Scribd
Upload
45 views46 pages

Landau Elementary Number Theory 1 5

Teoría de números

Uploaded by

Yani Betancourt Gonzalez
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
45 views46 pages

Landau Elementary Number Theory 1 5

Teoría de números

Uploaded by

Yani Betancourt Gonzalez
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/ 46
ELEMENTARY NUMBER THEORY BY EDMUND LANDAU TRANSLATED BY JACOB E. GOODMAN Columbia University WITH EXERCISES BY PAUL T. BATEMAN Professor of Mathematics Universit ry of Iilinois AND CHELSEA PUBLISHING COMPANY NEW YORK, N. Y. Copyricut By S. Hirzex at Lerpzic, 1927 ALLE RECHTE, INSBESONDERE DAS DER UEBERSETZUNG IM FREMDE SPRACHEN, VORBEHALTEN CopyricHT By Marranne Lanpau, 1955 CopyRIGHT VESTED IN THE ATTORNEY GENERAL OF THE UnrtTep STATES, PURSUANT TO LAW MANUFACTURED AND SOLD UNDER License No. A-1692 OF THE ATTORNEY GENERAL OF THE UNITED STATES THE PRESENT WORK IS A TRANSLATION INTO ENGLISH, By JAcoB E. GoopMAN, OF THE GERMAN-LANGUAGE work ELEMENTARE ZAHLEN- THEORIE (Vor. I, oF VorLESUNGEN UEBER ZAHLENTHEORIE), BY Epmunp LANDAU, WITH ADDED EXERCISES By Paut T. BATEMAN AND E. E. KoHLBecKER Liprary oF Concress CataLocue Carp No. 57-8494 Copyricut © 1958, sy Cuetsea PusiisHINc CoMPANY Copyricut 1958, py CHELSEA PusLisHinc CoMPANY PRINTED IN THE UNITED STaTES oF AMERICA CONTENTS PART ONE Foundations of Number Theory Cuapter I: The Greatest Common Divisor of Two Numbers.... 11 Cuapter IT: Prime Numbers and Factorization into Prime Factors 18 Cuapter III: The Greatest Common Divisor of Several Numbers.. 26 CuapterR IV: Number-Theoretic Functions . 29 CuapTer V: Congruences 37 CuapTer VI: Quadratic Residues .. 53 CuapTer VII: Pell’s Equation 76 PART TWO Brun’s Theorem and Dirichlet’s Theorem Introduction 2... 2... 6. cece eee eet eee eee 87 Cuapter I: Some Elementary Inequalities of Prime Number BT Frey tae eta eb eee eee 88 Cuapter II: Brun’s Theorem on Prime Pairs............6..005 94 Cuaprer III: Dirichlet’s Theorem on the Prime Numbers of an Arithmetic Progression ...........6..e eee eens 104 §1: Further Theorems on Congruences ....... 104 §2: Characters . . 109 §3: L-Series . 5 §4: Dirichlet’s Proof . 125 PART THREE Decomposition into Two, Three, and Four Squares Introduction 0.2.6.6... eee eee e cece eee ees 129 Cuapter I: Farey Fractions ... 131 Cuapter IT: Decomposition into Two Squares. . - 135 7 8 ConTENTS Cnapter III: Decomposition into Four Squares... .-. 141 Introduction -.. 41 §1: Lagrange’s Theorem ........-.--.-----5 142 §2: Determination of the Number of Solutions. 146 Cuapter IV: Decomposition into Three Squares . 151 §1: Equivalence of Quadratic Forms. 151 §2: A Necessary Condition for Decomposability into Three Squares .......-..-..000-. 161 §3: The Necessary Condition is Sufficient..... 162 PART FOUR The Class Number of Binary Quadratic Forms Introduction -.. 167 Cuapter I: Factorable and Unfactorable Forms ... --. 170 Cuapter II: Classes of Forms «+. 172 Cuapter IIT: The Finiteness of the Class Number... - 175 Cuapter IV: Primary Representations by Forms 180 CHapTer V: The Representation of h(d) in terms of K(d)....... 191 Cuarter VI: Gaussian Sums .. . 197 Appendix ... . 202 Introduction . - 202 §1: Kronecker’s Proof .. -- 203 §2: Schur’s Proof .. §3: Mertens’ Proof . Cuapter VII: Reduction to Fundamental Discriminants........... 219 Cuapter VIII: The Determination of K(d) for Fundamental Discriminants -. 221 Cuaprer IX: Final Formulas for the Class Number. . . 227 APPENDIX Exercises Exercises for Part One . -. 233 Exercises for Part Two . - 242 Exercises for Part Three 247 Inpex or Conventions ; INDEX oF DEFINITIONS . 