Scribd
Upload
427 views287 pages

Marcus Number Fields

Uploaded by

Đào Quang Đức
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF or read online on Scribd
427 views287 pages

Marcus Number Fields

Uploaded by

Đào Quang Đức
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF or read online on Scribd
You are on page 1/ 287
Daniel A. Marcus Number Fields 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 Fythagoreen triples, we can show thet Fermat was right for n end hence (automatically) also for any multiple of 4. (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 amiltiple of p. ‘Thus the problem is to show that if p is an odd prime, then x? +yP = 2? has no solution in nonzero integers x, y, 2+ Suppose, for some odd prime p, there is a solution x,y,z € 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 x°+y3= 23 has no case 1 solutions: If x, y, and 2 are not miltiples of 3, then in fact x+y) #23 (moa 9) since each of these cubes is =41 (mcd 9). Now assume p> 3; x, y, and z are not miltiples of p; and x+y? =P. Factoring the left side, we obtain (a) (x + y)Ge + w(x + yu) oe. Ge yuh 3) = oP 2/0 where w is the pth root of mity e’ (To see why this is true, note that Low, wP"+ are the p roots of the polynomial tP -1, hence we have the identity @) Peas (t- W(t - ot - a) .. (b= a4), fron which (1) follows by substituting the umber for the variable +t.) ‘Thus we have a miltiplicative proplem in the number field Q[w], and in fact 4n the subring Z[w].” Kummer attempted to prove Fermat's conjecture by consider- ing whether the unique factorization property of % and Z[i] generalizes to *etwl fey # aw tee ta, uP Ps a, €@ Vis €@ vi). Bla] = fay + aw woe +a) _yoP~? the ring Z[w]. Unfortunately it does not. For exemple if p = 23, then not all members of %[w] factor uniquely into irreducible elements: .@., elements @€ 2{w] which are not units and such that whenever @= py, either Bp or y is a wit (see exercise 20). In other words, Z[w] is not a unique factorization domain (UFD) for p= 23. It is, however, a UFD for all primes less than 23. For these primes it is not difficult to show that x° + y® = 2? has no case 1 solutions. ‘The argument can be organized as follows: Assuming that Z[w] is e UFD, 4t can be shown that x + yw has the form uo? for some @ € Zlw] and some unit u€ Z[w]. It can then be shown thet the equation x+yw= uo®, with xX and y not divisible by p, implies that x =y (mod p). (See exercises 16-28 for the details.) Similerly, writing x? + (-z)P = (-y)P,, we obtein x = -2 (mod p). Bub then 2xP a xP 4 yP = oP = xP (mod p) , inplying that pl3x®. since p}x and p43, this is a contradiction. Thus case 1 of Fermat's conjecture can be established for all primes p for which lw] is a UFD. Wost can be done for other primes? Unique factorization in Z[w] wes needed only for the purpose of deducing x + yw= uo from equation (1); might 4t not be possible to deduce this in sone other way? The answer is yes for certain values of p, Ancluding for exemple p= 23. This results from Dedekind's amazing iscovery of the correct generalization of unique factorization: although the elements of Z[w] may not factor uniquely into irreductble elements, the ideals in this ring always factor uniquely into prime ideals. Using this, it is not hard to show that the principal ideal (x + yw) is the pth power of some ideal 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 + yo) = P = (a)? = (oP) and thus egein we have x+yw= uo? for some wit u. As before, this implies x = y (mod p) and e 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 @A-= 6B for some a, BE Zu]. (Verify that this is an equivalence relation.) It turns out (see chapter 5) that there are only finitely many equivalence classes of idesls mder ~. ‘The mumber of classes is called the class number of the ring Blu] , end is denoted by the letter h, ‘Thus h is a fmetion of p. DEFINITION: A prime p is regular iff p{h. fo explain why I (in the equation (x + yw) = 1°) must be principal whenever p is reguler prime, we note first that the ideal classes cen be multiplied in the obvious way: the product of two ideal classes is obtained by selecting an ideel from each; multiplying thems and taking the ideal class which contains the product ideal. This is well-defined: The resulting ideal class docs pot depend on the particuler ideals chosen, but only on the two original ideal classes (prove this). Multiplied in this way, the ideal classes actually forme group. ‘The identity element is the class C, consisting of all principal ideals (vhich really 4s a class; see exercise 31). ‘The existence of inverses will be estebllished in chapter 3. ‘Thus the ideal classes form a finite abelien group, called the ideal class group. If p is regular then clearly this group conteins no elenent of order p, and it follows that if r is principal then sois I: let C be the ideal class containing I; then C” is the class containing I”, which is (p+ Since Cy is the identity in the ideal class group and C cannot have order p, it follows that C= Cy, which shows that I is principal. As we noted before, this leads to » contrediction, showing that 2? + yP = 2P has no case 1 solutions (i.e., solutions for which p}ayz) 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. In any case our attempt to prove Fermat's conjecture leads us’to consider various questions about the ring Z{w]: What are the mits in this ring? What are the irreducible elements? Do elements factor uniquely? If not, what Li can we sey about the factorization of ideals into prime ideals? How many ideal classes are there? ‘he 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 Q[w]. In every number field there is a ring, analogous to Zw], for which there are interesting answers. EXERCISES 2. 