Daniel A. Marcus Number Fields @)) SpringerDr. Daniel A. Marcus Department of Mathematics California State Polytechnic University Pomona, CA 91768 USA Editorial Board (North America) 8. Axler FW. Gehring Department of Mathematics Department of Mathematics Michigan State University University of Michigan East Lansing, Ml 48824 Ann Arbor, Ml 48109 USA USA P.R. Halmos Department of Mathematics Santa Clara University Santa Clara, CA 95053 USA Mathematics Subject Classification (1991): 12-01, 12Axx, 12Cxx, 11Rxx, 11Txx LUbrary of Congress Cataloging in Publication Data Marcus, Daniel A. 1945~ Number fields. (Universitext) “This book grew out of the lecture of a course -- at Yale University in the fall semester, 1972." Bibliography: p. Includes indexes. 1. Algebraic number theory. 2. Fields, Algobralc. |. Tile. A247. MB46 51274 77-2467 All rights reserved. No part of this book may be translated or reproduced in any form without written permission from Springer-Verlag. Printed on acid-free paper. © 1977 by Springer-Verlag, New York Inc. Printed in the United States of America. 987654 ISBN 0-387-90279-1 Springer-Verlag New York Berlin Heidelberg ISBN 3-540-90279-1 Springer-Verlag Berlin Heidelberg New York SPIN 10575510To my parents, Selma and Louis MarcusAcknowledgments I thank John Yang for the determination of the algebraic integers in a biquadratic field (exercise 42, chapter 2) and Dodie Shapiro for the very careful final typing, which 1s what you see here. Also, thanks go to my wife Shelley for her unfailing wit and to my son Andrew, age 4, for his cheerful presence.Table of contents FOREWORD 2 ee eee et et ee eee eet CHAPTER 1: A SFECIAL CASE OF FERMAT'S CONJECTURE. 2... . See CHAPTER 2: NUMBER FIELDS AND NUMBER RINGS «1... ...--- CHAPTER 3: PRIME DECOMPOSITION IN NUMBER RINGS... 2... 24. CHAPTER 4: GATOTS THEORY APPLIED 10 PRIME DECOMPOSITION... . CHAPTER 5: THE IDEAL CLASS GROUP AND THE UNIT GROUP... 2... CHAPTER 6: THE DISTRIBUTION OF IDEALS IN A NUMEER RING... . CHAPEER 7: THE DEDEKIND ZETA FUNCTION AND THE CLASS NUMBER FORMULA CHAPTER 8; THE DISTRIBUTION OF PRIMES AND AN INTRODUCTION TO CLASS APPENDIX 1; COMMUZATIVE RINGS AND IDEAIS 2. eee ee eee eee APPENDIX 2: GALOIS THEORY FOR SUBFIELDS OF ©. ...-....2. APPENDIX 3: FINITE FIELDS AND RINGS ©. ee. ee eee eee eee APPENDIX k: WO PAGES OF PRIMES 2. eee eee eee eee eee FURTHER READING 2 6. eee eee ee ee ee eee eee Boo INK ee ee ee ee ee eee ee eee INDEX OF THEOREMS 2 eee eee ee ee ee ee eee eee LEST OF SYMBOIS 6 ye ee ee ee ee ee eee ee ee vii 98 130 158 223 8 Be 213 276Foreword This book grew out of the lecture notes of a course which I gave at Yale University in the Fall semester, 1972. Exercises were added and the text was rewritten in 1975 and 1976. The first four chapters in their present form were used in @ course at Ohio State University in the Fall quarter, 1975. The first six chapters can be read, in conjunction with appendices 1-3, by anyone who is familiar with the most basic material covered in standard under graduate courses in linear algebra and abstract algebra. Some complex analysis (meronorphic functions, series and products of functions) is required for chapters 7 and 8. Specific references are given. ‘Te level of exposition rises as the book progresses. In chapter 2, for example, the degree of a field extension is defined, while in chapter b it is assumed that the reader knows Galois theory. The idea is to make it possible for someone with little experience to begin reading the book without difficulty and to be lured into reading further, consulting the appendices for background material when necessary. I have attempted to present the mathematics in a streightforward, "down to earth” manner that would be accessible to the inexperienced reader but hopefully stili interesting to the more sophisticated. ‘Thus I have avoided local methods with no apparent disadvantages except possibly in exercises 20-21 of chapter 3 and exercises 19-22 of chapter }. ven there I feel that it is worthwhile to have available "direct" proofs such as I present. Any aviwardness therein can be takenvidi vy the reader as motivation to learn about localization. At the same tine, it is assumed thet the reader is reasonably adept at filling in details of arguments. In many places details are left as exercises, often with eleborate hints. The purpose of this is to make the proofs cleaner and easier to read, and to promote involvement on the part of the reader. Major topics are presented in the exercises: fractional ideals end the @ifferent in chapter 3, ramification groups and the Kronecker-Weber ‘Theorem in chapter 4, fundamental units in non-totally real cubic fields in chapter 5, cyclotomic class nusbers and units in chapter 7. Many other results appear in step-by-step exercise form. Among these are the determination of the algebraic integers in pure cubic fields (chapter 2), the proof that prime divisors of the relative different are ramified over the ground field (chapter 1), and the Frobenius Density Theorem (chapter 7). I have taken the Liberty to introduce some new terminology (“nunber ring" for the ring of algebraic integers in a number field), a notational reform (([I\| for ‘the index of an ideal I in a number ring, rather then the more cumbersome N(I) ), and the concept of polar density, which seems to be the "right" density for sets of primes in a number field. Notice, for example, how easily one obtains Theorem 43 and its corollaries. Chapter 8 represents a departure from tradition in several ways. The distribution of primes is handled in an abstract context (Theorem 48) and without the complex logerithn. The main fects of class field theory are stated without proof (but, I hope, with ample motivation) and without fractional ideals. Results on the distribution of primes are then derived from these facts. It in hoped thet this chapter will be of some help to the reader who goes on to stuiy class field theory. Daniel A. Marcus Columbus, Ohio Sune, 1977Chapter 1 A special case of Fermat's conjecture Algebraic nuiber theory is esséntially the study of nusber fields, which are the finite extensions of the field @ of rational numbers. Such fields can be useful in solving problems which at first appear to involve only retional numbers. Consider, for example, this problem: Find ell primitive Pythagorean triples: i.e., integer solutions of x +y° = 2 having no common factor. Assuming that we heve such a triple end considering the equation modi, we find immediately that z must be odd. This will be used later. Now comes the Antroduction of e number field (namely Q[1] = (a + bi: a,b € Q}) into the problem: if we factor the left side of the equation we obtain (x + yilx- yi) =f ‘ond thus we have a multiplicative problem in the ring of Gaussian integers @[i] = {a+ bi: o,b € Z). It is well know (see exercise 7 at the end of this chepter) that Z{i] is e wique factorization domain: every nonzero Gaussien integer can be expressed in a unique way (up to order and unit factors) as a product of Gaussian primes. We will use this fact to show that x+ yi has the forn uo? for some Gaussian integer @ and cone Geussion integer uit u. If we then write @=m+ ni and observe that the only units in Zi] are +1 and4&1 (see exercise 2), we obtain {xy} = {4@%—n2),t2mn} and z = £ (ut 0”), It is obviously necessary thet m and n be relatively prime and not both odd. (otherwise x, y, and 2 would have a factor in common) end it is easy to see that fe primitive Pythagorean triple zesults from any such choice of m and n, anda choice of signs. Furthermore it is clear that nothing is lost if we take only positive m and ne ‘ms the problen will be solved if we can chow that for any primitive solution, x+yL has the form ud’. To do this, it is enough to show that if ff is a Gaussian prime dividing x + yi, then in fact 1 divides x+ yi an even number of times: 1 |x+yi ana t+ 4x4 yi for some even e. Since (x +yi)(x = yi) = 2% ana 1 obviously divides 2 an even nutber of times (twice as many times as it divides =), we need only show that ttx - yi. ‘Thus, supposing that 1 divides both x+yi and x - yi, we wanta contradiction. Adding, we get T|2x. Also we have f[z. But 2x and z are relatively prine integers (recall that 2 is odd, and if x and z had a non- trivial factor in common, then so would x, y, and z). So there exist integers m and n such that Qm+zn=1. But then M1 in @[i]. This is impossible since ff is a prime, not a unit. ‘Thus by working in the field @[i] we have determined all primitive Pythagorean triples. Since this was so successful, let us try to apply the same idea to the equation x® + y" = 2" for n > 2. Fermat, in his famous marginal note, claimed that he had a proof that there are no solutions in nonzero integers when n > 2. This is known as “Fermat’s last theorem" or "Fermat’s conjecture." For over three centuries it has been one of the most famous unsolved problems in mathematics.* Using our result on primitive Pythegorean triples, we can show that Femat was right for n= and hence (automatically) also for any miltiple of +. (See *rermat’s last theorem may have finally been proved in 1993-94 by Andrew Wiles using concepts from the theory of elliptio curves.exercise 15.) It is therefore sufficient to consider only the case in which n is an odd prime p, since if no solutions exist when n =p. then no solutions exist when n is a multiple of p. ‘Thus the problem is to show that if p is an odd prime, then x? + y? =z? has no solution in nonzero integers x, y, z+ Suppose, for some odd prime p, there is a solution x,y,z € % - (0). Clearly we may assume that x, y, 2 have no common factor (divide it out if there is one). We want a contradiction. It is convenient to separate the argument into ‘two cases: either p divides none of x, y, 2 (case 1), or else p divides exactly one of them (case 2). (If p divided more than one then it would divide all three, which is impossible.) We will consider only case 1. It is easy to show that x3 +y3 = 23 has no case 1 solutions: If x, y, and z are not multiples of 3, then in fact x3 +3 £23 (moa 9) since each of these cubes is = +1 (nod 9). Now assume p>33; x, y, and z are not mitiples of p; and xP +yP = 2? Factoring the left side, we obtain @) (x + y)(x + yo)(x + yu®) eee (ee yw) = DP ami fp where w is the pth root of mity e (fo see why this is true, note that 1, o, ow, wo} are the p roots of the polynomial t? -1, hence we have the identity (e) Pere (b(t - ot F) o. (e- P4H), from which (1) follows by substituting the number <* for the variable t .) ‘Thus we have a multiplicative problem in the number field @[{w], and in fact in the subring Z{w].” Kumer attempted to prove Fernat's conjecture by consider~ ing whether the unique factorization property of % and Z{i] generalizes to x Ole] = (ey Fawr ee Fe Blo] = (a # OF ee + Bgthe ring Z[w]. Unfortunately it does not. For example if p = 23, then not ell members of Z[w] factor uniquely into irreducible elements: i.e., elements @ € Z[w] which are not units end such that whenever a= py, either p or y is a unit (see exercise 20). In other words, Z{w] is not a unique factorization domain (UFD) for p= 23. It is, however, a UFD for ali primes less than 23. For these primes it is not difficult to show that x° + y® = 2? has no case 1 solutions. ‘The azgument can be organized es follows: Assuming that Z[w] is a UPD, it can be shown thet x + yw hes the form ua? for sone a € Z{w] and some wit u€ Z{w]. Tt can then be shown thet the equation x+yw= uo?, with x and y not divisible by p, implies thet x =y (mod p). (See exercises 16-28 for the deteils.) Similarly, writing x? + (-z)P = (-y)P, we obtein x= -z (mod p). But then xP a xP 4 yP uw oP = -xP (mod p) , implying that p|3x’. since pfx and p #3, this is a contrediction. ‘Thus case 1 of Fermat's conjecture can be established for all primes p for which Zw] is e UFD. What can be done for other primes? Unique fectorizetion in @[w] wes needed only for the purpose of deducing x + yw= uc? from equation (1); might it not ‘be possible to deduce this in some other way? ‘The answer is yes for certain values of p, including for exemple p= 23. ‘his results from Dedekind's enazing @iscovery of the correct generalization of unique fectorizetion: although the elenents of Z[u] may not factor uniquely into irreducible elements, the ideals in this ring alweys fector uniquely into prime ideals, Using this, it is not hard to show thet the principal ideal (x + yw) 4s the pth power of some 1deal I (see exercises 19 and 20). For certain p, called "regular" primes (defined below), it then follows that I must itself be a principal ideal, say (a), so that (x + yw) = TP = (oP = P)and thus again we have x+yw= ua for some unit u. As before, this implies x =y (mod p) anda contradiction follows. ‘Thus case 1 of Fermat's conjecture can be established for all regular primes, which we now define. There is an equivalence relation ~ on the set of ideals of Z[w], defined as follows: for ideals A and B A~B iff GA= pB forsome a, BE Zw]. (Verify that this is an equivalence relation.) It tums out (see chapter 5) that there are only finitely many equivalence classes of ideals under ~. ‘The number of classes is called the class number