253 INDEX CHAPTER I THE GREATEST COMMON DIVISOR OF TWO NUMBERS Until further notice, lower case italic letters will always represent in- tegers, i.e. 1, 2, 3,... (positive integers or natural numbers), 0 (zero), —1,—2, —3, ... (negative integers). The following facts will be used constantly: If a is an integer, then so are —a and |a|; if a and 6 are integers, then so are a+b, a—b, and ab; if a>b, then a=b-+-1; and if a0 and b is arbitrary, then there is exactly one pair of numbers q and r such that (1) b=qa+r, 0Srq, then uzqtl, b—uasb—(q4+ lha=r—a<0. The desired relations 01. Then any number a>0 can be expressed in one and only one way in the form: =O) +09+-->+6ng", N20, Co>0, OStm1, and assume the assertion true for 1,2,...,a@—1. a belongs to one of the intervals lca0, since ¢ag"=a—r>g"—g"=0; in addition, cn0, then, since r0, VSbmrZ=big'2=g' ; therefore aby th gt +: +bigif0-gttt+ +0. ge *+eng" 2) The proof of uniqueness goes as follows: Let B= Oy teygt ++ tong" =do+dygt ++ +deg', N20, Cn>0; OStm0, OSdn0, ee hence gSl49'|=lQ+-° +e—g9 1 SG—Y1+ +9 )=g— 1 ~9Oand b>0. Of all the common multiples of a and b (there are such multiples, and even positive ones: for example, ab and 3ab), let m be the smallest positive one and let n be any of them (nZo). Then min. In words: Every common multiple is divisible by the smallest positive one. Proof: By Theorem 7, the numbers q and r can be chosen such that n=qm-+r, OSr0. Therefore r=0, n=qm, min. Tueorem 10: If a0 and bla, then [o|slal, so that every a=-0 has only a finite number of divisors. Proof: a=qb and q-+0; therefore lal24, joj =|a|/o|25). TueEoreM 11: Letaand b not both be 0. Let d be the greatest common divisor of a and b. (d exists and is >0; for at least one of the numbers a, b is +0 and hence, according to Theorem 10, has only finitely many divisors ; and the number 1 is certainly a common divisor of a and b.) 1) If f ts any common divisor of a and b, then fd. In words: Every common divisor goes into the greatest common divisor. THEOREMS 9-11 15 2) If a>0, b>0, and m is the smallest positive common multiple of a and b, then md=ab. In particular, then: If a>0, b>0, and d=1, then m=ab. Proof: Case1: Let a>0and b>0. Since ab is a common multiple of aand b, then by Theorem 9, mab, oe is an int mm $8 an integer. Setting aoe mo we shall prove the following: a) that if fla and f[b, then fly. b) that gad (which will prove all our assertions in Case I). In fact, a) If flaand fb, then 6 a alas, b/b—. la, bbs cA) is thus a common multiple of a and b ; hence by Theorem 9, ab m 7 af ab gif? so that the quotient abab_g ai is an integer, and consequently : Hig. b) Since enue g b6’g a are integers, we have 16 Parr Ong. I. Tue Greatest Common Divisor or Two NumBErs g is thus a common divisor of a and b. Since, by a), every common divisor f of a and b goes into g, and g>0, we have by Theorem 10, iso, so that g is the greatest common divisor of a and b. Case II: Suppose that the assumption a>0, b>0 is not satisfied but that a and 0 are still both +-0. Then 1) follows from Case I, since a has the same divisors as |a| and } the same divisors as !b|. In fact, d is the greatest common divisor not only of @ and } but of |a| and [| as well. Case III: Let one of the two numbers be 0, say a=0, so that D4 Then obviously d=|b|, and from f/0 and f(b it follows that fid. Notation: For any a and b which do not both vanish, the greatest common divisor of a and b is denoted by (a, b). Examples: (4,6)=2; (0,—3)=3; (4, -6)=2; (1,0)=1. TuHeoreM 12: If aand b are not both 0, then (a, b)=(b, a) Proof: The definition of (a,b) is obviously symmetrical in a and b. 0. T DeFinition 2: If (a,b)=1, that is, if 1 is the only positive common divisor of a and b, then a and b are called relatively prime. We also say: a is relatively prime to b. 1 and —1 are then the only common divisors of a and b. Examples: 1) (6,35)=1, since 6 has 1, 2, 3, and 6 as its only positive divisors, and none of the numbers 2, 3, and 6 goes into 35. 