1-9: Define N: Zi] + Z by Na+ bi)sa +B. 1. Verity that for all a,p € Z[i], N(wp) = N(@)N(p), either by direct computation or by using the fact that W(a + bi) = (a + bi)(a - bi). Conclude that if aly in [i], then N(a)|M(y) in m. 2. let @€ Zi]. Show that @ is awit iff Mc) =,1. Conclude that the only units are 41 and +i. 3. Iet @€ @[i]. Show that if N(@) is prime in 2 then a is irreducible in [i]. Show that the sane conclusion holds if N(a) = Bp’, where p isaprimein Z, p#=3 (moah). 4, Show that 1-4 4s 4rreductble in Z[i] and that 2 = u(1- 4)* for sone uit u. 5. Notice that (2 + i)(2 - 1) =5 = (1+ 2i)(1 - 21). How is this consistent with unique factorization? 6. Show that every nonzero, non-unit Gaussian integer a is a product of Avreducible elenents, by induction on (a) . 7. Show that @[i] is a principal ideal domain (PID); i.e., every ideel I is principal. (As shown in Appendix 1, this implies that Z[i] is e UFD.) Suggestion: ‘ake @€I- (0) such that N(@) is minimized, and consider the miltiplies ya, y€ Z[i]; show that these are the vertices of an infinite family of squares which fi11 up the complex plane. (For example, one of the squares has vertices 0, @, ia, and (1+ i)a; all others are translates of this one.) Obviously I contains all ya; show by a geonetric argument that if I contained anything else then minimality of N(@) would be contradicted. 8. We will use unique factorization in [i] to prove thet every prime p21 (mod k) is a sum of two squares. (a) Use the fact that the multiplicative group a of integers mod p is cyclic to show that if p=1 (mod) then n® = -2 (moa p) for some n€ Z. (b) Prove that p cannot be irreducible in Z[i]. (Hint: pln? 4. = (n+ i)(n - i).) (c) Prove that p is a sum of two squares. (Hint: (b) shows thet p=(a+#bi)(e+ di) with neither factor a unit. eke norms.) 9. Describe all irreducible elements in @[{i] . ao-ay: tet w= 0/3. 24081. poring wm: alo] > % ty Ne + bu) =e - aber. 10. Show that if a+ bw is written in the form u+ vi, where u and v are veal, then N(e + bu) =v +. 11. Show that for all ap € Z[lw], Map) = N(@)N(p), either by direct computation or by using exercise 10, Conclude that if aly in Zlw], then N(@)|N(y) in Ze 12. let @€ Zw]. Show that @ ise uit iff W(@) =1, and find all mits in Zw]. (There are six of them.) 113. Show that 1- 4s irreducible in Z{w], and that 3 = ull - o)* for some unit u. hk. Modify exercise 7 to show that 2[w] is a PID, hence a UPD. Here the squares are replaced by parellelogrems; one of them has vertices 0, a, wa, (w+ jo, and all others are translates of this one. Use exercise 10 for the geonetric argument at the end. 15. Here is @ proof of Fermst's conjecture for n=: re x*4y! 2 2! hes a eclution in positive integers, then so does x'+y'aw. tet x, y, w be @ solution with smallest possible w. ‘hen x°, y", w isa primitive Pythagorean triple. Assuming (without loss of generality) that x is odd, we cen write Pant, y= om, wom sme with m end n relatively prime positive integers, not both odd. (a) Show thet xer-s?, neers, mare with r and s relatively prime positive integers, not both odd, (b) Show that r,s, and m are pairwise relatively prime. Using y = hrem, conclude thet r, 8, and m are all squares, say a, bY, ana (c) show that at 4 pt =, and that this contradicts minimality of w. amti/p 16-28: Ist p be an odd prime, w 16. Show that Q@-oa-o)... =P) =p by considering equation (2). 17- Suppose that Z[w] is a UPD and fi|x+yw. Show that {f 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 m+ ypn = 1 for some mn€ Z. How is this a contradiction? 18. Use 17 to show that if [lw] is a UFD then x+yw= ua®, ae lw], uawmitin Zo]. 19. Dropping the assumption thet Z{w] is a UFD but using the fact that ideals factor uniquely (up to order) into prime ideals, show that the principal ideal (x + yw) has no prime ideal factor in common with any of the other principal ideals on the left side of the equation qa) (x + y)(x + yo) .2. Ge + yuh “4 = (2)? in which all factors are interpreted eas principal ideals. (Hint: modify the proof of exercise 17 appropriately, using the fact that if A is an ideal dividing another ideal B, then A>B.) 20. Use 19 to show that (x +yw) = 1? for some ideal I. 21. Show that every member of Q[w] is uniquely representeble in the forn : 2 -2 ay + 80 + Ow Hoe #8, ooh > 8 EQ Vi by showing that w is @ root of the polynomial al. 22, 23. 2h. 25. 10 (continued) e(t) 2 Po Pm ta and thet #(t) is ixreductble over @. ” (Hint: Tt is enough to show that £(t +1) is irreducible, which can be established by Eisenstein's criterion (appendix 1). It helps to notice that £(t +1) = ((t+2)P - 1A.) Use 21 to show that if a € Zw] and pla, then (writing Ga ay + OW tore +e, ow? a, € %) oll a, ere divisible by p. Define congruence mod p for p,Y € Z[w] as follows: B=y¥ (mod p) iff p- Y= 8p for some 8 € Alw]. (Equivalently, this is congruence mod the principal ideal p Z[u] . ) Show thet if p = y (mod p), then § = ¥ (mod p) where the bar denotes complex conjugation. Show that (p + y)? =p? + yP (mod p) and generalize this to sums of erbitrarily many terms by induction. Show that vae€ Z[w] , OP is congruent (mod p) to sme 2 € BZ. (Hint: write @ in terms of w and use 2h.) 26-28: Now assume p>5. We will show that if x +yw= uc? (moa p), a€ Zw], wv evnit in Zw], x and y integers not divisible by p, then x =y (mod p). For this we will need the following result, proved by Kumer, on the units of 2[w] : IEMA: Te u isa unit in Z[o] and UZ is its complex conjugate, then u/@ is a power of w. (For the proof, see chapter 2, exercise 12.) 