of the ring Zlw], and is denoted by the letter h. Thus h isa function of p. DEFINITION: Aprime p is regular iff p{h. fo explain why I (in the equation (x + yw) = I”) must be principal whenever p isa regular prime, we note first that the ideal classes can be multiplied in ‘the obvious way: the product of two ideal classes is obtained by selecting an ideal from each; multiplying them; and taking the ideal class which contains the product ideal. This ts well-defined: ‘The resulting ideal class does not depend on the particular ideals chosen, but only on the two original ideal classes (prove this). Multiplied in this way, the ideal classes actually form a group. The Adentity element is the class Cy consisting of all principal ideals (which really is a class; see exercise 31). The existence of inverses will be established in chapter 3. Thus the ideal classes form a finite abelian group, called the ideal Class group. If p 4s regular then clearly this group contains no element of order p, and it follows that if I” is principal then sois I: let © be the ideal class containing I; then C” is the class containing 1”, which is Coe p, it follows that C= C,, which shows that I is principal. Since Cy is the identity in the ideal class group and C cannot have order As we noted before, this leads to a contrediction, showing that xP 4 yP = oPjhas no case 1 solutions (i.e., solutions for which p4xyz) when p is a regular prime. It is also possible, although somewhat more difficult, to show that no case 2 solutions exist for regular primes. (For this we refer the reader to Borevich and Shafarevich's Number Theory, p. 378-361.) Thus Fermat's conjecture can be proved for all regular primes p, hence for all integers n which have at least one regular prime factor. Unfortunately irregular primes exist (e.g. 37, 59, 67)« In fact there are infinitely many. On the other hand, it is not known if there are a infinitely many regular primes. Tn any case our attempt to prove Fermat's conjecture leads us‘to consider various questions about the ring Z[w]: What are the units in this ring? What are the irreducible elements? Do elements factor uniquely? If not, what ’ can we say sbout the factorization of ideals into prime ideals? How many ideal classes are there? The investigation of such problems forms a large portion of classical algebraic number theory. More accurately, these questions are asked in subrings of arbitrary number fields, not just @[w]. In every number field there is a ring, analogous +o [lw], for which there are interesting answers. EXERCISES 2,2 1-9: Define WN: Zi] > Z vy Na+bi)sa ey. 1. Verity that for all op € @[1], Nap) = N(@)N(p), either by direct computation or by using the fact that Na + bi) = (a + bi)(a - bi). Conclude that if aly in Zi], then Ma)|Ny) in Zw. + Conclude that the 2. let @€ Bi]. Show that a is a mit iff Ma) only units ere #1 and +i. 3. Iet @€ Zli]. Show that if N(@) is aprimein Z then @ is irreducttle in [i]. Show that the same conclusion holds if Ma) = p*, where p isaprime in Z, p#=3 (mod4).4, Show that 1-4 4s irreducible in [i] and that 2 = u(1- 4) for some unit u. 5. Notice that (2+ 4)(2 - 4) =5 = (1+ 2i)(1 -2i). How is this consistent with unique factorization? 6. Show that every nonzero, non-unit Gaussian integer @ is a product of irreducible elements, by induction on N(q) . 7. Show that [i] is a principal ideal domain (FID); i-e., every ideal I is principal. (As shown in Appendix 1, this implies that [i] 4s a UFD.) Suggestion: Take @ €I- (0} such that N(@) is minimized, and consider the multiplies yo, y € Z[i]5 show that these are the vertices of an infinite family of squares which fill up the complex plene. (For exemple, one of the squares has vertices 0, a, ia, and (1+ i); all others ere translates of this one.) Obviously I contains all ya3 show by a geonetric argument that if I contained anything else then minimality of N(a) would be contradicted. 8. We will use unique factorizetion in [i] to prove that every prime Ppl (mod +) is a sum of two squares. (a) Use the fact that the multiplicative grow of integers nod p is cyclic to show thet if p=1 (mod) then 2° = -1 (mod p) for sone n€ Z. (b) Prove that p cannot be irreducible in [i]. (Hint: pin? 4a = (n+ 4)(n-4).) (c) Prove that p is a sum of two squares. (Hint: (b) shows that p= (a4 bi)(e + ai) with neither factor a unit. ‘Teke norms.) 9. Describe ali irreducible elements in Zi] . emi/3 10-1: Tet w=e =-3+ Gi. Define N: Zlu] + Z vy Ma+ bw) = &- aba.10. 113. a be Show that if a+ bw is written in the form u+vi, where u and v are real, then Wa + bu) = 4v. Show thet for all a,p € Z{w], Mp) = M(a)N(p), either by direct computation or by using exercise 10. Conclude that if aly in Z{w], then N(c)|M(y) in Z. Tet @€ Zlw]. Show that a ise uit iff No) =1, ond find ali units in Zfw]. (There ere six of them.) Show that 1-w 4s irreducible in Z{u], end thet 3 = u(l-o)* for some unit u. Modify exercise 7 to show that Z{w] 1s a PID, hence a UFD. Here the squares are replaced by parallelograms; one of them has vertices 0, a, wa, (w+ 1), and all others are translates of this one. Use exercise 10 for the geonetric argument at the end. Here ds a proof of Fermat's conjecture for n=: re xy! = 2" has @ solution in positive integers, then so dees x ayleav. tet x, y, w be a solution with omelest possible w. then x, y*, w 4s 9 primitive Pythegorean triple. Assuming (without loss of generality) that x 4s odd, we can write os em, wane ane with m and n relatively prime positive integers, not both odd. (a) Show that PoP, neers, mea ek with r and s relatively prime positive integers, not both odd. (b) Show that r, s, and m are pairwise relatively prime. Using y= lirsm, conclude that r, s, and m are all squares, sey ef, and ce.16-268: Iet p be an odd prime, w= e 16. MW. 1. 20. al. 4 yt (c) Show thet et + bt =, ond thet this contradicts mininality of w. Pri /e Show that (-oQ-u) . V-PYep ty considering equation (2). Suppose that Z[w] isa UFD and |x + yw. Show that ff does not divide any of the other factors on the left side of equation (1) by showing that if it did, then 1 would divide both z and yp (Hint: use 16); but z and yp are relatively prime (assuming case 1), hence 2m + ypn=1 for some mné€ Z, How is this a contradiction? Use 17 to show that if Z[w] is a UPD then x+yu= uo’, ae alu], u awitin z{w]. Dropping the assumption that Z[w] is @ UFD but using the fect that ideals factor uniquely (up to order) into prime ideels, chow that the principal ideal (x + yw) hes no prime ideal factor in common with any of the other principal ddeals on the left side of the equation a) (x + y)Gc + yu) vee Ge uP) = (e)P in which all fectors are interpreted as principal ideals. (Hint: modify the proof of exercise 17 appropriately, using the fect that if A is an ideal dividing another ideal B, then A>B.) Use 19 to show that (x + yw) = I for some ideal I. Show thet every menber of @{w] 4s uniquely representable in the form aot aot eu tie + a, ow *, a, €@ Vi by showing thet w is a root of the polynomialaL. 23. ah. 25. (continued) a(t) = Pa Pe ta and that f(t) is irreducible over Q. ° (Hint: It is enough to show that £(t +1) is irreducible, which can be established by Eisenstein's criterion (eppendix 1). It helps to notice that #(t +1) = ((t +1)? - 1)/e.) Use 21 to show that if a € Z{w] and plo, then (writing Waagt aor. ta, puP*, a € Z) all a; are divisible yp. Define congruence mod p for p,v € Z{w] as follows: p= y (mod p) iff p- y= 6p for some BE Zu]. (Equivalently, this is congruence mod the principal ideal pZ{w].) Show thet if p = y (mod p), then B = ¥ (mod p) where the bar denotes complex conjugation. Show that (p+ y)P =p? + yP (mod p) and generalize this to sums of arbitrarily many terms by induction. Show that vae Z[w], o? is congruent (mod p) to some a€ Z. (Hint: write @ in terms of w and use 2h.) 26-28: Now assume p>5- We will show that if x + yw = ua? (mod p), 26. a@€ Zw), u awit in Z[w], x and y integers not divisible by p, then x =y (mod p). For this we will need the following result, proved by Kummer, on the units of z[w] : LEMMA: Tf u isa unit in Z{w] and U is its complex conjugate, then u/@ is a power of w. (For the proof, see chapter 2, exercise 12.) Show that x +yw = ua? (mod p) implies x + yw = (x + yo") (moa p)26. 28. 29. (continued) for some k€ Z. (Use the Lemma on units and exercises 23 end 25. Note that D- uw.) Use exercise 22 to show that a contrediction results unless k = 1 (mod p). (Recall that ptxy, p>5, and wba uP "4 og w e120.) Finally, show x = y (mod p). tet w= LMS verity that the product Pee er ee Taser soo eal ruta is divisible by 2 in Z{w], although neither factor is. It can be shown (see chapter 3, exercise 17) that 2 4s an irreducible element in Z[u] 5 it follows that Z[w] cannot be a UFD. 30-32: R 4s an integral domain (commutative ring with 1 and no zero divisors). 30. Show that two ideals in R are isomorphic as R-modules iff they are in the same ideal class. Show thet if A is an ideal in R andif GA is principal for some @E€R, then A 4s principal. Conclude thet the principal ideals form an ideal class. Show thet the ideal classes in R form a group iff for every ideal A there is an ideel B such thet AB is principal,Chapter 2 Number fields and number rings A number field is a subfield of @ having finite degree (dimension as a vector space) over Q. We know (see eppendix 2) that every such field has the form Q[o] for come algebraic nutber a €@. If a is a root of an irreducible polynomial over @ heving degree n, then @fe] = (ep Fase te, yobs a €@ VE) ne and representation in this form 1s unique; in other words, (1,c,-..,0°71) 4s basis for @[a] as a vector space over Q. We have ellvesdy considered the field @[y] wnere wo /P, p prime. Recall thet n=p-1 in thet case. More generally, et we, nm not necessarily prince. The ficld @[w] is called the m‘” cyclotomic field. Thus the first two cyclotomic fields are both just @, since w= 1, -1 (xesp.) for m=1, 2. Moreover the third cyclotonic field 4s equal to the sixth: If we set w= MVE | nen we nut = ~(u?)%, viteh shows thet @fu] = [uP]. In general, for oda m, the nt” cyclotomic field is the same as the 2n', (show that 1 w= eH then w= ut? € @fu®].) on the other hand, we will show that the cyclotomic fields, for m even (m>0), are all distinct. this will essentially follow from the fact (proved in this chapter) that the degree of the m®” cyclotamic field over Q@ is g(m), the number of elements in the set13 (er 1O, are called the reel quadratic fields; the Q/m], m<0, the imaginary quadratic fields. Thus Q@[4] is an imaginary quadratic field as well as a cyclotomic field. Notice that Q[/-3] is alsoa cyclotomic field (which one?). While working with the p“? cyclotomic field fw] in chapter 1, we promised to show that the ring {w] has certain nice properties: for example, every ideal factors uniguely into prime ideals. ‘his is true more generally for the ring Z{w] in any cyclotomic field. It is also true for 2{,/m] for certain velues of m. However it fails, for example, for Z(,/-3] (see exercise 2). Nevertheless we know that Q[,/-3] has a subring in which ideals factor uniquely into primes, nanely af+/3]- zo), w= ets, ‘his ring consists of all et ave m, a=b (nie). (Verify this; recall that Z{w] = (2 + bu: a,b € %).) We will see that every nunber field conteins a ring aitferent fron % (if the field is not @) having thie unique factorization property; it consists of the algebreie integers in the field. DEFINITION: A complex nunber is an elgebreic integer iff it is a root of some monic (leading coefficient 1) polynomial with coefficients in Z.as Notice that we have not required that the polynomial be irreducible over Q. ‘Thus we cen easily see that w= eM as an algebraic integer, since it is a root of x"~1. It 4s true, however, that every algebraic integer @ is a root of some monic irreducible polynomial with coefficients in Z: THEOREM 1: Tet be an algebrade integer, and let f be a monic polynomial over 2% of least degree having @ as aroct. Then f is irreducible over Q. (Equivelentily, the: monic irreducible polynomial over Q having @ as a root has coefficients in Z.) IGMA: Tet f be @ monic polynomial with coefficients in Z, and suppose f= gh where g and h are monic polynomials with coefficients in Q@. Then g ond h actually have coefficients in z. Proof: Let m (resp. n) be the smallest positive integer such thet mg (vesp. nh) has coefficients in 2%. ‘Then the coefficients of mg have no comon fector. (Show that if they did then m could be replaced by a smaller integer; use the fact thet g is monic.) The same is true of the coefficients of nh. Using this, we can show that m= 1: If m>1, take any prime p dividing mn and consider the equation mf = (mg)(nh). Reducing coefficients mod p, we obtain O-=img th where the bars indicate that coefficients have been Feduced mod p. (We have applied the ring-homomorphism [x] + [x] +) Bat [x] is an integral domain (since z, is; this is easy to show), hence if or h= 0. But then p divides all coefficients of either mg or mh; as we shoved above, this is impossible. Thus m=n=1, hence gh € Z{[x]. 