. Wang and J. Guo, Symplectic graphs modulo pq, Discrete Math. 313 (2013), 650–655. [32] Y. Meemark and T. Prinyasart, On symplectic graphs modulo pn , Discrete Math. 311 (2011) 1874– 1878. [33] Y. Meemark and T. Puirod, Symplectic graphs over finite local rings, Europ. J. Combin. 34 (2013) 1114–1124. [34] Y. Meemark and T. Puirod, Symplectic graphs over finite commutative rings, Europ. J. Combin. 41 (2014) 298–307. [35] Y. Meemark and B. Suntornpoch, Eigenvalues and energy of restricted unitary Cayley graphs induced from the square mapping, ScienceAsia 39 No.6 (2013) 649–652. [36] Y. Meemark and N. Wiroonsri, The quadratic digraph on polynomial rings over finite fields, Finite Fields Appl. 16 (2010) 334–346. [37] Y. Meemark and N. Wiroonsri, The digraph of the kth power mapping of the quotient ring of polynomials over finite fields, Finite Fields Appl. 18 (2012) 179–191. [38] J. W. Sander and T. Sander, The energy of integral circulant graphs with prime power order, Appl. Anal. Discrete Math., 5 (2011), pp. 22–36. [39] W. So, Integral circulant graphs, Discrete Math., 306 (2006), 153–158. [40] L. Somer and M. Křížek, On a connection of number theory with graph theory, Czechoslovak Math J., 54 (2004) 465–485. [41] L. Somer and M. Křížek, Structure of digraphs associated with quadratic congruences with composite moduli, Discrete Math., 306 (2006) 2174–2185. [42] Z. Tang and Z. Wan, Symplectic graphs and their automorphisms, Europ. J. Combin. 27 (2006) 38– 50. Index 5-lemma, 106 abelian extension, 170 abelian group, 6 act faithfully, 12 act transitively, 12 adjacency matrix of a graph, 102 algebraic closure, 150 algebraic element, 76, 179 algebraic extension, 76, 179 algebraically closed field, 68, 149 algebraically dependent, 179 algebraically independent, 179 alphabet set, 96 alternating group, 26 arc, 143 arc transitive, 143 Artinian module, 135 Artinian ring, 135 ascending chain condition (a.c.c.), 130 associate, 59 associative, 5 atom, 59 automorphism group, 22 automorphism group of graphs, 143 automorphism of graphs, 143 automorphism of groups, 9 basis, 108 bilinear form, 137 binary operation, 5 binomial theorem, 46 Burnside theorem, 16 butterfly lemma, 82 canonical projection, 18, 54 Cauchy theorem, 16 Cayley digraph, 99 Cayley theorem, 12 center of a p-group, 28 center of a group, 9 center of a ring, 51 central series, 89 centralizer, 9 chain ring, 58 character, 43 characteristic of a ring, 49 characteristic subgroup, 85 chromatic number, 143 coefficient, 66 cokernel, 104 commutative, 5 commutative ring, 45 commutator, 85 composition series, 83 congruence modulo n, 4 conjugacy class, 13 conjugation, 13 connected component, 99 connected digraph, 99 constant term, 66 content, 69 coset, left or right, 14 cycle, 23 cycle structure, 25 cyclic extension, 170 cyclic group, 8 cyclic module, 108 cyclotomic field, 172 cyclotomic polynomial, 71, 170 defining relation, 98 degree of a field extension, 75 degree of a polynomial, 66 degree of a regular graph, 100 degree of an algebraic element, 76 derivative, 152 derivative of a polynomial, 78 derived length, 86 derived series, 85 derived subgroup, 85 descending chain condition (d.c.c.), 135 diagram chasing, 105 dihedral group, 8 direct product of groups, 34 direct sum, 104, 107 directed path, 99 disjoint cycle, 24 divide, 2, 58 divisible module, 116 division algorithm, 1 division algorithm of polynomials, 67 division ring, 47 division ring of real quaternions, 49 dual group, 44 eigenvalues and eigenvectors of a graph G, 102 Eisenstein’s criterion, 71 elementary row/column transformations, 121 elementary symmetric functions, 159 embedded, 54 empty word, 96 endomorphism of groups, 9 energy of a graph, 102 entire, 47 epimorphism of groups, 9 185 INDEX 186 Euclidean algorithm, 5, 65 Euclidean domain, 63 Euler φ-function, 6, 20 even permutation, 26 exact, 19 exact diagram, exact sequence, 104 exponent of a group, 37 external weak direct product of groups, 34 factor group, 18 factor module, 104 factor ring, 54 field, 47 field extension, 75 field of fractions, 55 field of invariant, fixed field, 156 field of rational functions, 158 finitely generated free module, 108 finitely generated module, 108 free group, 97 free module, 108 Frobenius’ automorphism, 153, 167 fundamental theorem of algebra, 165 fundamental theorem of arithmetic, 3 fundamental theorem of Galois theory, 161 Galois extension, 157 Galois group, 156 Galois ring, 79 Gauss’ lemma, 72 gcd-graph, 102 general linear group, 6, 47, 92, 120 