26. Show that x + yw = ua? (mod p) implies x + yw # (x + yo Yul (mod p) 26. 28. 29. (continued) for sone k € Z. (Use the Lena on units and exercises 23 and 25. Note that B= wt.) Use exercise 22 to show that a contradiction results unless k =1 (mod p). (Recall that pay, p25, and wba uP"? a ewe 1 20.) Finally, show x = y (mod p). tet w= M3, verity that the product eho ews eso wee eel +c is divisible by 2 in Z[w], although neither factor is. It can be shown (see chapter 3, exercise 17) that 2 is en irreducible element in Z[w] it follows that Z[w] cannot be a UFD. 30-32: R is an integral domain (commtative 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 4s principal for sone @E€R, then A 4s principal. Conclude that the principal ideals form en ideal class. Show that the ideal classes in R form a group iff for every ideal A there is an ddeal B such that 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 @[@] for some algebraic number @€C. If a is o root of an irreducible polynomial over @ having degree n, then ve, yatta €@ vi) fat Qe] = fay taga+ . and representation in this form is unique; in other words, (1,0,...,0°74} isa basis for Q[a] as a vector spece over Q. We have already considered the field @{u] where w= e@/P, » prime. Recall that n=p-1 in that case. More generally, let w= e%/™, ww not necessarily prime. The field @{w] 1s called the né® cyclotomic field. Thus the first two cyclotomic fields are both just @, since w=1i, -i (resp.) for m=1, 2. Moreover the third cyclotomic field is equal to the sixth: If we set ete ws » then w= -ub = -(u2)®, whiten shows that @fw) = @fu]. In general, for oda m, the m®® cyclotonie field is the sane as the 2a", (show that if w= MM sen w= ut € @fuP].) on the other hand, we will show that the cyclotomic fields, for m even (m>0), are all distinct. This will essentially follow fron the fact (proved in this chapter) that the degree of the m™ cyclotomic field over @ is o(m), the nuiber of elements in the set 3B {kr 10O, are called the real quadratic fields; the Q[/m], m1, teke eny prime p dividing m end consider the equation mnf = (mg)(nh). Reducing coefficients mod p, we obtain O= img mh where the bars indicate that coefficients have been reduced mod p. (We have applied the ring-homomorphism z[x] > mx] +) But [x] is en integral domain (since Z, is; this is eesy to show), hence ig or th=0. But then p divides all coefficients of either mg or mh; as we showed ebove, this is impossible. Thus m=n=1, hence g,h € Z[x]. 0 Proof of Theorem 1: If f is not irreducible, then f= hg where g and h are nonconstant polynomials in @[x]. Without loss of generality we can essume thet g end h are monic. Then g,h € Z[x] by the lemma. But a is a root of either g or h and both have degree less then thet of f. This ise contradiction. O b COROLLARY 1: The only algebraic integers in @ are the ordinary integers. 1 COROLLARY 2: Iet m be @ squarefree integer. The set of algebraic integers in the quedretic ficla Q[/m] is {e+bJm: abe Z} if m=2 or 3 (moak), 2B, aye z, aed (nae) if m= (nodh). Proof: Iet @ar+s/m, x,8€Q. If s #0, then the monic irreducible polynomial over @ having a as e root is = 2rx + x = ms”, Thus a is on algebraic integer iff 2r and x - ms” are both integers. We Jeeve it ea an exercise to show that this implies the result stated shove. O Corollary 2 shows that the algebraic integers in Q@[,/m] forma ring. The seme is true in any mmber field. To prove this, it is enough to show that the sui and product of two algebraic integers are also algebraic integers. For this it is helpful to establish sone alternative characterizations of algebraic integers. THEOREM 2: ‘The following ere equivalent for a €@: (1) a is an algebraic integers (2) The additive group of the ring [a] is finitely generated; (3) @ 4s © menber of sone subring of © having e finitely generated additive group; (4) GACA for some finitely genereted additive subgroup ACE. Proof: (1) = (2): If a is @ root of e monic polynomial over Z% of degree n, then in fact the additive group of [a] is genereted ty 1,0, ..., a®~1, (2) = (3) = (4) trivially. (4) = (4): Tet o),..+,a, generate A. Expressing each ca, as a linear combination of a,,.+.,a, with coefficients in Z, we obtain 6 - © where M isan nxn matrix over 2. Equivalently, es (of - M) is the zero vector, where I denotes the nxn identity matrix. Since the a; are not all zero, it follows that of -™M has determinant O. (In other words, we have shown thet a is an eigenvalue of M.) Expressing this determinant in terms of the n° coordinates of OI - M, we obtain oP + lower degree terms = Thus we have produced a monic polynomial over Z having @ ese root. O COROLLARY 1: If a and p are algebraic integers, then so are +f and ap. Broof: Vie know that lai] and aZ{p] have finitely generated additive Groups. Then so does the ring Z[a,p]. (If O),.