0 Proof of Theoren1: If f is not irreducible, then f= hg where g and h ere nonconstant polynomials in @[x]. Without loss of generality we can assume thet g ond h are monic. Then g,h € Z[x] by the lemma. But @ is a root of either g or h and both have degree less than that of f. ‘his is a contradiction. Oa COROLLARY 1: The only algebraic integers in @ are the ordinary integers. Q COROLLARY 2: Iet m be @ squarefree integer. The set of algebraic integers in the quadratic field Qf[,Jm] is (a+b/m: a,b€ Z} if m=2 or 3 (modh), a,b € Z, a=b (mod 2)} if m=1 (moa h). Proof: let @=r+s/m, r,s€Q. If s #0, then the monic irreducible polynomial over @ having @ as a root is ~ are - ms. Thus @ is an algebreic integer iff 2r end r°-ms* are both integers. We leave it as an exercise to show that this implies the result stated shove. 0 Corollary 2 shows that the algebraic integers in @[,/m] forma ring. The sone is true in any number ficld. To prove thie, it ie enough to show that the suri and product of two algebraic integers are also algebraic integers. For this At 4s helpful to esteblish some alternative charecterizations of algebraic integers. THEOREM 2: The following are equivalent for a €€: (1) @ 4s an algebraic integers (2) The additive group of the ring Z[c] is finitely generated; (3) @ de a member of come stibring of € having a finitely generated edditive group; (1) CA CA for some finitely generated additive subgroup ACC. Proof: (1) (2): If @ 4s a root of a monic polynomial over % of degree n, then in fact the additive group of Za] is generated ty 1,0, ..., 71, (2) = (3) = (4) trivietiy. (4) = (1): Tet Ay,+++8, generate A. Expressing each aa, as a linear combination of a,,...,8, with coefficients in Z, we obtein16 (--Q vhere M isan nxn matrix over %. Equivalently, ( (ot - M) ‘1 is the zero vector, where I denotes the nxn identity matrix. Since the ay ere not all zero, it follows that OL -M has determinant ©. (In other words, we have shown that @ is an eigenvalue of M.) Expressing this determinant in 2 ‘terms of the n° coordinates of OL -M, we obtain o” 4 lower degree terns = 0. ‘Thus we have produced a monic polynomial over 2% having a as a root. O COROLLARY 1: If a and f are algebraic integers, then so are a+ 6 end ap. Proof: We know that fa] ana @[p] have finitely generated additive groups. Then so does the ring @Z{a,p]. (If ,...,0, generate Z[a] and Byy+++sB, generate z[p], then the m products @,p, generate z{o,p) -) Finally, 2[0,p] contains a+ p and ap. By characterization (3), this implies that they are algebraic integers. 0 EXERCISE: Pick your two favorite algebraic integers and apply the determinant Procedure to obtein monic polynomials for their sum and product. Tais result shows that the set of algebraic integers in ¢ is a ring, which we will denote by the synbol A. In particular ANK is a subring of K for any nunber field K. We will refer to A (MK as the mumber ring corresponding to the number field K. We have determined the number rings corresponding to @ end the quadretic fields. For the cyclotomic fields we have AN Qu] = Zw] 5vy however at this point ell that is clear 1s that AN@[w] contains Zlw] (since w€A end AN@fw] is a ring). To esteblish equality we will need come further information about @[w]: specifically, its degree over @ ond its discriminant. THE CYCLOTOMIC FIELDS tet w= T/ proxy conjugate cf w (root of the sane szreducttile polynomial over @) 4s clearly also an m” root of 1 and is not ann’ 1 for any n Z [x], we obtain the fect that g(x?) 4s aivisitie ty Hx) in z, {#1 - But (x?) = (&(x))? (see exercise 5) and m,[x] is © wmique factorization domain (see appendix 1); it follows that F and § have a common factor h in Bod. men w]iy =x"- 1. Mis implies (see exercise 6) that h divides the derivative of x"-1, which is m™"1, (tere the ber denotes m reduced mod p.) Since ptm, m0} then in fect h(x) ds just e monomial (agin using unique factorization in Z,[x]). But this is impossible since nx” - 1. That completes the procf. O COROLLARY 1: @[w] hes degree {m) over @. Proof: w has {m) conjugates, hence the irreducible polynomial for w over Q has degree Ym). O COROLLARY 2: The galoie group of @[w] over @ is isomorphic to the multiplicative group of integers mod m m= (cr 1gk e(u*) for each g € 2[x] )- Broof: An automorphism of @{o] 4s uniquely determined by the image of w, and Theorem 3 shows that qw can be sent to any of the ul, (k,m)=1. (Clearly it can't be sent anywhere else.) ‘This establishes the one-to-one correspondence between the Galois group and the multiplicative group mod m, and it remains only to check that composition of automorphisms corresponds to miltiplication mod m. We leave this as an exercise. O As an application of Corollary 2, we find that the subfields of fu] correspond to the subgroups of ,. ‘In perticuler, for p prime, the p‘™19. cyclotomic field contains a unique subfield of each degree dividing p-1 (since 2, is cyclic of order p- 1). Thus for each odd prime p, the p“" cyclotomic field contains a unique quadratic field. This turns out tobe Q[,/+p] with the sign depending on 7 (see exercise 8). We will exploit this fact in chapter 4 to prove the quadratic reciprocity lav. conortany 3: tet w= 4/9, re m 4s even, the only roots of 1 in Q{w] are the m™ roots of 1. If m is odd, the only ones are the ont” roots of 1. Proof: It is enough to prove the statement for even m, since we know that the n’™ cyclotomic field, m odd, is the same as the 2n°". Tus, assuming m is even, suppose @ is @ primitive x" root of 1 in @fu]. (ie, @ is ax root of 1 but not ann” root for any n and [asl = |4||B] for matrices A end B. 0 COROLLARY : dise(O,,---,4,) €Q; and if ell a, are algebraic integers, then Aise(ay,---,C,) €z.o0 Among other things, the discriminart determines whether the @, are linearly dependent: THEOREM 7: dise(a4,--+,0,) = 0 Aff ),...,0, are linearly dependent over Q. Eroof: Tt is easy to see that if the cj are linearly dependent over @ then so are the colums of the matrix fo,(05)1 3 thus the discriminant is O. Conversely, if disc(oy,-..,4,) = 0, then the rows Ry of the matrix (2,0,)1 are linearly dependent. Suppose that 4,---,,, ere linearly independent over @. Fixing rational numbers 9,,-.-,8, (not all ©) such that aR) tee +aR ie the zero vector, consider @ = a0) ++. +80,. Necessarily a#O. Moreover ty considering only the j* coordinate of each row, we obtain the fact that Taa,) =0 foreach J. Since the a, are assumed to be linearly independent over Q, they forma basis for K over Q} it then follows (since a #0) that the same is true of the ca, . But then T(p) = 0 for every PEK (why?).26 This is clearly a contradiction since, for example, (1) =n. O Theoren 7 shows that every basis for K over @ has a nonzero discriminant. We can obtain a relatively simple forma for the discriminant in the case of a ‘vasis consisting of the powers of a single element: TawoREM 6: Suppose K = Q[o], and let cy,...,0, denote the conjugates of @ over Q. Then dise(tja,..08)- TH @,-0,% = ame) isr 4 igr3): Z{l-] = Z[w] end dise(1 - w) = dise(w) - Eroof: {1 -u] = Zw] 4s obvious, since w=1-(1-w). Mis in itself implies that the ciscrininants are equal (see exercise 26). However it is probably easier to use Theorem 8: As a, runs through the conjugates of w, 1 - Oy runs through the conjugates of 1-w. ‘Thus T (@-o)-@-a)? lsrI: gcd(r, gca(m)) = 1. ‘To prove the theorem, we have to show that, for any such a, rld. Clearly it will be enough to show that r|disc(R) ; by symmetry r will also divide @ise(S) and we will be done. ‘The lemma shows that every embedding o of K in @ extends to an embedding (which we also call o) of KL in ©, fixing each point of L. Hence for each we have ofa) = AE ofay)o, «Setting for each i=1,...,m, we obtain m equations n E_ ofa, )x, = o(@), del avn one for each o. Now sclve for the x, by Cramer's rule: x, = y,/8, where 6 is the determinant formed by the coefficients o(o,), and y, is obtained from 8 by replacing the 1** are algebraic integers, since all o(a,) and o(a) eres moreover &° = disc(R). column by the o(0)). It is clear thet 8 ondall y, Setting ¢ = aisc(R), we have ex, = by; EA then in fact ex, = z Hp, €AnL=s. jer Recalling that the 6, form an integral basis for 5, we conclude that the rational numbers en s/e must all be integers: Thus r divides all ems Since by assumption r is relatively prime to goa(m, 5) » it follows that rle = disc(R). O Using Corollary 1, we can prove COROLLARY 2: Tat K=e@fw), w=e™ poanK. men R= ao]. Proof: ‘This has already been established if m is a power of a prime. If nm is not a power of a prime, then we can write m= mym,, for sove relatively prine integers m,,m,>1. We will show that the results for m, and for m imply the result for m. (‘hus we ere proving R= Z{w] by induction om m.) Setting36 R= ANK, RB =ANK, we assume (inductive hypothesis) that R, = Z{w,], R, = Z{u,]. 1 apply Corollary 1, we have to show that K= IK, and that the degree and discriminant ™ conditions hold. Clearly w* =a, oem. It follows that w= ulub for some r,s € % (why?) andhence K=I\K,. Moreover this shows that Zl] = Z[w,] Wu] The degree condition holds: g(m) = o(m,)o(m,) since m and m, are relatively prime. For the discriminant condition, recall that we have chown that dise(w,) divides @ pover of mj and aisc(a,) divides a power of m,- Finally, then, we conclude that = RR = Ao] Ze]- zo]. 0 Tt would be nice if every number ring had the form Z{o] for some a. Unfortunately, this is not always the case (see exercise 30). Equivalently, there may not exist an integral basis of the form 1,a,...,0°"*. mis suggests the following vague question: Does there always exist an integral basis whose members are expressed in terms of a single element? Of course the answer is yes, since K = Q{o] for some a and hence every menber of K is a polynomial expression in @ with coefficients in @. ‘his ie not particularly illuminating. However what if we require that these polynomials have some special form? An answer is provided by the following result: THEOREM 13: Let @€R and suppose @ has degree n over Q. Then there is an integral basis #,(@) ty 1) vq ay where the d, ere in % and satiety alale--la, 45 the f, are monic 1, polynomiels over @%, and f, has degree i. The d, are uniquely determined, Broof: For each k, 1), Seen Se a2 and this is in mR, ,1) (why?). Tt follows that Oa. = mff(B) for some m€ Z. Defining 4, = md 1, we have m(p) = cM/a,, which implies thet B= £,(0)/a, for sone £,(c) = oF + lower degree terms. However ve cannot yet say that f£, has integer coefficients; all that is clear is that a8, /a, hes38 integer coefficients. However since f,(a)/d,_,=mB €R, we have £,(@) - af, _1(@) Sead =yeR, and in fact this has been selected so that y €R,. Using our basis for R, ve can write y= B(0)/4, 1 for some g € Z[x] having degree ) Show that the 8°" cyclotomic field contains JZ. (c) Show that every quadratic field is contained in a cyclotomic field: In fact, Q[/ Ti] is contained in the a" cyclotomic field, where d= disc(AN QL Ym]). (More generally, Kronecker and Weber proved thet every abelian extension of @ (normal with abelian Galois group) is contained in a cyclotomic field. See the chapter ) exercises. Hilbert and others investigated the abelion extensions of an arbitrary nunber fleld; their results are known as class field theory, which will be discussed in chapter 8.) With notation as in the proof of Corollary 3, Theorem 3, show that there exist 2M fe _ yx integers u and v such that e _ eY . (Suggestion: First write o- eMib/k 1, yelatively prime to k.) Complete the proof of Corollary 3, Theoren 3, by showing if m is even, nlx, and g(r) ) Show thet 260+ 3b is a root of (Eg) + 2) ov. Use this to find Hear + 3) 6 (c) show that atse(a) = =(le? + 276%). (a) Suppose oF =a 41. Prove that (1,0,0°) is an integral basis for ANn@c]. (See 27e.) Do the sane if @4a=i. Tet K be the biquedratic field Q[,/m,,/u] = (2+ b/m+ en + afm: a,b,c,d € Q}, where m and n are distinct squarefree integers. Suppose m and n ere relstively prime. Find an integral basis and the discriminant of A&M K in each of the cases (a) mn =2 (aoa 4). (b) m=1 (woah), n #1 (modh). (See exercise 23¢. For the general case, see exercise 2.) Tet K=Q[/7,,/ a0] and fix any GO €ANK. We will show that Ank? @a]. Let £ denote the monic irreducible polynomial for a over % and foreach g€ Bx] let & denote the polynomial in m,[x] obtained by reducing coefficients mod 3. (a) show that g(a) is divisible by 3 in Zla] iff g is divisible by Fin wx]. (b) Now suppose ANK= 2Z[a]. Consider the four algebraic integers a= + S70 + JT) O% = +470 - 90) % = @- VG + 4%) a, = - VG - JT).30. 3l. 3h. MT (») (continued) Show thet a11 products 015 (i 4 4) are divisible ty 3 in Z[o], but that 3 does not divide any power of any a, . (Hint: Show thet 0/3 is not an elgebraic integer by considering its trace: Show thet Hal) ah + hs hah and that this is congruent mod 3 (in z{a]) to (+a, +a, + a,)h =u. Why does this imply that (a) = 1 (moa 3) in x 2) (c) Tet a, =£,(c), £, € @[x] foreach i=1, 2, 3,4. Show that FIRE, G45) in a(x] but FTF. Conclude that for each i, F has an irreducible factor (over 2) which does not divide F, but Which does divide all. fj, d#4. (Recall that a(x] is a unique factorization domain.) (a) ‘his shows that F hes at least four distinct irreducible factors over %z+ On the other hand f hes degree at most h. Why is that a 3° contradiction? Show that (/3+,/7)/2 is an algebraic integer, hence the discriminant condition is actually necessary in Corollary 1, Theorem 12. Find two fields of degree 3 over Q@, whose composition has degree 6. (You don't have to look very fer.) : ret wn eMm 53, We know thet Mu) =41 sine o isa mit. Show that the + sign holds. tet w= ETM, wn positive integer. (a) show that Leora ...4u87* ase mitin Zo] if k is relatively prime to m. (Hint: Its inverse is (w -1)/(u" - 1); show that w= ul tor come h € Z.)3h 35+ 36. 