2) (a,0)=1 for a=1 and for a=—1, but for no other a. THEOREM 13: If (a,b)=d, then G 3) =1. Proof: If #>0, iff, i[f. then by Theorem 3, 2) we have fdja, fa/b, and therefore by Theorem 11 tdfd, so that by Theorem 3, 1) Hi), fel. Turorem 14: If c>0, ofa, o/b, é, 2 )=1, then c=(a,b). Proof: Since 4 and eS do not both vanish, a and b are not both 0. If we set (a, })=d, then c/d by Theorem 11, so that a is an integer. From ¢ [Der. 2] THEOREMS 12-16 7 it follows that i a 6 and therefore, since (é, *)= 1, d>0, c>0, wi c=d. c TueoreM 15: If albe and (a,b)=1, then ajc. In words: If a number divides the product of two numbers and is rela- tively prime to one of them, then it divides the other. Proof: By assumption, a=-0. 1) If b=0, then a=+1, since (a, 0)=1; and hence ajc. 2) If b=+:0, let m be the smallest positive common multiple of the rela- tively prime positive numbers |a| and |b|. By Theorem 11, m=|al|b|. Since, by hypothesis, bc is a common multiple of |a| and |b|, we have, by Theorem 9, |o| |>|lbe, ablbc (Theorem 1), alc (Theorem 3,1)). . TueoreM 16: If a/H ap, v=2, (a, Gn)=1 for lSn2, Theorem 15 yields, successively, ° » . aI dn, af ET an, ve @] TD dy lay. n= n=v—1 CHAPTER II PRIME NUMBERS AND FACTORIZATION INTO PRIME FACTORS The number 1 has only one positive divisor, namely 1; every number a>1 has at least two positive divisors, namely 1 and a. DerFinition 3: A number a>1 is called a prime number (or simply a prime) if it has only two positive divisors (namely 1 and a). Examples: The first few primes are 2, 3, 5, 7, 11. The letter p will be reserved for prime numbers only; likewise, symbols such as /3, po, .-., ~’, p”, ... will always represent primes. Our next aim will be to prove that every number a>1 can be represented as a product of primes (this will be easy) and that this representation is unique apart from the order of the factors (this is somewhat deeper). TueEorem 17: Every a>1 can be represented as a product of prime numbers: ) a=Ipy, rel. re 1 (For primes a=, this is obvious, and the product reduces to p= IT p,.) Proof (by mathematical induction): 1) For a=2 the assertion is true, since 2 is a prime. 2) Let a>2 and assume the theorem true for 2, 3,..., a—1. 21) Ifais prime, the assertion is true. 22) Otherwise, by Definition 3, there exists a factorization a=,d, 11 which is not a prime is called a composite number. The natural numbers thus fall into three classes : 18 [Ders. 3-4] THEOREMS 17-18 19 1) The number 1; 2) The primes ; 3) The composite numbers. There are, of course, infinitely many composite numbers; for example, all numbers of the form 2”, n=2. Tueorem 18: There are infinitely many primes. Proof: We must show that to any finite set of primes there can be adjoined yet another prime. Let pi, ..., Po be distinct prime numbers. Then ® a=1+IIp, n=l is, to begin with, >1 and, in addition, is not divisible by any of the prime numbers $1, ..., Pv, so that by Theorem 17 it is divisible by a prime number different from fi, ..., Po. Theorem 18 can be expressed as follows: For any §>0, let 1(€) repre- sent the number of primes S&. Then as approaches infinity, so does 2(€) ; i.e., given w>O0 there exists y= (w) such that x(é)>o if é>n=n(@). The question as to whether, and with what degree of accuracy, 2(&) can be approximated by the functions of analysis, can be answered only later on. In Part 7, Chapter 2, § 3 of my Vorlesungen iiber Zahlentheorie, the reader will find a very accurate result, the methods used being those of complex function theory. This result contains as a special case the “Prime Number Theorem” ee log é this theorem can be found in Part 7, Chapter 1, § 2 of the work cited. Let us also note here that the question, for example, of whether there exist infinitely many primes whose decimal representations end in the digit 7 will be answered (in the affirmative) in Part Two, Chapter III ; specifically, the answer will appear as a special case of Dirichlet’s well-known Theorem on Arithmetic Progressions (Theorem 155). None of this will be made use of, however, until it has first been proved. 