++,0, generate [a] and ByrreP, Generate mp], then the mn products @,6, generate m{o,p] .) Finally, [0,6] contains @+ 6 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 obtain monic polynomials for their sum and product. This result shows that the set of algebraic integers in € is a ring, which we will denote by the symbol A. In particular AK ise subring of K for eny mmber field K. We will refer to A{K es the number ring corresponding to the muber field K. We have determined the nuiber rings corresponding to @ end the quadratic fields. For the cyclotomic fields we have AN Q[w] = Zlw] ly however et this point all that is clear is that AN Q[w] contains Zw] (since wEA end AN lw] is aring). To esteblish equality we will need some further information about Q[w]: specificelly, its degree over Q and its discriminant. ‘HE CYCLOTOMIC FIELDS Tet w= oP pyeny conjugste of w (root of the sane ixreduethle polynomial over @) is clearly also an m” root of 1 and is not an nt” root of 1 for any n Ze], we obtain the fect thet Bx?) is divisible by Fx) in a Ix). But G(x") = (G(x))P (see exercise 5) and Z,[x] 4s a unique factorization domain (see appendix 1); it follows that F end @ have e common factor h in Z,[x] . Thon w[ FE =x" - 1. tis dmplies (see exercise 6) that h divides ‘the derivative of x"-1, which is m™"+, (flere the par denotes m reduced mod p.) Since ptm, m#O3 then in fact h(x) is just e monomial (again using unique factorization in z,{x]) + But this is impossible since h|x™ - That completes the proof. O COROLLARY 1: Q{w] has degree “{m) over Q. Proof: w has {m) conjugates, hence the irreducible polynomial for w over @ has degree Am). O COROLLARY 2: The gslois group of @[w] over Q is isomorphic to the multiplicative group of integers mod m gh= cr lbJa (2b €@), since JH and -/H are the two conjugates of JH. The mé? cyclotomic field has g(m) embeddings in ©, the @(m) automorphisms. On the other hand, the field Q[7@] has three embeddings in ©, only one of which (the identity mapping) is an automorphism. ‘The other two embeddings correspond to the conjugates wiF ana WIE of FB, were w= eM/3, not in [3/2] since they are not reel. These are clearly More generally, if K and L are two mmber fields with KCL, then we now (see appendix 2) that every embedding of K in € extends to exactly [L:K] embeddings of L in €. In particular, L has [L:K] embeddings in © which leave each point of K fixed. It is often preferable to work with automorphisms of e field, rether then embeddings in € (particularly if we want to compose them with each other). A useful trick for replacing embeddings of a nutber field K with eutonorphisns is to extend K to anormal extension L of Q (which is always possible; see appendix 2); each embedding of K extends to [L:K] embeddings of L, all of which are automorphisms of L since L is normal. For example, the field @ JZ] can be extended to L= QZ, wl, over Q. Each embedding of K extends to two automorphisms of L. (Exercise: K 27/3 | mich is normal describe the six automorphisms of L in terms of where they send xe and w.) ‘THE TRACE AND THE NORM Tet K be @ number field. We define two functions T and N (the trace and the nom) on K, as follows: Tet o,,+ 10, Genote the embeddings of K in €, where n= (K:Q]. For each @€K, set (0) = 0,(4) + (a) +... + 0,0) Ha) = 6;(@)eQ(2) «+ 6,2) « Clearly 1(c) end N(@%) depend on the field K es well es on a, When more than one fiela is involved we will write T(a) end m¥(a) to avoid confusion. ‘Immediately from the definition we obtain T(a +6) = T(a) + (s) and Nap) = N(a)N(p) for ell O68 €K. Moreover for r €Q we have T(r) = nr, N(x) =x". Also, for r€Q@ and @EK, Mra) =rt(a) end Mra) =x"N(a). We wii2 establish another formula for the trace end nom ani show that the values are always rational: Iet a heve degree a over Q (i.e., the irreducible polynomial for @ over @ has degree 4; equivalently, a has a conjugates over @5; equivalently @[a] has degree ad over Q). Tet +(c:) and n(a) denote the sum and product, respectively, of the a conjugates of @ over Q. Then we have TTBOREM 4s 2a) = F(a) wa) = (a(ay)"/ where n= [K:@]. (Note that 3 is an integer: in fact, it is the degree [k:@[a]] .) Proof: +(c) and n(a) are the trace end nom 2%) gna wl) oe a, a Bach embedding of Q[a] in @ extends to exactly 3 embeddings of K in €. That establishes the formas. ( COROLLARY 1; (cr) and N(a) are rational. Proof: It is enough to show that t(~) end n(a) are retional. This is clear since -t(a) is the second coefficient of the monic irreducible polynomial for @ over @, end sm(a) is the constant tern. 0 If @ is an algebraic integer, then its monic irreducible polynomial over @ hes coefficients in Z; hence we cbtein COROLLARY 2: If a@ is en algebreic integer, then 1(@) and N(a) are integers. O EXAMPIZ: For the quadratic field K = @[/m], we have (a + bm) = 20 wa + b/m) =o - me for a,b €Q. In this case @ is an algebraic integer iff its norm and trace are both integers. (Thet is not true in general, of course: consider, for exemple, any root of x3 +dx +1.) SOME APPLICATIONS Suppose we went to determine the units in the ring ANK of algebraic integers in K. Using Corollary 2 above and the fact that the norm is mltiplice- tive, it is easy to see that every unit has norn 41, On the other hand if a is an algebraic integer having