4B (continuea) (o) Set (a) (vo) (e) (a) (e) (t) (s) Iet m=p", p aprime. Show that p= u(l-w)" where n = gp") and u isaunitin 2[{w]. (See Lemma 2, Theorem 10.) ee wee, vrere wee, nog, Show that w is a root of a polynomial of degree 2 over @[e]. Show that @[6] = R New] and that Qw] has deeree 2 over this field, (Hint: Qu] > Rn Qw] > ee] -) Show that @[@] is the fixed fleld of the automorphism o of @[w] determined by o(w) = wt. Notice that o is just complex conjugation. Show that AN@(@] = RN alu]. Tet n= olm)/25 show that (oo ur,0%,..,0 ta Yuh) as an integrel basis for Z[w]- Use this to show that (2,0,0,00, ©, 80,4, 4,08" 4y) is another integral basis for z[o]. (Write these in terms of the other basis and look at the resulting natrix.) -1 Show that (1,¢,0,-.-,8""4} is an integral basis for A Nee]. Conclude that AN ee] = ze]. Suppose m is an oda prime p. Use exercise 23 to show that aise(e) =: pP-3)” , show that the + sign must hola. (Hint: (w+ of = F(w- 1)(w4 1); tirst eateutate Mele of tats, For the + sign, note that @[¢] contains ,/disc(6). In the proof of Theoren 13, show that (1,.+.,p) isa basis for R.,1 over (First show that m(p) # 0.) Tet @ de an algebraic number of degree n over @ andlet f and g be ‘two polynomials over @, each of degree 4 so. (Suggestion: see exercise 28.) Suppose @=aa+l. Prove that ANQa]= Za]. (x -x-1 is irreducible over Q; this can be show by reducing mod 3. See appendix 3.)43. ah, 45. 53 (continued) (c) Tet a be squarefree and not #1. ‘Then x 40x +a is irreducible by Bisenstein, Let @ bea root and let a, a, a, and 4, de as in Theoren 13. Prove that ir lite a= 4 <1 ona aga, fe". (See exercise 10. Actually a, = a, = 15 see exorcise 26(a), chapter 3.) Verify that ha 45° ic squarefree 45? is also squarefree then when a = -2, -3, -6, ~7, -10, -11, -13, and -15. Note that if a positive integer m is nota square and not divisible by any prime pP,). By condition (2) for Dedekind domains, we must have P =P).58 Finally, recall that (a) cannot contain a product of fewer than r primes; in particular, Eb € (P,P; P,) - (a). Then y=b/a €K-R and ACR. (Prove this lest assertion.) 0 We resume the proof of Theoren 15. Consider the set A=Z17. this is contained in R (recall IJ c (a)), and in fact A is an ideal (verify this). If A=R then IJ = (a) and we are finished; otherwise A is a proper ideal and we can apply Lemma 2. Thus YACR, Y€K-R. We will obtain a contradic tion from this. since R is integrally closed in K, it is enough to show that Y isa root of a monic polynomial over R. Observe thet A= 217 contains 5 since GET; thus WCYACR. It follows that CJ; to see why this is true go back to the definition of J and use the fact that \ and YA are both contained in R. We leave it to the reader to fil in the detadis. Finally, fixe finite generating set 0,,-+ for the ideal J and use the relation YI CJ to obtein a matrix equation where M icon mxm matrix over R. As in the proof of Theorem 2, we obtain (via the determinant) a monic polynomial over R having Y as a root. That completes the proof. 0 An immediate consequence of Theorem 15 is COROLLARY 1: The ideal classes in a Dedekind domain forme group. (See exercise 32, chapter 1.) O Theorem 15 has two further consequences which, will ensble us to prove unique factorization:59 COROLLARY 2: (cancellation law) If A, B,C are ideals in @ Dedekind domain, and AB = AC, then B=C. Proof: here isan ideal J such that AJ is principal; let AJ = (a). Then CB = OC, from which B=C follows easily. 0 COROLLARY 3: If A and B are ideals ine Dedekind domain R, then A|B iff ADB. Froof: One direction is trivial: AJB=A>B. Conversely, assuming A>B, fix g such that AJ is principal, AJ = (a). We leave it to the reader to verify that the set ¢=235 is an ideal in R (first show it is contained in R) and that AC a Using these results, we prove THEOREM 16: Every ideal in a Dedekind domain R is uniquely representable as a product of prime ideals. Proof: First we show that every ideal is representable as a product of primes: If not, then the set of ideals which are not representable is nonempty and consequently has a maximal member M by condition (1"). MR since by convention R 4s the empty product, being the identity element in the semigroup of ideals, (If you don't like that you can forget about it and just consider proper ideals.) It follows that M is contained in sone prime ideal P (see the proof of Lema 2 for Theoren 15). Then M= PE for some ideal I by Corollary 3 ebove. Then I contains M, and the cancellation lew shows that the conteinment is strict: If I=M then RM= FM, hence R= P which is absurd. Thus I is strictly bigger than M and consequently I is a product of primes. But then so is M, contrary to assumption. Tt remains to show that representation is unique. Suppose P,P, «+ P, Q% +++ Q, where the P, and Q are primes, not necessarily distinct. Then Py PQ «++ Q,, implying that P, > some Q, (see the proof of Lemma 2 forTheoren 15). Rearrenging the Q, if necessary, we can assume that P, >Q, 3 then in fact P, = Q, by condition (2). Using the cancellation law we obtain Pose PL Q eee Q,+ Continuing in this way we eventually find that r= 6 and (after rearrengenent) P, = Q, forell i. O Combining Theorens 1: and 16 we obtain COROLLARY: ‘he ideals in a number ring factor uniquely into prime ideals. O ‘As an example of this, consider the principel ideals (2) and (3) in the number ring R= Z[f-5]. It is easily show that (@) = @ 1+)" (3) = 3, 1+) 53, 1- J 5) (verity this; note that the product (@,p)(¥,d) is generated by OY, BY, @d, pd +) Moreover ali ideals on the right are primes: This can be seen by Observing that |R/(2)| = 4, hence R/(2, 1+ ./-5) has order dividing }. The only possibility is 2 because (2, 1+,/-5) contains (2) properly and cannot be all of R (if it were, then ite square would also be R). This implies that in fact (2, 1+. J-5) is maximal as an additive subgroup, hence a maximal ideal, hence a prime. Similerly the factors of (3) are primes. This sheds some light on the example of non-unique factorization given in exercise 15, chapter 2: (2)(3) = (1+, -5)(1 -/-5)~ Verity that (2+ JS) = (2, 1+ JS)3, 147-5) and 2 ~ J-5) = (2, 1 + J5)3,1-J-5)« ‘Thus When all elements (more precisely, the corresponding principal ideals) are decomposed into prime ideals, the two factorizations of 6 become identical. In view of Theoren 16, we can define the greatest common divisor ged(I,J) and the least common miltiple Jem(I,J) for any two ideals, in an obvious way via their prime decompositions, The terms "greatest" and "least" actually have the opposite meaning here: Corollary 3 of Theorem 15 shows that "multiple" means sub- ideal and "divisor" means larger ideal; thus ged(I,J) is actually the smallest61 ideal containing both I and J, and lem(I,J) is the largest ideal contained in both. ‘Therefore we have gc(I,d) - 14a len(L3) = INT. Using this observation we can show that every ideal in a Dedekind domain is genereted as an ideal by et most two elements; in fact one of them can be chosen arbitrarily. THEOREM 17: Iet I be an ideal ina Dedekind domain R, and let o be ony nonzero element of I, Then there exists p €I such that I= (a,p). Proof: By the observation above, it is sufficient to construct p €R such ‘that I = ged((~),(p)). (Automatically p will be in Ij why?). AL fe a, Tet PB," Be, -++y B,* be the prime decomposition of I, where the P, are a, distinct. Then (a) 1s divisible ty all Pi. Tet Q),-..,@, denote the other primes (if any) which divide (7). We must construct p such that none of the 1, Q aiviae (p), and for each i, P;) is the exact power of P, diviaing (p). Equivalently, xr ( ny att) s ( ) € P.” - P, nn R- . pe A Ry 1PoY del = This cen be accomplished via the Chinese Remainder Theorem (see appendix 1): Fix ap nyel 6, EPS ~ Py (which is necessarily nonempty by unigue factorization) and let 6 satisfy the congruences ), Leder 6 #1 (moa %), d= Lee. (to show that such a 6 exists we have to show that the powers of the Pj and the Q, are pairwise co-maxinal: that the oun of any two is R. This is easy to€ verify if one interprets the sum as the greatest common divisor, Another way of seeing this ie given in exercise 7-) O We know that every principal ideal domain (PID) is e unique factorization domain (UFD) (see appendix 1). In general the converse is falee: Z{x] isa UFD but not e PID (exercise 8). However the converse is valid for Dedekind domains: THEOREM 18: A Dedekind domain is a UFD iff it is @ PID. Proof: As we have noted, PID always implies UFD; for Dedekind domains we can also get this result by using Theorem 16. Conversely, assuming that the Dedekind domain R is a UFD, let I be any ideal in R. By Theorem 15, I divides some principal ideal (a). The element a is a product of prime elements in R, and it is easily shown that each prime element p generates a principal prime ideal (p): If ab € (p), then plab, and then pla or p|b, implying that a or bis in (p). Thus I divides a product of principal prime ideals. By unique factorization of ideals in R, it follows that I is itself a product of principal primes and therefore a principal ideal. SPLITTING OF PRIMES IN EXTENSIONS We have seen examples of primes in 2% which are not irreducible in a larger number ring. For example 5 = (2+4i)(2-4) in Zf{i]. And although 2 and 3 are irreducible in Z[,-5], the corresponding principal ideals (2) and (3) are not prime daealss (2) = (2, 1+ JS) and (3) = (3,2 4 5)(3, 2-5) + This phenomenon is called splitting, Abusing notation slightly, we say that 3 splits into the product of’ two primes in ZLJ-5] (or in @L/-5], the ring being understood to be AN @[,/-5]). We will consider the problen of determining how a given prime splits in a given number ring. More generally, if P is any prime ideal in any nusber ring R=AMK, K a number field, and if L isa number field conteining K, we consider the prime decomposition of the ideal6 generated by P in the number ring S=ANL. (This ideal is PS = (Op, ++. FOP,20, EP, BLES}. If P is principal, P = (@), then PS is just oS = [ops p € 8} .) Until further notice, let K and L be number fields with KCL, and let R=ANK, S=ANL. The term "prime" will be used to mean “nonzero prime ideal." THEOREM 19: Iet P bea prime of R, Q a prime of S. ‘Then the following conditions are equivalent: (2) @lrs (2) qos (3) @>P (4) QNR=P (5) QnkeP. Broof: (1) # (2) by Corollary 3, Theoren 15; (2) # (3) trivially since Q is an ideal in S; (4) = (3) trivially, and (4) @ (5) since QCA. Finally, to show that (3) = (4), observe that QAR contains P and is easily seen to be an ideal in R; since P is a maximal ideal, we have QNR=P or R. If QNR=R, then 1€Q, implying Q=8, contradiction. O When conditions (1)-(5) hold, we will sey that Q lies over P, or P Les mder Q. "THEOREM 20: Every prime Q of § lies over a unique prime P of R; every prime P of R lies under at least one prime Q of 8. Proof: ‘he first part is clearly equivalent to showing that QNR isa prime in R. ‘his follows easily fron the definition of prime ideal and the observation thet 1¢Q. Fill in the details, using a norm argument to show that QQMR is nonzero. For the second part, the primes lying over P are the prime @ivisors of PS; thus we must show thet PS #8, so thet it has at least one prime divisor. Equivalently, we mst show 1¢ PS. (We know 1¢P, but whycan't 1= a6) +. +06, O EP, B, E82) To chow 1¢ PS, we invoke Lemma 2 for Theoret 15: There exists y€K-R such that YPCR. Then YS CRS=8. If 1€PS, then YES. Bat then Y is an algebraic integer, contradicting YEK-R. O As we have noted, the primes lying over a given P are the ones which dccur in the prime decomposition of PS. ‘The exponents with which they occur are called ‘the ramification indices, Thus if Q° is the exact power of Q dividing PS, then e is the ramification index of Q over P, denoted by e(Q|P). BMAMPIZ: Let R= Z, S= Z[i]s then the principal ideal (1-4) in 8 lies over 2 (we are writing 2 but we really mean 2%) and in fact (1-1) is a prime. (‘his can be seen by considering the order of S/(1- 1), as we did in the case of ZLy-51/(2, 1+ 5) after Theoren 16.) We have 28 = (1 - i), hence e((i-1)[2)=2. On the other hand e(Q|p) = 1 whenever p#2 and Q lies over p. More generally, if R= Z and $= Zlu], w= ef m= p" for some prime p € %, then the principal ideal (1-w) in § ise prine lying over p and e((1 - )[p) = g(m) = pp -1) (sce exercise 3b, chapter 2, and the renarks following the proof of Theorem 22.) On the other hand Where e(@la) = 1 vhenever gp and @ lies over q; this will follow from Theoren eh. There is encther important number associated with a pair of primes P end Q, Q lying over P. We know that the fector rings K/p and 8/Q are fields since P and Q are maximal ideals. Moreover there is an obvious way in which R/P can be viewed as a subfield of S/Q: ‘The containment of R in S induces a ring- homomorphism R+S/Q, and the kernel is RNQ. We know that RW Q=P (Theorem 19), so we obtain an embedding R/p»8/Q. These are called the residue fields essociated with P and Q. We know that they are finite fields (see the proof of Theorem U+), hence §/Q is an extension of finite degree over R/P; let f£ be the degree. Then f is called the inertial degree of Q over P, and isdenoted by £(Q|P) . EXAMPLES: Iet R= Z, S= Z[i]5 we have seen that the prime 2 in Z lies under the prime (1-4) in Z[i]. 