20 Part One. II. Prrme NUMBERS AND FACTORIZATION INTO PRIME Factors TueoreM 19: If pa, then (p,a)=1. Proof: p has as positive divisors only 1 and p. Hence (p,a)=1 or p and, since p+a, the latter is impossible. THEOREM 20: If 7 plan, not then for at least one n we have plan. Proof: If, for all n, p+a,, then by Theorem 19 we would always have (~, @)=1, so that, by Theorem 16, ’ pt an. n=l TueoreM 21: If : Pi pn, n=l then for at least one n we have P=Pn Proof: By Theorem 20, Plbn for at least one n; but since the prime p, has 1 and p, as its only positive divisors, and since p+-1, it follows that p=pn. TueEoreMm 22: The representation (2) of any number a>1 is unique up to the order of its factors. Inwords: Every prime number appearing in a decomposition into “prime factors” of a given number appears equally often in every such decomposition. Every a>1 is therefore of the form a=ITp', Pla where p runs through the various primes that divide a; and where every l=I,,p>0 and is uniquely determined by a and p. (This is the so-called canonical decomposition of a.) Example: 12=2.2.3=2.3.2—3.2-2—2?.3=3.2% Proof: It is obviously sufficient to prove the following: If 2 of t= U p= Py PSPS -* SPo SPS S Pe n=l then v=v', Pr=Pn for lonxv. Tueorems 19-23 21 1) For a=2 the assertion is true, since we merely have avs], p= pl=2. 2) Let a>2 and suppose that the assertion has been proved for 2, 3, 4,...,a—1. 21) Ifaisa prime, then v=v=1, p\=p, =a. 22) Otherwise, we have v>1 and v’>1. Since ie v pi lI pa, py|T pas n=l n= it follows by Theorem 21 that Pi= Pm Pr=Pm- for at least one 2 and at least one m. Since PySPa=PiSPm= Py » we have P= Pi Now (since 11, let T(a) be the number of positive divisors of a, and let feat a= I ps n=l be the canonical decomposition of a (i.e., Pi, ..., Pr are distinct and every In>0). Then a has for its positive divisors the numbers (3) I pm, OSmaSly for lansr n= and no others. Hence r T (a)= IT (ln +1). n=1 (That the numbers (3) are distinct follows from Theorem 22.) 22 Part Ong. II. Prrme NuMBERS AND FAcTorIzaTION INTO Prime Factors Proof: 1) Each number of the form (3) obviously divides a. 2) If d>0 and dja, then a=qd, so that d cannot contain any prime factor that does not divide a, nor can d contain any prime factor of a a greater number of times than that factor appears in a itself. Dertinition 5: For any real number &, let [§] denote the largest integer SSE, that is, the integer g for which gséé then aZ[6]+1> [6]. TueoreM 24: The number q of Theorem 7 is equal to [2]. Proof: qasb=ga+r<(q+))a, 6 qsy0and 9>0, then the number of positive multiples of k which are Sn is (z “I: Proof: Since h>0 and hkSy, it follows that n 09. TueEoreM 26: Ifk>0 and 0, then ()-(7] k kl (For 9>0 this also follows from Theorem 25, for there are just as many positive multiples of k up to 7 as there are up to [7].) (Der. 5] THeEorems 24-27 23 Proof: From n 9S7-<9+1 it follows that Eo kgO0, and let p be any prime. Then p divides n! exactly ° IF m= lp” times. (This infinite series converges, since the general term vanishes for suffi- ciently large m, and in particular for m >i , since we then have p">n, n - O< ims} . The series can therefore also be written as ie wea |S SS gp where, in case p>m, the sum stands for zero—as shall every empty sum henceforth.) In other words: We have Sin 2 [pl (4) nl =I p= P psn (where, in case n=1, the product represents the number 1—as shall every empty product henceforth) ; for the primes p>x do not divide n!. We can equally well write ie Ie nl =p"? p™ Pp where the product is taken over all primes arranged in increasing order of magnitude, for every factor is 1 for p>n. Proof: In preparation for later on, I present two proofs. 