norm +1, then Theorem i shows that 2 is also an algebraic integer (since all conjugates of a are algebraic integers), ‘This shows that the units in AK are the elements having nom +1. ‘hus, for exemple, the only units in Z[/2] are +1, A similar result holds for all ‘but two of the imaginary quadretic fields (see exercise 13). On the other hand, ‘the units in Z[JZ] correspond to integer solutions of the equation a? . op? = 415 there are infinitely many (see exercise Ih). The norm cen elso be used to show that certein elements are irreducible in QNK, es defined in chapter 1. Clearly @€ANK is irreductble in ANK whenever its norm is e prime in 2% (but not necessarily conversely). Thus, for exemple, 9+, 10 is irreducible in Z[ 10]. As an epplicetion of the trace, we can show that certain fields cannot contain certain elenents. For example, J3 ¢ OYE] (see exercise 16). te trace is also closely connected with the Aiscriminant, which we will define in the next section. First, however, we generalize the trace and the norm by replacing Q 23 with en exbitrery number fiela: Iet K and L be two number fields with KCL. Denote by oy,+++,0, the n= [L:K] embeddings of L in € which fix K pointwise. For a€L, define the relative trace and relative norn by TEA) = o(@) + (a) +. +o, (0) HCO) = oy (209 (4) + 9,(0) Thus, in this notation, we have me @ and wea. Again we have Rl + p) = TECa) + TRB) ana e(op) = Myo nE(B) for el. ap Ls TES) = mB ana we(b) = 8" for all 6 € Ks and TH(ba) = OTE) and mf(oa) = oN) for 5 €K and @€L. Exactly as before we can prove THEOREM 4": let a €L and let a be the degree of a over K. Let +(c) and n(@) be the sum and product of the d conjugates of a over K. Then (a) = Bela) We) = @o). 5 coronary: (a) and w(c) are in K. If @EANL, then they are in ANK. oO When there are three different fields, the relative traces and norns are related in the following way: THEOREM 5: Let K, L, and M be number fields with KCLCM. Then for all @€M we have TEC)) = C@) , WAH Ca)) = melo) - (This is referred to as transitivity.) Proof: Let o,,.++,0, be the embeddings of L in © which fix K pointwise Eroot: oy n po! and let ty,-++,7,, be the enbeddings of M in € fixing L pointwise. We want to compose the o, with the ,, but we can't do that yet first we heve to extend all of the enbeddings to automorphisms of some field. Thus fix a normal extension N of @ such that MCN. ‘Then all oj and 1, can be extended to automorphisms of 3 fix one extension of each and again denote these extensions wy oy and ty (to confusion will result fron this.) Now the mappings can be composed, and we have aka) = a0? YO) = 2 exrsod eH nEGR(@)) = Text AUG) = i o,74(2) « Tt remains only to show thet the mm mappings o,7,, when restricted to M, give the embeddings of M in € which fix K pointwise. Since eli oats fix K pointwise and there is the right number of them (mm = [M:L][L:K] = [M:K]), it is enough to show that they are all distinct when restricted to M. We leave this to the reader (exercise 18). 0 ‘THE DISCRIMINANT OF AN n-TUPLE Iet K be a number field of degree n over Q. Let o,---,0, denote the n enbeddings cf K in €. For any n-tuple of elements 04,.+.,0, €K, define ‘the @iscriminant of 04,...,0, to be aise(ay,..-,0,) = Joyo)? , tues, the square of the determinant of the matrix heving 9, (a) in the 1 row, i® cotum. (Notation: we will write [a, j] t© denote the matrix having 4,5 in the i" you, 3°" coum, ana ag! to denote its determinant.) Notice that ‘the square makes the discriminant independent of the ardering of the o, and the ordering of the oy . ‘As with the norm and trace, we can generalize the concept of the discriminant by replecing @ with an axbitrary number field. (See exercise 23.) We can express the discriminant in terms of the trace T= 7; THEOREM 6: Aise(Oy,+++5%,) = [Xo,0,)| . Proof: This follows immediately from the matrix equation [oj (4) Ilo, (05)1 = Lo (0405) tenet (425) = [2@,e,)) and familiar properties of the determinant: = Taya, ana [aa] = [4][B] for matrices A ona B. 0 COROLLARY: dise(a,,---,0),) €@3 and if ell o, are algebraic integers, then disc(a,,...,%,) € Z. 0 Among other things, the discriminant determines whether the , are linearly dependent: .%, axe linearly dependent THORN 7: dise(@,,..-,0,) = 0 AFF O4,.-+,0,, over Q. Proof: Tt is easy to see that if the a, then so are the colume of the matrix [o,(,)] 3 thus the discriminant ie 0. are linearly dependent over @ Conversely, Af dise(o,.-+,0j,) = 0, then the rows Ry of the matrix [0(0,0,)] are linearly dependent. Suppose that 04,..-,0,, are linearly independent over Q. a Fixing rational mubers ,,...,8, (not all ©) such that aR) +...+@R) is ‘the zero vector, consider G = ac, +... +40. Necessarily o #0. Moreover by considering only the j*" coordinate of each row, we obtain the fact that Mac,) =0 foreach j. Since the oy over Q, they forma basis for K over Q; it then follows (since @ #0) that are assumed to be linearly independent the same is true of the ca,. But then 1B) = 0 for every p €K (why?) « 26 This is clearly a contradiction since, for example, 7(1)=n. 0 Theorem 7 shows that every basis for K over @ has a nonzero discriminant. We can obtain e relatively simple formula for the discriminant in the case of a basis consisting of the powers of a single element: TAROREM 8: Suppose K = @[O], and let 04,...,0, denote the conjugates of G@ over Q. ‘Then Py TM w= a = 2 whee) igr 3): disc(1 ~ w) = dise(w) . Proof implies that the discriminants are équel (see exercise 26). However it is probably easier to use Theoren 8: As cq, runs through the conjugates of w, 1 - a, ‘runs Zl - 0] = Zw] 4s obvious, since w=1~(1-). This in itself ‘through the conjugates of 1-w. ‘Thus = -aF= WT -a,) - (1-a,))? = @ise(1 - w). a 32 IEMA 2: (for m =p) Tha - oh) =p k where the product is taken over ell k, 1 and both have the same sign. (Note that both of these discrininants ere integers since all p, and all y, are algebraic integers.) 