8/28 has order 4, and (1 - i) properly contains 28; therefore |s/(1 - 1)| must be a proper divisor of 4, and the only possibility is 2. So R/P end 8/Q are both fields of order 2 in this case, hence f= 1. On the other hand 36 isa prime in S (by exercise 3, chapter 1, and the fact that S isa PTD), and |s/38] + So £(38]3) = 2. Notice that ¢ end f are multiplicative in towers: if PCQCU are prines in three mumber rings RCS CT, then e(u]P) = e(u]aye(a|P) 2(0|P) = 2(0]@)e(a]P) . We leave it to the reader to prove this (exercise 10). Tn general, if Q is any prime in any number ring 5, we know Q lies over a mique prine p€ Z. Then 6/Q isa field of order p', whore f= £(a[p)- We know that Q contains pS, hence p* is at most [s/sS|, which is xf, where n is the degree of L (the number field corresponding to 8) over Q. This gives the relation f Pye tt, First, fixing any « € P - H+, we neve an obvious isonarphism R/P + oR/OR . Next, the inclusion oR ¢ P* induces the homomorphism on > BRyge +t whose Kernel is (oR) NPS*} ana whose image is ((or) + Pt) /K+1, go prove what we want we must show that (or) nPS*+= oP ana (or) + ts. mis is easily done by considering these as the least common multiple and greatest common divisor of oR and P**+, ana noting that PS is the exact power of P dividing OR. (Convince yourself.) O Proof of Theorem 21, Special Case: We prove Theorem 21 for the case in which K=Q. Then P=p% for some prime p€ Z. We have hence TT go = Th co's" woot = TT tat = To. On the other hand we know that |[ps|| = p°. Thus the result is established in this special case. O Broof of 22(b): In view of 22(a), it is sufficient to prove this for the case in vhich I isa prime P; the generel result will then follow by factoring I into primes. Notice thet S/PS is a vector space over the field R/P. (Verify this; show thet in fact 8/PS is a ring containing R/P.) We claim that its dimension in n.68 First we show that the dimension is at most n. It will be sufficient to prove that any n+1 elements are linearly dependent. ‘Thus, fixing aq, ‘linearly dependent over R/P. ‘this is not as easy as it locks. We know of course “8,41 € 5, we must show that the corresponding elenents in /PS are that 04,+++0,,) ve Lineerly dependent over K, and it follows thet they are linearly dependent over R. (See exercise 25, chapter 2.) ‘Thus we have Py Fee * Pn ne Pa ad is to show that the , need not all be in P, so that ven we reduce moa P they = 0 for some fy, €R, not all 0. ‘he prob do not all become 0, For this we require the following generalization of Iemma 2 for Theorem 15: TEMA: Iet A and B be nonzero ideals in a Dedekind domain R, with BCA and AR. ‘Then there exists Y€K such that YBCR, YB¢A. Proof of the Temmat By Theorem 15 there is a nonzero ideal C such that BC is principal, say BC = (@). ‘Then BC ¢OA; fixany BEC such that pB SOA and set y= pf. Tt works. O ‘The lemma is applied with A= P and B= (Byy+++By 4a) « We leave it to the reader to fill in the details. ‘Thus we have esteblished the fact that 8/PS is at most n-Gimensional over R/P. To establish equality, let PN Z=p% and consider all primes P, of R dying over p. We know 8/P,8 is a vector space over R/P, of dimension n, 1 for some prime Q of S lying over P. (In other words, PS is not squarefree.) We heve seen that p is ramified in z[u] (w= eV", n= 3") and we claimed that no other primes of Z are ramified in Z[w]. We saw thet 2 end 3 are ramified in z[ 3/2] but 5 is not, and that 23 is ramified in zlo],2 where oF =: +1. Recall that the discriminants of these rings are, respectively, a power of p; -33+2°3 and -23. In general, a prine p€ % is ramified ina number ring R iff p|disc(R). We will prove one direction now, postponing the converse until chapter h. THEOREM 2: Let p bea prime in Z, and suppose p is ramified ine number ring R. Then p|dise(R). Proof: let P be © prime of R dying over p such that e(Plp) >1. Then pR= PI, with I divisible ty all primes of R lying over p. 1G, denote the enbeddings of K in © (where K is the number field corresponding to R) and, as usual, extend all oj to automorphisms of some extension L of K which is normal over @. Tet 04,...,0,, be any integral basis for R. We will replace one of the O, by a suitable clement which will ensble us to sce that p divides dise(R). In particular, teke any @ € I - pR (we know I properly contains pR); then @ is in every prime of R lying over p, but not in pR. If we write Gamo, +++. +mG,, m, € Z, then the fact thet @ ¢ pR implies that not all m, are @ivisible ty p. Rearranging the @, if necessary, we can assune that ptm. Set. = dise(R) = dise(ay, +++ ,0,) 5 then it is easy to see that Aise(\O,,.-- 4c) ama (see exercise 18 if necessary). Since ptm, it will be sufficient to show that pldisc(a,o,,--.,0,) « Recall that a is in every prime of R lying over p. It follows that a is in every prime of S=A(ML lying over p. (Bach such prime conteins p and intersects R in some prime; this prime of R contains p, hence lies over p,3 hence contains a.) Fixing any prime Q of 8 lying over p, we claim that in fact o(@) €@ for each automorphisn g of L: ‘To see this, notice that o (a) is a prime of oS) =S lying over p, and hence contains @. In particular, then, we have o,(@) €Q forall i. It follows that Q contains aise(4,a,,..+,0,)- Since the discriminant is necessarily in %, it is in QN Z= pH. That completes the proof. O If we work a little harder, we can obtain a stronger statement about the power of p dividing disc(R). See exercise 21. COROLLARY 1: Let @€R, K=@[o], andlet £ tbe any monic polynomial over % such that £(0) = If p isa prime such that ptw'rt(a), then p is unramified in K. (See exercise 21, chapter 2.) O ‘COROLLARY ring R. O Only finitely many primes of % are ramified in a nunber COROLLARY 3: Let R ana S be number rings, RCS. Then only finitely many primes of R are ramified in 8. Proof: If P isa prime of R which is ramified in 8, then PN Z=pZ is renifiea in 8 (recall that e is miltiplicative in towers). There are only finitely many possibilities for p, and each one lies under only finitely many primes of R. Hence there are only finitely many possibilities for P. O Of course p canbe ramified in S without P being ramified in 8. Exercise 19 provides a finer tool for showing primes of R are unramified in S. Better yet, there is a criterion for determining whether a particular prime Q of 8 is ramified over R (meaning that e(Q/p)>1, where P=QNMR),. There is @ special ideal in 8, called the different of 8 (with respect to R) which is divisible by exactly those primes @ which are ramifled over R. We will develop ‘the concept of the different in the exercises at the end of this chapter and provemh one direction of the ramification statement. The other will be proved in chapter J, We now consider in detail the way in which primes p€ Z split in quadratic fields. Iet R=AN@L/m], m squarefree. Recall that R has integral basis (1m) ena discriminant 4m when m=2 or 3 (mods), and integral basis (1, (2 +/m)/2} and discriminant m when m #1 (mod 4). Tet p be a prime in 2%. Theorem 21 shows that there are just three possibilities: i, t(e[p) = 2 pR=(P , £(B|p) = 2 PP, £(P,|p) = |p) = 1 - THEOREM 25: With notation as above, we have: rf pm, then eR = (pay. : a) If m 4s oda, then @,1+ JR) if m= 3 (moa) @) on =| (2, 24 0Bye, 1-8) se mea (moa 8) (3) prime if m=5 (moa 8). (4) If p is oda, ptm, then ia n+ Jap n- fm) if men’ (mod p) (6) prime if m is not a square mod p (6) and in (3) ana (5), the factors are distinct. Proof: For (1), we have (p,JH)° = (s°,p/im). ‘This is contained in pR since plm. 0n the other hand, it contains the ged of & and m, which is Ps hence it contains pR.ue) (2), (3), and (5) are all similar to (1), and we leave them to the reader. The distinctness of the factors in (3) and (5) follows from the fact that ptdaise(R) in those cases. Finally we prove (11) and (6): In each case it will be enough to show that if P ie any prime lying over p, thon R/P is not isomorphic to 2, (since then we will have £(P/p) = 2). Assuming first that p is oda, p/m, and m is not a square modp, consider the polynomial x°-m. This has a root in R, hence a root in R/P. But by assumption it has no root in ,. This shows that R/P and a, cannot be isomorphic which, as we have observed, implies (6). (4) is similar, using the polynomial 2 i-m x-xeee, We leave the details to the reader. Note that it makes sense to consider this polynomisl over R/p and Z, since 1-m is assumed to be divisible ty 4. O The prime ideals involved in these factorizations do not look like principel ideals, but we know that in certain cases they must be principal: for example when n= -1 or -3 (exercises 7 and 14, chapter 1). Can you describe principal generators for the various prime ideals in these two cases? To apply Theoren 25 for a given prime p it is necessary to te able to determine whether or not m is a square mod p. This of course can be done with the aid of Gauss' Quadratic Reciprocity Law, which we will establish in chapter 4 ‘vy comparing the way a prime splits in a quadratic field with the way it splits in cyclotomic field. We turn now to the letter problem tet = c°/™ ana fix a prime p€ %- Since Qu] isa normal extension of @, the corollary to Theorem 23 shows that we have BR = (QQ «++ @)° where the Q, are distinct primes of [uw] , all having the same inertial degree £ over p. Moreover we have ref = ¢{m).76 TMORM 26: Write m in the form pin, pin. ‘then (with notation as above) we have e = dp"), and f is the (miltiplicative) onder of p moan. (in other woras, p* = 1 (mod n) and f£ is the smallest positive integer with this property. ) Proof: Set a =u, paw. Then @ and p ere, respectively, oth and nth roots of 1. We will consider how p splits in each of the fleas ela] and Q[p]; the result for @[w] will then follow easily. When ptm we have k=0 and n=m, sothat @=1 and pay. Assuming that plm, we consider hov p splits in [a], which is the pitn cyclotomic field. We know that peut - a?) where u isa unit in Zo] (exercise jib, chapter 2), As we have noted before, this implies that pZ[o] is the o(p")th power of the principal ideal (1 - a), and since o(p") = [elo]:@] , Theoren 21 shows that this must be the prime fectorizetion of p Zlo]. Wow we consider what happens in @[p], which is the nth cyclotomic field. We know that p is unremified since pfn end aise( Z[p]) is a divisor of n%") (estentistea in chapter 2, after Theoren 8). ‘aus ve have pZlp] = PB + where the P, are distinct primes of 2[p], each with the same inertial degree £ over p, and rf=¢(n). (This rv and f will tum out to be the r and £ for the splitting of p in Q[w], but we don't know that yet.) We claim that f£ is the order of p modn. To establish this, recall first that the Galois group G of e[p] over @ is isomorphic tom, the multiplicative group of integers mod ns an auto morphism ¢ of @[p] corresponds to the congruence class &€ Z. (a € Z) ife In particular, let o denote the eutomorphism corresponding to p.Tt Tet (c) denote the subgroup of G generated by 3 thus (c) consists of the powers of go. ‘The order of the group (c) is the same as the order of the element co, Which is the same as the order of p modn. ‘Thus we have to show that (c) has order f. Fixing any P= Pj, we note that the field z[p]/P has degree f over Zs since that was the definition of f= f(P/p). Consequently the Galois group of uip]/P over 2, is cyclic of order f, generated by the automorphism + which sends every elenent to its pth power (see appendix 3). To prove what we want, it will be sufficient to show that o*=1 iff ;*=1, for every a € %. This will show that the cyclic groups (co) end (r) have the same order. Clearly we have o® =1 iff p? =p, and the latter holas iff p°=1 (moan). On the other hand, it is not hard to see that +*=1 iff a 2 BP =p (mod P). ‘Thus, assuming fp = 6 (mod P), we must show that a p’ =i (moan). Clearly we can write p°=b (moan), 11 then n€P; but this is clearly impossible since p€P and (n,p)=1. So bal. ‘This completes the proof that f(P/p) is the order of p modn, for each prime P lying over p in fp]. Finally we put together our results for Z[o] end Z[p]. Fix prines Qye++40, of lu] lying over Py,...,P,, respectively. (Theorem 20 shows that the Q, exist.) All @, Ie over pé€ Z, hence all Q; must lie over (1 - @) in @[@], since we showed thet (1-0) is the unique prime of [a] lying oN, (2 e((2 - ap) = os") £(Q|p) > 2(Py|p) = £- Moreover we have rf = g(n) by Theorem 21, and hence gp" )rf = gm). ‘Then Theoren 21, applied to the splitting of p in Z[u], shows thet the @, are the only primes of Z[w] lying over p and equality must hold in the inequalities ebove. ‘That completes