generator, 8, 98 greatest common divisor (gcd), 2, 61 group, 6 group action, 12 group algebra, 51 group of symmetries, 7 group of units, 47 group ring, 51 groupoid, 5 Hilbert basis theorem, 133 homomorphism of groups, 9 homomorphism of rings, 50 hyperbolic pair, 138 hyperbolic plane, 138 ideal generated by, 53 ideal, left or right, 52 ideal, two-sided ideal, 52 identity, 6 independent modules, 107 indeterminate, 66 index of subgroup, 14 injective hull, injective envelope, 119 injective module, 115 inner automorphism, 22 integers modulo n, 4, 6 integral domain, 47 internal direct product of groups, 35 internal direct sum, 108 internal direct sum of groups, 34 internal weak direct product of groups, 34 inverse, 6 inverse modulo, 4 irreducible element, 59 isometry, 137 isomorphic (series), 81 isomorphism of graphs, 143 isomorphism of groups, 9 isomorphism theorems of groups, 18 isomorphism theorems of rings, 54 Jacobson radical, 134 Jordan-Hölder theorem, 84 kernel, 10, 52 kernel of a module homomorphism, 104 Kronocker’s theorem, 145 Lüroth’s theorem, 181 Lagrange theorem, 14 leading coefficient, 66 least common multiple (lcm), 5, 61 left regular representation, 13 line, 137 linear fractional transformation, 13 linearly dependent, 108 linearly independent, 108 local ring, 57 lower central series, 88 Lucas’ congruence, 34 maximal ideal, 56 maximal normal subgroup, 83 metabelian group, 92 minimal polynomial, 76 module, 103 module homomorphism, 104 modules over a PID, 119 monic polynomial, 66 monoid, 6 monomorphism of groups, 9 multiple root, 78, 152 multiplicity, 78, 152 Nakayama’s lemma, 134 negative Pell’s equation, 61 nilpotent element, 55, 134 nilpotent group, 88 nilradical, 55, 134 Noetherian module, 132 Noetherian ring, 131 norm, 167 √ norm map on Z[ d], 60 normal basis, 178 normal basis theorem, 169, 178 normal closure of a group, 98 normal extension, 157 normal series, 81 normal subgroup, 8 normalizer, 9 odd permutation, 26 orbit, 12 orbit-stabilizer theorem, 15 order of a group, 6 INDEX order of an element, 8 orthogonal complement, 137 orthogonal graph, 143 p-group, 28 Paley graph, 169 partition, 25 Pell’s equation, 61 perfect field, 153 permutation, 7 Poincaré upper half plane, 13 polynomial, 66 presentation of a group, 98 prime element, 59 prime field, 74 prime ideal, 57 prime number, 2 primitive, 69 primitive element, 154 primitive element theorem, 154 primitive root of unity, 170 principal ideal, 53 principal ideal domain (PID), 53 principal ideal ring, 53 principal series, 83 projective linear group, 93 projective module, 112 projective special linear group, 93 purely transcendental extension, 179 quaternion, 48 quotient field, 55 quotient group, 18 quotient ring, 54 rank of a free group, 97 rank of a module, 111 reduced word, 96 refinement, 81 regular n-gon, 8 regular module, 103 retraction, 106 Riemann sphere, 13 ring, 45 ring of Gaussian integers, 63 ring of polynomials, 66 ring of quadratic integers, 60 root of unity, 170 rotation group, 8 section, 106 semi-direct product, 17 semigroup, 5 separable element, 152 separable extension, 153 separable polynomial, 152 short exact sequence, 104 short exact sequence of groups, 19 simple extension, 76, 154 simple group, 27, 83 skew field, 47 solvable by radicals, 175 solvable group, 85 special linear group, 92 187 spectrum of a ring, 57 split (polynomial), 145 split exact sequence, 106 splitting field, 146 stabilizer, 12 strongly regular graph, 139 subgroup, 7 subgroup generated by, 8 submodule, 104 submodule generated by, 108 subnormal series, 81 subring, 45 subring generated by, 45 sum, 107 syllable, 96 Sylow p-subgroup, 30 Sylow theorems, 28 symmetric group, 7 symmetric rational function, 159 symplectic bilinear form, 137 symplectic graph, 139 symplectic group, 137 torsion element, 127 torsion free, 127 torsion subgroup, 35 torsion submodule, 127 torsion-free rank, 120 trace, 167 transcendence basis, 180 transcendence degree, 180 transcendental element, 76, 179 transposition, 25 transvection, 93 trivial G action, 13 trivial character, 43 unimodular, 137 unipotent matrix, 93 unique factorization domain (UFD), 59 unit, 46 unitary Cayley graph, 101 unity of a ring, 45 universal mapping property of a free group, 97 universal mapping property of a free module, 108 upper central series, 89 valuation map, 63 vector space, 103 vertex transitive, 143 von Dyck’s theorem, 99 Wedderburn’s theorem, 49, 174 well-ordering principle, 1 Wilson’s theorem, 28 word, 96 zero, 45 zero divisor, 47 zero ring, 46