0n the other hand a similar argument shows that: dise(p,,-..,8,) is @ divisor of dise(y,,+++,¥,) « We conclude that the discriminants are equal. 0 Thus the discriminant of an integral basis can be regarded as an invariant of the ring R. Denote it by disc(R). We will also write disc(K) when R=ANK. 33 For exemple, we have disc(,/m) = 4m if m=2 or 3 (mod k) aisc(A n afm) = aise(ht lm 4AM) om if m= 1 (mod k) essuming m is squarefree. (Exercise: Verify this computation in four different way8, using various formilas which have been established.) One spplication of the discriminant is in identifying integral bases: 2, are in R, they form an integral basis for R iff Assuming Q,,...,0,, disc(o,,...,0,) = disc R (see exercise 274). As another application, we will generalize Theorem 10 to any cyclotomic field. This will follow from a more general result relating the algebraic integers in a composite field KL to those in K and L. We know that if K and L are ‘two number fields, then the composite KL (defined as the smallest subfield of € containing K and L) actually consists of ell finite sums GP, + +O,p,, all a EK, all A, EL (see appendix 2). Ifwe let R, 8, and 1 denote the rings of algebraic integers in K, L, and KL, respectively, then it is obvious that T contains the ring BS = (0,6, +... +@6,: ell a, ER, all p, €8) and it is natural to ask whether equality holds. In general it doesn't (see exercise 31). However we can show that T= RS under certain conditions which are, conveniently, satisfied by cyclotomic fields. let m and n denote the degrees of K and L, respectively, over @. Iet d denote the greatest conmon divisor gced(dise R, dise 8). THROREM 12: Assume that [Ki:Q] =m. Then TCH RS. 3h ‘Thus in particular we have COROLLARY 1: Tf [KL:@] =m end d=1, then T= FS. 9 To prove Theorem 12, we need a lenma from field theory: TEMA: Assune that [KL:@] =m. Let o be an embedding of K in €, and let 7 be an enbedding of L in €. ‘hen there is an embedding of KL in € which restricts to o on K andto T on L. Proof: We know (see appendix 2) that go has n distinct extensions to enbeddings of KL in €3 no two of them can agree on L, hence they have n distinct restrictions to L. One of these must be +, since L has only n embeddings in ¢. O Proof of Theorem 12: Iet (Q,,.++,0,,} be a basis for R over Z (Lee, an integral basis for K) and let {6,,++458,} be a basis for S$ over %. ‘Then the m products 0,6, forma basis for RS over 2%, and also for KL over 4P5 @ (why?). Any a € 7 can be expressed in the form = oe Ee Ps Where r and ell m,, axe in 2, and these mn +1 integers have no comnon factor >1: gear, gea(m, 5)) ‘fo prove the theorem, we have to show that, for eny such a, r]d. Clearly it will be enough to show that r[disc(R); by symmetry r will also divide @isc(S) and we will be done. ‘The lemma shows that every embedding o of K in € extends to an embedding (which we also call oc) of KL in ©, fixing each point of L. Hence for each o we have 2 of) = Ht olae, « Setting nn, = aL xy ga x 3 for each i =1,....m, we obtain m equations B_ of@y x, = of0) ser ema 9) one for each ¢. Wow solve for the x, by Cremer's rule: x, = y,/5, where 5 is the determinant formed by the coefficients o(0,) » and y, is obtained from 8 by replacing the 1°? are algebraic integers, since ell o(0,) and o(@) are; moreover & = dise(R). colum by the ofa). It is clear that 6 andall y, Setting e = dise(R), we have ox, = By €A; then in fact ne ox,= 2 tp ecanuas. gaa 3 Recalling that the p, form an integral basis for S, we conclude that the rational numbers en, ,/t must all be integers: Thus r divides all ems: Since by assumption r is relatively prime to gea(m, 5) , it follows that rle = dise(R). O Using Corollary 1, we can prove COROLLARY 2: Let K=@lul, w-e™™, Roank. mon R= z[u]. Proof: his has already been established if m is a power of a prime. If m is not a power of a prime, then we cen write m~=mm,, for sone relatively prime integers mm, >1. We will show that the results for m, and for m imply the result for m. (Thus we are proving R= Z[w] by induction on m.) Setting 36 4 SANK, R= ANK, we assume (inductive hypothesis) that R, = Z{w,], R, = Zw]- To apply Corollary 1, we have to show that K=K,K, and thet the degree and discriminant my me conditions hold. Clearly w*=u,, 0° =u). Tt follows thet w= uius for sone 3,8 € % (why?) and-hence K = K,K,. Moreover this shows that Z{u] = Zw] Zw]. The degree condition holas: q{m) = @(m, )e(m,) since m, and m, sere relatively prime. For the discriminant condition, recall that we have show that disc(w,) divides a power of m, and dise(w,) divides a power of m,. Finally, then, we conclude that R=RR = Bo] Zlyl= Zw]. O Tt would be nice if every number ring had the form Z[G] for some a. Unfortunately, this is not always the case (see exercise 