1) The number of positive multiples of the number p up to n is, by Theorem a5,[*]; the number of positive multiples of p? up to n is [a] 3 ete. The multiplicity with which p divides n! therefore oo = & number of positive multiples of p” up ton m=1 Ss fz == “|; mai Lp” 24 Part One. II. Prrme NUMBERS AND FACTORIZATION INTO PRIME Factors for each of the numbers 1, ..., » is counted / times (and so not at all if J=0) asa multiple of p" for m=1, 2,...,1, if p divides it exactly / times (20). 2) (This proof is longer than the former, and introduces the logarithmic function—which can, of course, be eliminated by the use of exponents—but the proof is otherwise useful.) Let us henceforth set _Jlogp for a=p*, cZ1, (5) A= [8 ior hee a>0. (thus A()=0, A(2)=log2, A(3)=log3, A(4)—log2, .4(5)—logs, A(6)=0, ...). Let the symbol Z1@ aja mean, on principle, for a>0, that the sum is taken over all positive divisors dofa. Then we have (6) loga=2'A (d). dja For (6) is obvious if a=1 (0=0), and if aH pt (r=re,p) pla is the canonical decomposition of a>1, then loga=Zrlogp=Z(A@)+A@) +--+ A@)=ZA (%). c 0 is From (6) it now follows that @ fa] @ é 7 I N= =, = 2}: oO) og (=Zhga=F FE A@=F4 (a4); (Iam generalizing the result somewhat, in that I replace » by any real €>0) ; for.A(d) appears only for 10 and every /, the terms of Ss fz lal m=ilp™ can be most expeditiously computed one after the other by use of the result feal-[le], which follows from Theorem 26. Example: n=1000, p=3; the calculations should not proceed as fol- lows (every 3 is >0 and <1) ume 333-49,, rove m1+6, 2Posr+a, ue 246, rae = 14043 but rather o 333-48, 38 un, Has, Forte, emt, f=16, in order to compute the terms 333, 111, 37, 12, 4, 1, and the final result of 498. CHAPTER III THE GREATEST COMMON DIVISOR OF SEVERAL NUMBERS TueoreM 28: Let a=l and b=1. Let their canonical decompositions be written a=IT p', b=IT p™ (l=hp >0, m=m5,y>0) pla pid (where, in case a or b=1, the empty product shall mean 1). /f / and m are allowed to assume the value 0, then a and b may be written in uniform nota- tion as a=IT pf, b=II pp. Plab piad Then (8) @, |= p> Gm), piad If v1, ..., Yr are real numbers, then Min (y1,..., Y-) represents here— as it shall from now on—the smallest and Max (y1,..., Yr) the largest, of the numbers y1,..., Yr. Examples: Min (—3, 0, —3)=—3; Max (1,0)=1. Proof: The positive divisors of a are (by Theorem 23) the numbers II pt, O0 3 for if they are not, then we merely discard those that equal zero and change the sign of those that are negative. 1) For r=2 the assertion is true by Theorem 11. 2) For r>2, I give two proofs. 21) Set a= I py.jqm I ph he Play ++ Oy Ply + + Oy + b=). Then (compare the proof of Theorem 28), we obviously have (Gy eee, Qe) TT pa Gyr), Play +++ dy and every common divisor is + Tp’, 0Sv< Min (1,.., 1), Ply * * 6 and hence goes into (a;,...,¢,). 22) Let the assertion be already proved for r—1. Every common divisor of a,..., a, divides a,,...,a@,-, and hence (a,..., @;—1) ; it also divides a,, and hence ((a1,...,@r—1), 4), This number divides (a,...,¢--1) and a,, and hence @4, @)..., @p—1, Gy} it is therefore equal to (a,..., ar). ‘We should make note of the relation (9) y+ +21 Gr) = (Gy 2+) Gr—2)y Oe) for r>2, a:>0,..., @,>0, which we found during the second proof. Tueorem 30: Let r=2, a>0,..., and a,>0. Then every common multiple of n of a,,..., 4, is divisible by their smallest positive common mul- tiple v (which obviously exists). Proof: 1) For r=2, we know this by Theorem 9. 