the proof. O We restate Theorem 26 for the special case in which ptm: COROLLARY: If ptm, then p splits into gm)/f distinct prime ideals in lu], where f is the onder of p mdm. O We have not yet given a general procedure for determining how a given prime splits ine given mmber ring. Such a procedure exists, and it works almost all the time. Tt will explain in particular how we found the prime decompositions in the cubic flelds between Theorems 22 and 23. Tet R, 8, K, and L be as always, and let n= [L:k]. Fix an elenent @€S ‘of degree n over K, so that L=K[o]. In general R[@] is a subgroup (additive) of §, possibly proper. However the factor group 8/R[@] is necessarily finite. (One way to see this is to observe that § and R[a] are both free abelian groups of renk mn, where m= [K:@]; another way 1s to show that 8/R[o] is a finitely generated torsion group.) We will show that for all but finitely many primes P of R, the splitting of P in S canbe determined by factoring a certain polynomial mod P. Specifically, it will work whenever P lies over a prime p € % which does not divide the order of S/R[o] ; thus if S = R[@] it will work for all P. Fixing a prime P of R, we establish the following notation: for a polynomial h € R[x], let H denote the corresponding polynomial in (R/P)[x] obtained by reducing the coefficients of h mod P. Now let g be the monic irreducible polynomial for @ over K. ‘he79 coefficients of g ere algebraic integers (since they can be expressed in terns of the conjugates of the algebraic integer a), hence they are in ANK=R. ‘Thus g € R[x] and we can consider g € (R/P)[x]. fectors uniquely into monic ixreducible factors in (R/P)[x], and we can write this factorization in the forn e, = <1 =O b+ &, where the g, are monic polynomials over RK. It is assuned that the & are distinct. THEOREM 27: Let everything be as ebove, and assume also that p does not divide |8/R[@]|, where p is the prime of @% lying under P. ‘Then the prime decomposition of PS is given by e, r ee, OP BP oes G where Q, is the ideal (P,6,(@)) in S generated by P and g, (a); in other words, Q = BS + (a; (a))- Also, £(@,|P) is equal to the degree of 6, + Broof: et f, denote the degree of @,. This is the same as the degree of By. We will prove . f, (2) For each 4, either @,=58 or else S/Q, isa field of order |R/P| (2) +9; = 8 whenever afd; ee, e. (3) wit ae. ar. Assuming these for the moment, we show how the result follows: Rearranging the Q, . (tt wi if necessary, we assume that Q),+++,Q, £8, and Q. qr, turn out that r= 6s.) In any case, we find that @,.--,Q, are all prime ideale of S, end they obviously le over P since they contain P. ‘This also showsthat £(0,|P) f, for igs. (2) shows that Q>++-,Q, are distinct, and (3) e + @,° upon setting Og gyre = 8. Tt follows that the Oy ote 4, prime decomposition of PS is Q," @, -.- Q,, with a Se, for i e pecomes psla," a,” Applying Theorem 21, we obtain n= ajf; +... 4+ 4f,. On the other hand, n is the degree of g, which is easily seen tobe ef) +... te,f,. Tt follows that we must have r=6 end a =e, forall i. Taus it remains to prove (1), (2), and (3). Proof of (1): We look eround for a field of the desired order, and we find that Fy = (a/P)EXI/G,) is such a field. (See appendix 1.) In order to establish a connection between Ry and S/Q,, we observe that R[x] can be mapped homomorphically onto each: R[x] + F, de defined in the obvious way, reducing coefficients mod P and then reducing mod the ideal (&,). This is obviously onto and it is not hard to see that the kernel ie the ideal in R[x] generated ty P and g,: (P,8;) = Plx] + (g,) (see exercise 25). ‘Thus we have an isomorphism RIx]/(P,8;) > Fy + Now map R[x] into § ty replacing x with a; this induces a ring-homonorghism R[t] + $/Q,, and it is casy to sce that (P,g,) 8 contained in the kernel. ‘the isomorphism ebove shows that (P,g,) is e maximel ideal, so the keel is either (,6,) or ell of R[x]. Moreover R[x] is mapped onto 8/Q,: To prove this, we must show that 6 = RIC] + Q,. We know thet pEPCQ,, hence PSC. We claim that in fact S = R[o] + pS; this follows from the assumption that pt|s/Rlo]|. (the index of Ro] + pS in S$ is a como divisor of |s/xic]| and |8/p5| , and these ere relatively prime since |S/ps| is a power of p.) Tus R[x] +5/Q, is onto. ‘Taking into eccount the two possibilities for the Kemel, we conclude that either Q,=S or else S/Q, is isomorphic toa R[x]/(P,g,) , which is isomorphic to Fy. Proof of (2): Recall that the @, are distinct irreducible polynomials in ‘the principal ideal domain (R/P)[x] ; hence, given i # j, there exist polynomials h and k over R such that ‘This implies that gh + ake 1 (moa Plx]) 5 replacing x by Q, we obtain the congruence 6,(a)n(a) + 6 (a)k(a) = 2 (moa PS) « (Convince yourself that all of this is valid.) It follows that LE (P, g,(2), 8,(2)) = a + @, Proving (2). Proof of (3): fo simplify notation, set ¥, = e. e Tt is easy to see that the product @,"... @." is contained in, and hence (@). Ten Q, = (P,¥,)+ @ivieible by, the ideal eo ¢ e aie re (By Wy Yoo eee Ved e. e. This ideal is just PS. To prove this, we must show that the product » ee isin PS. We know that hence & (mod Plx]) « 1 £2 e By Be ote Be As in (2), this implies that e e ‘1 oe r a WY we Yy = g(a) = O (mod PS)end we are finished. O We note that the condition on p is satisfied, in particular, whenever L= Qa] and p-faisc(a). This is because |s/la]|? aivides |s/z[oj|* (which is finite in this case), and the latter number is a divisor of disc(a) (see exercise 27c, chapter 2). We give some applications of Theorem 27. Taking @ =m, we can re-obtain the results of Theorem 25 except when p=2 and m#1 (mod4); in this exceptional case the result can be obtained by taking a= (1+,/Jm)/2. As another example, we can determine how any prime splits in z[o], where sal, dy factoring the polynomial x2 -x-1 modp. Further examples are given in exercises 26 and 27. See exercises 29 and 30 for some interesting applications of Theoren 27. EXERCISES 1. Prove the equivalence of conditions (1), (1'), and (1") for a commtative ring R. (Hints: For (1) = (1'), consider the ideal generated by all 1,3 for (1') = (1"), construct an increasing sequence; for (1") = (1), consider the set of finitely generated sub=ideals of a given ideal.) 2. Prove that a finite integral domain is a field; in fact show that for each @ #0 we have o=1 for sone n, hence oM~+ is the inverse of o. 3. Let G bea free abelian group of rank n, with additive notation. Show that for any m€ Z, G/mG is the direct sun of n cyclic groups of order m. 4, Let K be a mmber field of degree n over @. Prove that every nonzero ideal I in R=ANMK is a free abelian group of rank n. (Hint: ORCICR for any MET, af0.) 5. Complete the proof of Lena 2 for Theorem 15.6. oa 8 a 10. 83 Fill in any missing details in the proof of Theorem 15. Show thet if I and J are ideals in a commutative ring such that 1€I+3, then 1€I" 49" forall mn. (Hint: Write l=a+p,a0€1,peéa and raise both sides to a sufficiently high power.) (a) Show that the ideal (2,x) in Z[x] is not principal. (b) Iet f£,g € Z[x] and let m and n be the god's of the coefficients of £ and g, respectively. Prove Gauss' Lemma: mn is the ged of the coefficients of fg. (Hint: Reduce to the case in which m=n=1 and argue as in the lemma for Theorem 1.) (c) Use (b) to show thet if f¢€ [x] and f is irreducible over Z, then f£ is irreducible over @. (je already knew this for monic polynomials. ) (a) Suppose f is irreducible over 2% and the gcd of its coefficients is 1. Show that if f|gh in [x], then fle or tlh. (Use (b) and (ec) .) (e) Show that [x] is 9 UFD, the irreducible elements being the polynomials £ as in (a), along with the primes pe Z. let K and L be number fields, KCL, R= ANK, S=ANL. (a) Iet I and J be ideals in R, and suppose Is|JS8. Show that IJ. (Suggestion: Factor I and J into primes in R and consider what happens in S .) (b) Show that for each ideal I in R, we have T=ISMR. (Set J=ISMR and use (a) .) (c) Characterize those ideals I of S such that I= (INR)S. Prove that e and f are multiplicative in towers, as indicated before ‘Theorem 21,13. Tet Vl (a) (e) (e) (a) (e) (v) let 8h K be a mumber field, R= ANK, I a nonzero ideal in R. Prove that aiviaes W(a) for all a €1, end equality holds iff I= (a). Verify that 98 = (5, a +2)(5, 0 + 30-1) inthe ring $= 2[F2], aaefe. Show that there ie @ ring-isonorphien BixI/5, © + 3x - 1) + By tx /OP +3m-). Show that there is a ring-homomorphism from alxI/(5, © + 3x -1) onto 8/5, P+ x -1). Conclude that either s/(5, 0 + 31-1) is a field of order 25 or clee ©, P+3x-3)=8. Snow that (5, oF + 32-1) #8 ty considering (2). Tet S= Bla], F=a+1. verity that 238 = (23, « - 10)°(e3, o - 3). Show that (23, @- 10, @= 3) =8; conclude that (23, a-10) and (23, @ ~ 3) are relatively prime ideals. K and L be number fields, KCL, R=ANK,S-ANL. Moreover assume that L is normal over K. Let G denote the Galois group of L over K. ‘Then |G| = [lek] =n. (a) (b) Suppose Q and Q" are two primes of 8 lying over a prime P of R. Show that the number of automorphisms go €@ such that o(a)=Q is the same as the number of o €G such that o(Q)=Q". (Use ‘heoren 23.) Conclude that this number is e(q|P)f(Q|P) . For an ideal I of 8, define the norm Ng(T) to be the ideal rn Tl oft). o€G Show that for a prime Q lying over P,we 15. () (c) (a) (e) (continued) XQ) = (Ql) | (Use exercise 9b.) Show thet for en ideal I of 8, off) = (a(T))8 ole ) = ages (Suggestion: First show thet the product hes the form JS for some ideal J of R3; then use exercise 9b.) Show thet HAC) = NEC) for ideals I and J in 8. (Suggestion: Use (c) and exercise 9b .) Tet 2€8, 040. Show that for the principal ideal (@) = as, W((@)) is the principel ideal in R generated by the element ny) 7 Parts (b) and (a) of exercise 1h suggest defining (I) for arbitrary extensions, not necessarily normal, by setting H(Q) = pf(alp) for primes Q, and extending multiplicatively to all ideals. ‘his is consistent with the other definition in the normal case. (a) () (ec) Show that for three fields KCLCM, We(I) = Netly 2) for an ideal I in ANM. Show that the result in exercise lhe is still true in the general case. (Hint: Let M be the normal closure of L over K.) In the special case K=@, show that N{(I) is the principal ideal in @ generated by the number |[Il]. (Suggestion: Prove it first for1b. 16. My. 86 (c) (continued) prime ideals and use Theorem 22a .) let K and L be number fields, KCL, R=ANK, S=ANML. Denote by G(R) and G(S) the ideal class groups of R and 8S, respectively. (See chapter 1.) (a) Show thet there 1s 8 homomorghism G(S) + G(R) defined by teking any I in a given class © and sending C to the class containing Ny(I). (Why is this well-defined?) (b) Tet Q be a prime of S lying over a prime P or R. Tet a) denote ‘the order of the class containing @ in G(8), 4, the order of the clase containing P in G(R). Prove that ap] age(Q]P) let K= QL/23], b= Qo), where w= c@/3 We iow (exercise 8, chapter 2) that KCL. let P be one of the primes of R=ANK lying over 2; specifically, take p= (2,6) where @= (1+,/-23)/2 (see Theorem 25), Iet Q be a prime of Z[w] lying over P. (a) Show that £(Q|p) =11. (Use Theorems 25 and 26 and the fact that f is multiplicative in towers.) Conclude that in fact Q = (2,9) in Zw] - (b) show that P= (9-2), but that P is not principal in R. (Hints Use Theorem 22(¢) to show that P is not principal.) (c) Show that @ is not principal. (Use 16b .) (a) Show that if 2=08, with a,pé€ Zlw], then @ or 6 is a unit in Zw]. (See the proof of Theorem 18 if necessary.) Iet K be a number ficld of degree n over @, and let O,...,0, EK. (a) Show that dise( roy ,05,+++,0),) = Paise(ay,-+.,0,) for all reQ.18. (continued) (b) Let p be a linear combination of Oy, +450, with coefficients in Q. Show that disc( + P,O5)+++,,) = dise(ay +4450) « 19, let K and L be number fields, KCL, andlet R=ANK, S=ANL. Let P be a prime of Re (a) Show that if a €S, BER, and OB € PS, theneither AEP or PEP. (Recall that 8/PS is a vector space over R/P. Also give a more straightforward proof using the fact that P is a maximal ideal.) (b) Tet 0,04,+++,0, € 85 BB rereshy ER} and oF PS. Suppose OB = 0,8, +++. +0,8,+ Prove that there exists Y €K such that py and ell of the p,Y are in R, andthe p,Y (i =1,...,n) are not all in P. (Hint: See the proof of Theorem 22(b) .) (c) Prove the following generalization of Theoren 2h: Tet 4,...,@, be 8 basis for L over K consisting entirely of members of 8, and let P be aprine of R which is remified in S. then aiscy(a,,...,0,)€P. (See exercise 23, chapter 2, for the definition and properties of the relative discriminant. ) 20. Tet K, L, R, and S be ae usual, and fixaprime P of R. We know (see the proof of Theorem 22(b)) that S/PS is an n-dimensional vector space over R/P. Call a set of elements of 8 independent mod P iff the corresponding elements in 8/PS are linearly independent over R/P + For each prime Q, of S lying over P, fix a subset B, CS corresponding to a basis for 8/Q, over R/P. (‘thus B, contains £, = £(Qj|P) elements.) For each i=1,...,x end for each j=41,...,e, (where e; = e(Q,|P)) fix an element ay, (> - Gy 06 9,8: ‘The Chinese Remainder hy ‘Theorem shows that such an element exists. Consider the n= Ee,f, clenents