30). Kquivalentiy, there ney not exist an integral basis of the fom 1,0,...,0°"1. mis suggests the following vague question: Does there always exist an integral basis whose members are expressed in terns of a single element? Of course the answer is yes, since K = @[O] for some @ and hence every menber of K is a polynomial expression in @ with coefficients in @. ‘This is not particularly illuminating. However what Lf we require that these polynomials have some special form? An answer is provided by the following result: "THEOREM 13: Let a €R and suppose a has degree n over Q. Then there is an integral basis £,() £, ©) aa 1 Snaa3 polynomials over Z, and f, has degree i. ‘The where the 4, are in 2% and satisfy ala the f£, are monic are uniquely deternined. i Broof: For each k, 1 f_1(0)/4, 1, 6) isa besis over z for K,,, (exercise Seen Seen ef, _,(@) Ce) and this is in WR, ,,) (wiy?). Tt follows that oy = mi(p) for some mé %. Defining 4, = md, _,, we have 1p) = da, , which implies that B= £,(a)/a, for sone £,(cc) = cM + lower degree terms. However we cannot yet say that f, has integer coefficients; all that is clear is that af,/d, has 38 integer coefficients. However since £,(@)/a,_,=mp €R, we have (a) - af, _1(@) Aaa and in fact this has been selected co that y€R,. Using our basis for R, we =yeR, ean write y= B()/4, for some g € Z[x] having degree @ prime) then t(xP) = (£(x))?. (suggestion: Use induction on the number of terms.) Show that if f and g are polynomials over a field K and f|g in K(x], then flg'. (itint: write g = 1h ana aifferentiate.) Complete the proof of Corollary 2, Tasoren 3. 8. 9. co 4o (a) Let oak psi (mod), and J-p if p=-1 (mod). (Hint: Recall that we Pp an odd prime. Show that Q[w] contains Jp if have shown that dise(w) = + pP7? with + holaing iff p =i (mod h).) Express ,f-3 and 5 as polynomials in the appropriate w. (b) Show that the &°" cyclotomic field contains JZ. (c) Show that every quadratic field is contained in a cyclotomic field: In th fact, QL] is contained in the d°” cyclotomic field, where a= disc(AN@L/m]). (More generally, Kronecker and Weber proved that every abelian extension of Q (normal with abelian Galois group) is contained in @ cyclotomic field. See the chapter i exercises. Hilbert and others investigated the abelian extensions of an arbitrary nutber field; their resvilts are know as class field theory, which will be @iscussed in chapter 8.) With notation as in the proof of Corollary 3, Theoren 3, show thet there exist aTtife _ yuge integers u and v such that e e’. (Suggestion: First write = eMib/k 1, relatively prime to Kk.) Complete the proof of Corollary 3, Theorem 3, by showing if m is even, nlr, and g(r) < om), then rem. (a) Suppose all roots of a monic polynomial f € Q[x] have absolute value 1. Show that the coefficient of x” has absolute velue < ® |) where n is the degree of f and @ is the binomial coefficient. (b) Show that there are only finitely many algebraic integers of fixed degree n, all of whose conjugates (including @) have sbsolute value 1. (Note: If you don't use Theorem 1, your proof is probably wrong.) (ec) Show that a (as in (b)) must be a root of 1. (Show that its powers are restricted to a finite set.) 1b. 1. sn Now we can prove Kunmer's Jemma on units in the p‘” eyelotomie field, as stated before exercise 26, chapter 1: tet w=0"/P » an oad prine, end suppose u is aunitin Zw]. (a) Show thet u/% 1s a root of 1. (Use 11(c) above and observe that complex conjugation is a member of the Galois group of @[w] over Q@.) Conclude that u/U=+w" for sone k. (b) Show that the + sign holds: Assuming u/U=-u, we have uP =P; show that this implies that wu? is divisible by p in Zlw]. (Use exercises 23 and 25, chapter 1.) But this is impossible since uw is a unit. Show that 1 and -1 are the only units in the ring AN@LJm], m square- free, m<0, mf-l, -3. What if m=-1 or -3? Show that 1+/2 isawit in ZLJ2], but not a root of 1. Use the powers of 14+,J2 to generate infinitely many solutions to the diophantine equation o” - 20% = 41. (Tt will be show in chapter 5 that all units in ZLJZ] are of the fon + (1+/B)K, ke m.) (a) Show that z{J-5] contains no element whose norm is 2 or 3. (b) Verity that 2+3 = (1+,/-5)(1- J -5) is an example of non-unique factorization in the number ring 2[J-5]. set a= YS. use the trace T= 1) to show that J3¢ alo}. (Write Y3=2a4 bat ca’ + ao? and successively show that a=0; b=0 (wnat is 1(JB/o) 7); c= 03 and finally obtain a contradiction.) Here is another interpretation of the trace and norm: Iet KCL and fix OQ €L; multiplication by a gives a linear mapping of L to itself, considering I as a vector space over K. Iet A denote the matrix of this mapping with respect to any basis (0 ,0,,-..] for I over K. (Thus the 3* cotum of A consists of the coordinates of Gay with respect to the ly. 18. iy. (continued) G+) Show that (a) ana RE(@) are, respectively, the trace and determinant of A. (Hint: It is well known that the trace and determinant are independent of the particular basis chosen; thus it is sufficient to calculate them for any convenient basis. Fix a basis {By ,Bp2++} for L over K[@] and multiply by powers of a to obtaine basis for L over K. Finally, use Theorem kt .) Conplete the proof of Theoren 5 by showing thet the ot, have distinct restrictions to M. (Hint: If two of them agree on M then they agree on L3 what does this show?) Iet R be a commutative ring and fix elements @,,8,,-.. €R. We will prove ‘by induction that the Vandermonde determinant Lae a is equal to the product. (a, - 8,)+ Assuming that the result rerdecn holds for some n, consider the determinant L 8, seeeeeeee £8) 2d Syaa cers Ppa) 43 19. (continued) for eny monic polynomial f over R of degree n. ‘hen choose f cleverly so that the determinant is easily calculated. 