a6, 6 €B, 1) Suppose p°|m. Writing m= hi’ as in exercise 41, chapter 2, set v=cP/k. Show that p does not divide |R/ZLv]| ; use this to determine the prime decomposition of ER. (c) Determine the prime decomposition of 3R when m #41 (mod 9). (a) Determine the prime decomposition of 3R when m=10, (Hint: Set B= (w@~1)°/3 and use exercise 18 to show that disc(p) = 4 aisc(R) . Also note exercise 1d, chapter 2.) Show that this always works for m1 (mod 9) except possibly when m = 4 6 (mod 27). (e) Show that 9+4dise(R) when m= 41 (mod 9); use this to show that 3R is not the cube of a prime ideal. (See exercise 21.) Assuming the converse of Theoren 2l, show that 3R= Q where P and Q are distinct primes of R. Tet © =5(4+1), R= ANAC]. Tet pf¥3 be aprine of Z. Show that the prime decomposition of pR can be determined ty factoring 2 -5x-5 modp. Doitfor p=2. (See exercise 43, chapter 2.)28. Tet fx) =P ee pete + Fa, all a, € Z, endict p bea prime divisor of 99. Iet p* be the exact pover of p dividing og, and suppose all a, ere divisible ty p’. Ascume moreover that f is irreduc- ible over Q (which is automatic if r=1) andlet a be a root of f. Tet K= Q[o], Re ANK. (a) Prove thet (p”) = p"R is the nth power of an ideal in R. (Hint: First show that o” = p%p, with (p) relatively prime to (p) .) (b) Show that if v is relatively prime to n, then (p) is the nth power cf an ideal in R. Conclude that in this case p is totally remifica in R. (c) Show that if rv is relatively prime to n, then disc(R) is divisitie ty gt. (See exercise 21.) What can you prove if (nr) =m>1? (a) Prove that as = 4, = 1 in exercises 43e and Lie, chapter 2. Tet @ be an algebraic integer and let f be the monic irreducible’ polynomial for @ over Z. Tet R=AN Ql] and suppose p is a prime in @ such that f has arcot r in 2%, and p+|R/ z[a)}| - (a) Prove that there is e ring homoncrphiem R > Z, such thet a goes to vr. (Suggestion: Use Theorem 27.) (bv) Tep oF zae1. Use (a) to show that JO¢elo]. (Hint: r=2. Find a suiteble p. See exercise 26, chapter 2.) (c) With @ as in part (b), show thet 3G ana JO-2 are not in ela]. (Suggestion: fry various values of r with small absolute value.) (a) tet o 420-2. Prove that the equation tayhadtea hes no solution in AN @[a]. (See exercise 43, chapter 2. It's easy if you pick the right r.)30. 31. (a) (») (e) (a) -fe) Tet m Tet f be any nonconstant polynomial over Z. Prove that f has a root mod p for infinitely many primes p. (Suggestion: Prove this first under the essumption £(0) =1 ty considering prime divisors of the numbers f(ni). ‘Then reduce to this case by setting a(x) = £(x8(0))/#(0) «) Iet K be ony number field. Prove that there are infinitely many primes P in K such that £(P|p)=1, where p is the prime of Z% lying under P. Prove that for each m€ Z% there are infinitely many primes pl (mod m). Tet K and L be muber fields, KCL. Prove that infinitely many primes of K split completely (split into [L:K] distinct factors) in L. (Hint: Apply (b) to the normal closure of L over K.) Tet f be a nonconstant monic irreducible polynomial over e number ring R. Prove that f splits into linear factors mod P for infinitely many primes P of R. R be a Dedekind domain, K its field of fractions. A fractional ideal of K is a set of the form of, for some @€K and some ideal I of R. We will assume moreover that @ and I are nonzero. (2) () (e) Define the product of two fractional ideals ty the formla (@1)(p5) = OpIy. Show that this is independent of the representation of the factors. Iet I bea fractional ideal in K; define Ls fe ek: areR). 1 Prove that II” =R. (See the proof of Theorem 15.) Conclude that the fractional ideals of K form a group under multiplication. Show that every fractional ideal of K is uniquely representable as a31. (c) (continuea) product, pt ee wee er where the P, are distinct prime ideals of R and the mj are in 2. In other words, the fractional ideals of K form a free abelian group. It follows that every subgroup is free abelian. (We have only proved this in the case of finite rank, but it is true in general.) In particular this shows that the group of principal fractional ideals OR (@ €K) is free abelian. How is this group related to the mltiplicative group of K? (a) A free ebelian semigroup is any semigroup which is isomorphic to a direct sum of copies of the non-negative integers. Theorem 16 shows that the non-zero ideals of R form a free abelian semigroup under multiplication. Show thet the nonzero principal ideals form a free abelian semigroup iff R isa PID. (e) Show that the ideal class group of R (as defined in chapter 1) is isomorphic to the factor group G/H, where G is the group of fractional ideals of K and H is the subgroup consisting of the ~ principal ones. (£) Considering K as an R-module, show that every fractional ideal is a finitely generated submodule. Conversely, show that every nonzero finitely generated submodule of K is a fractional ideal. 32. Tet K be a number field, R=AMK, andlet I be an ideal of R. Show that |R/t| = |3/tJ| for 011 fractional ideals J. (First prove it for ideals J by using Theoren 22(e), then generalize.) 33-39: K and L are number fields, KCL, R=ANK S=ANL. 33. Iet A be an additive subgroup of L. Define “1 AY = (@ EL: aac 8} Abs (2 EL: mi(oa) cR33. 93 (continued) (a) (e) (ec) We can consider 1 as an Remodule, and elso as en S-nodule. Show that ATT 4s an S-submodule of L and A¥ is an Resubmodule. (In other words, SA a> ana RAY = AY.) Also shoy ATCA. Show thet A isa fractional ideal ifr SA= A and At ¥ (0). Define the aifferent diff A to te (A*)">. Prove the following sequence of statements, in which A and B represent subgroups of L and I is 9 fractional ideol: AcBeA% a5, mops areacat+; ri-lor, aife Tc; @iff I is 2 fractional ideal ; AcIs diff A isa fractional ideal ; I* is an S-submodule of 15 Tc (aiee 1); I* 4s © fractional ideal 5 im ce* ond prc 5 ID* = 8% 5 TAD 2 8% 5 De = 15 aiff I=. difrs. Die Leb (04,+-+,0,} be a basie for L over K. (a) Prove that there exist P,,+++)P, €L such that Tl) el if i=, 0 otherwise, (Hint: Recall that the determinant [aK] is nonzero, hence the corresponding matrix is invertible over K.) Show ‘that (6),+++,8,) is another basis for L over K. (This is called ‘the dual pasis to (0,,++.,0,} +)3h. 35 (continued) (b) Tet A= Ro, @... © Ro, the (free) R-module generated by the a, . Show that A¥ = B, where B is the Rumodule generated by the p; . (Hint: Given y GAY, obtain BEB ouch that T((y-p)A)=0, and show that this implies Y= p.) Suppose GEL, L= Ka]. Let f be the monic irreducible polynomial for @ over K, andwrite f(x) = (x~c)g(x). ‘then we have nal B(x) = ot YX Hee HH Ge for some Yo)+++5%,_y € E+ We claim that %o Yna2. Pray? ERG) 4s the dual basis to (1,0,...,0°74). (a) let Oy, +++,0, be the embeddings of L in € fixing K pointwise. Then the o,(@) are the roots of f. Show that £(x) = (x ~ 0,8, (x) where 8, (x) is the polynomial obtained from g by epplying o, to ell coefficients, and oy = o,(0) i o aif if (b) Show that 8;(05) = a TT (See exercise 20, chapter 2.) (c) let M be the matrix to 1) Girere 4 denotes the row nurber, 3 the column number), and let N be the matrix fox(¥5 _,/#))1 + Show thet Mt is the identity matrix. Tt follows that MN is also the Adentity matrix (why?). What does this show? (a) Show that if © €S then the R-module generated by YorerYp iy te R[o]. (Hint: Multiply (x - a)g(x) .) (ec) Prove that (Rfa])* = (£"(0)) nfo] af aes.35+ (continued) (2) Prove that aiff Rfe] = 2'(@)s if aes. (g) Prove that if @€8, then f(a) € difrs. 36. let P be aprime of R, Q a prime of S lying over P. Let %, (See exercise 20.) are independent mod Pe +50, €8, where n= [ItK]. Suppose OQ), (a) ‘Show that @,,-++,0, forma basis for L over K. (Suggestion: Use ‘the lemma in the proof of Theorem 22b.) (b) Iet A be the free Rmodule RO, @+.. © Ro,. Show that PSN A= PA. (c) Let I = Ann,(8/a) = (r € Bz x8 cA} (this is clearly an ideal in R); show that T¢P. (itint: show TcP#IScMPlrcr.) (a) Show that aifr | aire A| (1S) aiff 5; conclude that the exact power of Q in aiff S 4s the same as that in aiff A. (e) Suppose e(Q|P) = [L:K]. Fixing any 1 EQ - &, show that the exect power of Q@ in diff S is the same as that in £(1)S, where f is the monic ixreducible polynomial for ff over K. 31. Note that diff S is an ideal in 8 depending on R. We will show that if P is aprime of R and Q is aprime of S lying over P, then diff S is divisible by @°71, where e = e(Q|P). (a) Wetting Ps = o° "11, show that P contains (I). (See the proof of ‘Theoren 2.) (v) ret Pt denote the inverse of P in K (notin L). Show that pls = (re). (ec) Show that (ps) co, (a) show that & “aires. (ce) show that @ "rr(a)s for any a €S8, where f 4s the monic37+ 36. 39; ko. (e) (continued) irreducible polynomial for @ over R. Tet M bea subfield of K, and set T=ANM. Tet aife(s|T), aire(s|R) and aiff(R]T) denote the differents corresponding to each of the pairs of number rings. (‘hus the first two ave ideals in 8, the third an ideal in R.) Prove that the different is multiplicative in towers in the sense that aife(s|t) = aize(s|R)(aire(R])s) (Suggestion: Employing the obvious notation, show that 87> 8_(R8) and sy (aize(R|2)s) csp by using the transitivity property of the trace.) Now consider the ground field to be @, so that R¥ = (a €K: Tor) c Zz} and diff R= aize(R|Z) = (R")"+. tis is called the absolute different of R.) (a) Tet (0,,+.+,0,] be an integrel besis for R and let {Byy+++56y) be ‘the dual basis. ‘Then (f,,-++56,) is basis for R* over Z. (see exercise 3.) Show that disc(a,,...,,)dise(p,,.-.,6,) = 1. (Hint: Consider the natrix product [o,(0;)]fo,(6)] -) (b) Show that |R*/r| = |@isc(k)]. (Hint: Write the o, in tems of the fy + If necessary, see exercise 27, chapter 2.) (c) Prove that |laitz Rl] = |disc R]. (See exercise 32.) (a) Give a new proof that disc R 4s divisible ty oY, k= zle,- If, as shown in exercise 21. (See exercise 37+) (e) Prove that disc(S) is aivisitie ty aisc(r)™*], (set m= q@ in exercise 36 and use Theorem 22.) Compare this with exercise 23, chapter 2. Tet p bee prime, r>1.40. (continued) (a) Show that g(p")>r+41 when p>3, ana 2") >r. r vr, (v) tet w= e/P" | ghow that dise(y) is divictble by pXP)-2, Conclude that p*] 2 disc(w) « (c) Tet R be a mmber ring and suppose R contains eM, nem, Prove that m[2 disc(R) .Chapter 4 Galois theory applied to prime decomposition Up to now the Gelois-theoretic aspects of mumber fields have not figured prominently in our theory. Essentially all we did was to determine the Galois group of the mth cyclotomic field (it was the multiplicative group of integers mod m) and to show that, in the case of a normal extension, the Galois group permutes the primes over a given prime transitively (‘Theorem 23). Galois groups also turned up in the proof of Theorem 26 on splitting in cyclotomic fields. In this chapter we apply Gelois theory to the general problem of determining how a prime ideal of a number ring splits in an extension field, Tet K end L be number fields, and assume that L is a normal extension cf K. ‘Thus the Galois group G, consisting of all automorphisms of 1 which fix K pointwise, has order n= [L:K]. As usual we let R and S denote the corresponding number rings. Fixing a prine P of R, recall that ell primes @ of S lying over P have the sane ramification index e and inertial degree f (corollary to Theoren 23). ‘Thus if there are r such primes Q, then ref =n (Theorem 21). For each prime Q lying over P, we define two subgroups of G: ‘The decomposition group: D = D(Q|P) = (co EG: o@= a). ‘The inertie group:9 E = R(Q{P) = (o €Gs ofa) = a(moaQ) vaEs)}. Tt is clear that these are actually subgroups of G, and that ECD. (the condition o@=Q can be expressed as o(a) = 0 (mod Q) iff @ = 0 (mod Q) 3 obviously the condition for E implies this.) ‘The members of D induce automorphisms of the field 8/Q in a naturel way: Every o €G restricts to an automorphism of §, and if g€D then the induced mapping S+S/Q has kernel Q; thus each o €D induces an automorphism G of 8/Q, in such 8 way that this diegran commutes: 8 ——2+8 oe 8/Q—S-+8/Q - Moreover it is clear that G fixes the subfield R/P pointwise since co fixes K, hence R, pointwise. ‘Thus G is a member of the Galois group @ of 8/Q over R/P. All of this can be sumed up by saying that we have a mapping D+@, and it is easy to see that is is a group homomorphism: composition of eutomorphisms in D corresponds to composition in @. ‘the kernel of the homomorphism D> @ is easily seen to be E} this shows that E is a normal subgroup of D and that the factor group D/E is embedded in @. We will see that D>G is actually onto, hence D/E>@ is actually a group isomorphism. We know the structure of @: It is cyclic of order f£ (see appendix 3), hence the same is true for D/E. Now look at the fixed fields of D and E; denote them by L, and ly, respectively. L, is called the decomposition field and 1, the inertia field. In general, we adopt the following system of notation: For any subgroup H of G, Ty denotes the fixed field of H; thus Ij) =L and ],=K. More generelly, for any subset XCL, let X, denote XNL,. Thus §, is the number ring in Ty, and Q, is the unique prime of §, lying mder Q. Obviously @, les over P, and it follows easily from the way we have defined things that S,/Q, is an intermediate field between 8/Q and R/P. (Verify ell of this.) We can now state the mein result:100 THEOREM 28: Tet K, L, R, 8, P, Q, G, D, Br, e, end f be as above. ‘hen we have the following: xemification inertial Segrees, L Q —indices__ —Segrees, e | e 1 ty & ft | | 1 ft ty % r 1 1 K P Eroof: We begin by showing that [L):K] =r. By Galois theory we know that {L):K] 4s the same as the index of D in G. Bach left coset oD (c €G) sends Q to oQ (i.e., each member of the coset does this to Q), and it is clear that oD= 1D iff oQ=1Q. ‘This establishes a one-to-one correspondence between thé left cosets cD and the primes o@3 as shown in Theorem 23, these primes include all primes of S ying over P, hence there are r of them. That proves what We want. Next we show e(@,|P) and £(@,|P) are both 1. Notice first that Q is the only prime of S lying over Q), since such prines are necessarily permuted transitively by the Galois group of I over 1, (L is automatically a normal extension of 1,)3 this Galois group is D, which doesn't send Q to anything __ else, It follows by Theorem 21 that [hat] = e(a]a,)£(@]a,) » ‘The number on the left is ef since we heve already shown thet [L):K] =r and we know ref =n, Moreover the individual factors on the right cannot exceed e end £, respectively; consequently equality must hold in both cases and we must also have10. e(@,]P) = £(@,1P) = Next we prove that £(@JQ,) = 1. Equivalently, 5/Q is the trivisl extension of §,/Q,+ It will be sufficient to show that the Galois group of 8/@ over 8,/Q, is trivial (see appendix 3). Zo do this, we will show that for each 6 €8/Q, the polynomial (x - @)" has coefficients in S,/Q, for some m>13 it will follow that every member of the Galois group sends @ to another root of (x - 0)", which can only be @+ ‘That will prove what we want. Fix any a €S corresponding to € €8/Q; clearly the polynomial a(x) = Tl (x - ca) o€E has coefficients in §,; reducing coefficients mod Q, we find that E € (S/Q)[x] actually has coefficients in 55 /%q - But all o(a) reduce to @ mod Q (why?), hence g(x) = (x - 6)", were m= |B]. ‘that completes the proof that (Qla,) = 2+ Together with £(@)|P) = 1, this shows that £(@,19,) = 2(a|P) =f. Toen vy Theorem 21 we mst have [I,:L)] >. But we have seen (remarks before the theorem) that E is a normal subgroup of D and‘the factor group D/E is embedded in G, which is © group of order f. ‘hus [I,:1)] = |p/el < £, hence exactly £. ‘hen (‘Theorem 21 again) e(@,|@,) = 1. Finally we easily obtain [L:1,] =e and e(Q|9,) =e by considering the degrees and remification indices elready established. Cl COROLLARY 1: D is © mapped onto @ by the naturel mapping ot G3 the kernel is B, hence D/E is cyclic of order f. Proof: We have already seen that D/E is embedded in G. Moreover both {gt}. oO groups have order f, since |D/E The following special case indicates a reason for the terms "decomposition field" and “inertia field.”102 COROLLARY 2: Suppose D is 9 normal subgroup of G. ‘hen P splits into ¥ distinct primes in L). If & 4s also norm in G, then each of then remains prime (is "inert") in I. Finally, each one becomes an eth power in Lb. Broof: If D 4s normal in G, then ty Gelois theory L, is @ normal extension of K. We know that a has remification index and inertial degree 1 over P, hence so dees every prime P' in 1, lying over P (corollary to Theorem 23). ‘Then there must be exactly r such primes (Theorem 21). It follows that there are exactly r primes in I, lying over P since this is true in both L, and L. ‘his implies that each P' lies under a unique prime P" in i 3 however it seems conceivable that P" might be ramified over P'. If E is normal in G (so that I, 1s normal over x), then e(P"|P) = e(@,|P) =1, hence e(p"|P') = 1. This proves that P' is inert in Ig? P" = P'S, . Finally, we Jeave it to the reader to show that P" becomes an eth power in L. (1 We have already seen an exemple of this phenomenon: the prime 2 in % splits into two distinct primes in @[,/23], and each remains prime in @Lul , w= 4/3 (exercise 17, chapter 3). This could have been predicted by Corollary 2, Theorem 26 shows that 2 splits into two primes in @w], hence the decon- position field hes degree 2 over Q; moreover there is only one quadratic subfield of Q@{w] since the Galois group is cyclic of order 22. So the decomposition field must be @[/-23]. Finally, 2 is unramified in Q@[w] so the inertia field is ell of @fw]. Slightly nore generally, whenever L is normal over K with cyclic Galois group end P (a prime in K) epllite into r primes in L, then the decomposition field is the unique intermediate field of degree r over K, and P splits into © primes in every intermediate field containing the decomposition field. As another exemple, consider the field L = @[i, 72, 5]; this is normal of “degree 8 over @, and the Gelois group is the direct sum of three cyclic groups of order 2. ‘the prime 5 splits into two primes in @[i], is inert in Q[J2],103 and becomes a square in Q[/5]+ Consequently L must contain at least two primes lying over 5, and each must have remification index and inertial degree at least 23 it follows that each of these numbers is exactly 2. ‘he inertia field must be a field of degree 4 over Q in which 5 is unremified. ‘The only choice is eli, J2]. tus (244) and (2-4) remain prime in @fi, 2] and become squares of primes in L. Here is a nonsbelian exemple: Let = @[YI9, wu) were w= ™/3, men L is normal of degree 6 over @ with Galois group 83 (the permutation group on three objects). Consider how the prime 3 splits: It becomes a square in Q[w] , and hes the form ¥q in @[FI9] by exercise 26, chapter 3. Consequently 1 nust contain at least two primes lying over 3, and each must have renification index divisible by 2. ‘he only possibility is for L to contein three primes over 3, each with e=2 and f=1. For each of these primes, the decomposition fiela has degree 3 over ©. ‘here are three such fields: eff i5], ews iI0], and ata 7m . Any one of them can be the decomposition field, depending on which prime over 3 is being considered. (‘he fact that they ell can occur is easily seen ty the fact that each of them can be sent to any other one by an automorphism of L.) ‘he inertia field is the seme since f=1. Notice thet 3 does not split into three distinct primes in any of the possible decomposition flelds since in fact it is ramified in each (it mplits into FQ in one of them, hence in ell since all are isomorphic extensions of @). ‘this shows that the normality condition on D was actually necessary in Corollary 2. We now consider a variation on the situation. What happens if K is replaced vy a larger subficld K' of L? We know that K' 4s the fixed field of some subgroup HCG; in our previous notation, K'= 1. Moreover the ring R'=ANK is G,, and Pt=QNR! is the uigue prime of R! lying under Q. P' also lies over P, but it need not be unique in that respect. We know that 1 is a normal extension of K', so the decomposition and inertia groups (Q|P') and E(QJP) can be considered. From the definition, we immediately find that p(qjP') = DAH x(g|P") = BNE10% where D and EB are as before (for Q over P). ‘Then by Galois theory, the fields 1,K' and I,K" are the decomposition and inertia fields, respectively, for Q over P'. We use this observation to establish certain maximal and minimal conditions for decomposition and inertia fields. THEOREM 29: With ell notation as above, (1) L, is the de2gest intermediete field K' such that e(P'|P) = 2(P']P) = 135 (2) 1 is the smallest K' such thet @ is the only prime of 8 dying over P'; (3) 1, is the Jergest K' such thet e(Pt|P) = 15 (4) 1, is the smallest K' such that Q is totelly ramified over Pt (dee, e(Q[P*) = [LeK']) . Broof: Notice first that we have elresdy shown thet I) end 1, heve these qroperties: For exemple we showed in the proof of Thecren 26 that @ is the only wrine of 8 lying over @,; this could also be recovered from the fect thet e(ala,)e(@]a,) = ef = [L:t,] . Suppose now that K'= 1, is any intermediate ficld in which @ is the only prime lying over P'. We know that every ¢ €H sends Q to another prime lying over Pt, sowe must have HCD. ‘his implies L) CK’, esteblishing (2). his result could also have been obtained by considering the @iagrem at the right, in which the indicatea degrees have been obtained by applying Theorem ap ts rt 26 to both situations (@ over P and Q over P'). Here e', £%, and r* are the numbers i associated with the splitting of P' in the eet normal extension L. ‘Thus r' is the number of primes lying over P'. ‘he diagram shows that if r'= 1, then K''= I,K", hence 1) CK". x105 Next assume that e(P'|P) = £(p'|P)= 1. ‘hen e=e' and f£=f' by nultiplicativity in towers. Considering the disgram, we find thet I) and I,K" oth have the same index in L, Since one is contained in the other they must be equal, implying K' CL). ‘Thus we obtein (1). (3) is similar: If e(P'|P) =1, then e = Kel. Finally, if Q is totally ramified over pt then [L:K'] = » henee I, = I,K", hence + Considering the diagram, we find that K'~= I,K", hence I, cK". 0 This theorem has some interesting consequences. We will use it to prove the Quadratic Reciprocity Lew. It will be helpful to introduce the following concept: Aprime P in anumber field K splits completely in an extension fiela F iff P splits into [F:K] distinct primes, in which case all must have e and f=1 ‘by Theorem 21. Conversely, if all primes of F lying over P have e and f=1, then P splits completely in F (again by Theorem 21). It follows that if a prime splits completely in an extension F of K then it also splits completely in every sub-extension. Combining this observation with (1) of ‘Theorem 29, we obtain COROLLARY: If D is a normal subgroup of G (for some Q lying over P) then P splits comletely in K' iff K'CI). Proof: If P splits completely in K’, then in particular, e(P'|P) = e(pt|P) = 1, where Pt =QNR'. Then Kt CL, by (1). Conversely, Corollary 2 of Theorem 26 shows that P splits completely in 1, and hence also in any K, KewteL. O This will be applied in a situation in which G is abelian, so that all sub- groups of G are normal. Tet p be an odd prime in Z. For né€ Z, p-+n, define the Legendre syuibol. 1 if n is a square mod p -1 otherwise .106 ‘The Quadratic Reciprocity Law states that @-{* if p =41 (moa 8) P “(i if p = +3 (moa 8) and for odd prines q fp we have ® if p or qi (moa k) @- -@) if peas (woah). We will establish a criterion for a prime to be a dth power mod p, for any @ivisor @ of p-1. All of the action tekes place inside the cyclotomic fleld ful, w= 4/2, we iow thet the Galois croup G of @{u] over @ is cyclic of order p-1, hence there is a unique subfield F,c@[w] having degree a over @, for each divisor @ of p-1. (Ff, is the fined field of the unique subgroup of G having order (p -1)/a.) Moreover Fa oe, art ala - ‘THEOREM 30: Let p be an odd prime, and let q be any prim fp. Fix a divisor a of p-1. ‘hen q isa dth power modp iff q splits completely in Fa + Proof: We kmow that q splits into r distinct primes in @[w] , where f= (p-1)/r is the order of q in the multiplicative group (1,...,p-1) mod p. This is a cyclic group of order p-1, so the dth powers form the unique subgroup of order (p-1)/d, consisting of all elements whose orders divide (p - 1)/a. (Be sure you believe this. ‘That's all the group theory we need.) ‘Thus the following are ell equivalent: q is a dth power mod p t\(p - D/a alr Fy c RK . Finally we observe that F,, is the decomposition field for Q over q, for any107 prime Q of Z[w] lying over q. (his is because the decomposition field must have degree r over Q, and F., is the only one.) Thus the condition Fy ¢F,, is equivalent to q splitting completely in F,, ty the corollary to ‘Theorem 29. 0 COROLLARY: ‘THE QUADRATIC RECIPROCITY LAW (above). Proof: ® =1 iff q splits completely in Fj. What is E,? We recall (exercise 8, chapter 2) that Q{w] contains @fJip], with the + sign iff pel (moah). So this mst be Ey+ ‘the result then follows from Theorem 253 we leave it to the reader to check the details. See exercise 3. 0 ‘Theoren 29 can also be used to establish the following result, in which there is no normality essumption: THEOREM 31: Let K be a number field, and let L and M be two extensions of K. Fixaprime P of K. If P is umramified in toth L ana M, then P is unremified in the composite field LM. If P splits completely in both L and M, then P splits completely in IM. Proof: Assuming first that P is unremified in L and M, let P’ be any prime in LM lying over P. We have to show that e(p'|p)=1. Iet F be any normal extension of K containing LM, and let Q be any prime of F lying over Pt. (Sucha Q exists by Theorem 20.) Q also lies over P; let E = E(Q|P) ‘ve the corresponding inertia group, so that Fy is the inertia field. Theoren 29 shows that contains both L ani M, since the primes QML and QNM are necessarily unramified over P. ‘hen F, also conteins LM, implying that QNLM= Pt is unremified over P. ‘The proof for splitting completely is exactly the same, except E is replaced by D. We leave it to the reader to check this. Recall that splitting completely in LM is equivalent to the condition e(P'|P) = £(P'|P) =1 for every prime P* of IM lying ovr P. O