20, Let £ be a monic irreducible polynomial over a mumber field K and let @ be one of its roots in €. Show that f(a) = 7 (a - 8) with the product fa taken over all roots pf#G. (Hint: Write £(x) = (x - a)g(x).) 21. Let @ be an algebraic integer and let f be & monic polynomial over Z (not necesserily irreducible) such that £(a) = 0. Show that dise(a) aiviaes nllp(q) , 22, Iet K bee mmber field of degree n over Q and fix algebraic integers Os that @=0 or 1 (modh). Letting Oy,+++,0,, denote the embeddings of 0%, EK. We know that a= disc(a,,+++,a,) is in %3 we will show K in ©, we know that 4 4s the square of the determinant lo,(e,)] - ‘This determinant is e sum of nt terms, one for each permutation of (1,...,n). Let P denote the sun of the terms corresponding to even permutations, and let N denote the sum of the terms (without negative signs) corresponding to oad permutations. Thus a= (P- 1) = (P+N)* - KEN. Complete the proof ‘by showing that P+N and PN are in Z. (Suggestion: show that they ere algebraic integers and that they are in Q; for the latter, extend all oy ‘to some normal extension L of @ so that they become automorphisms of L.) In particular we have disc(A NK) =O or 1(mod4). ‘This is known as Stickelberger's criterion. 23. Just as with the trace and norm, we can define the relative discriminant disc ( of an n-tuple, for any pair of muber fields KCL, [L:K] =n. (a) Generalize Theorems 6-8 and the corollary to Theorem 6. (b) let KCLCM be number fields, [L:K] =n, [M:L] = m, and let fay, +00,Q,) and (6y)-+- 8) be bases for L over K and M over L, respectively. Esteblish the formila 23. (bv) (continuea) atschlag y+ -s0yfg) = (atsef(cy,--+,0,,)) HE aise es, 48) « Suggestion: Tet o),.+-,0, be the embeddings of L in € fixing K pointwise, and 1),...,7, the embeddings of M in € fixing L point- wise. Extend ell o's and 7's to automorphisms of a normal extension of @ as in the proof of Theorem 5. Define (mn) x (mn) matrices A and B es follows: A has o,1(f,) in row m(i- 1) +h and colum m(i- 1) 4k, and zeroes everywhere else (so that A consists of n m xm blocks, arranged diagonally from top left to bottom right); B has a, (¢,) in vow m(i- 1) +t and colum m(j-1)+t for each t, isgtsm, and zeroes everywhere else (so that B is obtained from the nxn matrix fo, (45)] by replacing each entry by the corresponding multiple of the mx m identity matrix). Show thet the desired forma follows from the equation |ap|* = |p||a[?. (catcwlate |B] ty rearranging rows and columns eppropriately.) (c) Tet K and L be number fields satisfying the conditions of Corollary 1, Theoren 12. Show that (aise 7) = (aise RI aise syK (mis can be used to obtain e formila for dise(w), w= oTiAD ) Let G be a free ebelian group of rank n and let H be a subgroup, Without loss of generality ve take G= Z@...@ % (n times). We will show by Anduction that H is @ free ebelian group of renk Z denote ‘the obvious projection of G on the first factor (so that an n-tuple of integers gets sent to its first component). Let K denote the kernel of TT. (a) Show that HK 4s a free abelian group of rank Sn-i. (bo) The image 1(H) Cc % is either {0} or infinite cyclic. If it is {0}, then H=HNMK; otherwise fix h €H such that 1(h) generetes 1(H) and show that H 4s the direct sun of its subgroups Zh and HN K. 25. 26. aT. 45 Show that for any algebraic nunber a, there exists m€ Z, m#O, such that ma is an algebraic integer. (Hint: Obtain £€ Z[x] such that f(x) = 0 and take m to be @ power of the leading coefficient.) Use this to show that for every finite set of algebraic mubers a, , there exists me %, m#O, such that oll ma, € A. Prove the following generalization of Thecrem 11: Let Py,++.,8, and Yystr+s¥, be any menbers of K (a mumber field of degree n over @) such that the 6, and y, generate the same additive subgroup of K. ‘then aise(Ay-++58,) = Qisc(yy,.++,4,) + (Mus we can define aise(G) for any additive subgroup G of K which is generated by n elements. This is only interesting when the n elements are linearly independent over @, in which case G is free ebelian of vank n.) Let G and H be two free abelian subgroups of rank n in K, with HeG. (2) Show that G/H is a finite group, (b) ‘The well-known structure theorem for finite abelian groups shows that G/i is a direct sum of at most n cyclic groups. Use this to show that G has a generating set BprrB, Such that (for appropriate integers a;) G\Py,+++4,8, 18 & generating set for H. (c) Show that dise(it) = |e/i[Paise(a) . (a) Show that if Oy ye0e,%, ER=ANK, then they form an integral basis for R iff disc(o,,...,0,) = disc(R). (his can actually be esteblished without using (c): Express the @, in tems of an integral basis and show that the resulting matrix 4s invertible over % iff the discriminants are equal.) (e) Show that if Oy ,o02, ERS ANK and ise(o,,++-5,) is squarefree, then the a, form an integrel besis for R. (This result can also be obtained from Theorem 9.)

You might also like