2) For r>2, I give two proofs (as for Theorem 29). 21) In the notation of the previous proof, we clearly have v= I pM G4, Bla, ++ * Oy 28 Part One. III. THe Greatest Common Drvisor Either n=0 (so that it is certainly divisible by v), or |n| contains every pla, ... a, at least J; times, ... , at least J, times, and therefore at least Max (i, «++, ly) times. 22) Let the assertion already be proved for r—1. Let w represent the smallest positive common multiple of a,...,a,;-1, so that is divisible by @,...,,—, and hence by w; but it is also divisible by a, and hence by the smallest positive common multiple of w and a,. Since this number is itself a positive common multiple of a, ..., @-—1, a,, it must equal v. CHAPTER IV NUMBER-THEORETIC FUNCTIONS Derrnition 7: A function F(a) which is defined for every a>0 is called a number-theoretic function. The value of the function is not required to be a positive integer, nor an integer, a rational number, or even a real number. Examples: F(a)=a!, F(a)=sina, F(a)=(a+2)-*, F(a)=T(a) (the number of positive divisors of Theorem 23), F(a)=A(a) (Formula (5)), F(a) =Zd=S(a) (the sum of the positive divisors of a). dja TueroreM 31: If a>l and a=IIp! is its canonical decomposition, then pia 1 S(a)—7 7 za pl Proof: If we add the positive divisors p,™ p,™ +++ gy," of a enumer- ated in (3), and use the fact that 2 saa 2 pn Sao m= p—l’ then the result follows. Derinition 8: Any divisor of a other than a itself is called a proper divisor of a. DEFINITION 9: ais called even if 2/a; odd, if 24a. Examples: 0 is even; of two successive numbers a and a-+1, exactly one is always even, the other odd; every p>2 is odd. DEFINITION 10: a>0 is called a perfect number if a equals the sum of its proper divisors, that is, if S(@=2a. Examples: 6=14+243, 28=14+24+447+14. 29 30 Part One. IV. Numper-THeoretic Functions This old-fashioned concept of perfect number, and the questions asso- ciated with it, are not especially important; we consider them only because, in so doing, we will encounter two questions that remain unanswered to this day: Are there infinitely many perfect numbers? Is there an odd perfect number? Modern mathematics has solved many (apparently) difficult prob- lems, even in number theory; but we stand powerless in the face of such (apparently) simple problems as these. Of course, the fact that they have never been solved is irrelevant to the rest of this work. We will leave no gaps; when we come to a bypath which leads to an insurmountable barrier, we will turn around, rather than—as is so often done—continue on beyond the barrier. Tueorem 32: If p=2"—1 (so that n>1; for example, n=2, p=3; n=3, p=7), then PE} =2>41(91) is an (of necessity even) perfect number, and there are no other even perfect numbers. Proof: 1) For a=2-(20_1), —1=p we have, by Theorem 31, Bt =a 1) (p-+)=(2"—1) 2"=2a, 2) Ifq@isan even perfect number, then a=2°—1y, n>1, u>0 and odd, so that, by Theorem 31, ux 2a=§(a)= 1 (a) = (2"— 1) $(u) 2-1 and A u In this formula, eI (=S(u)—w) is an integer, and hence (since n>1) it is a proper divisor of u. The sum S(x) of all the divisors of w is therefore equal to the sum of u and a certain proper divisor. Hence u is a prime, and the proper divisor u Poi =1, so that u=2"_1. This proves the theorem. THEOREMS 32-33 31 Now, are there infinitely many perfect numbers? I do not know. It was already mentioned above that 2*—1 is prime for n=2 and n=3. For n=4, 2"—1=15 is composite. More generally, 2"—1 is always composite if n is composite ; for if n=bce, with b>1 and c>1, then Qn— 1 = Qhe_ 1=(2— 1) (226) 4 26-9) 4 see D4 ), where both factors are >1. For n=5, 2"—1=2'—1=31 is a prime that yields the perfect number 16 - 31=496 ; for n=7, 2"X—1=2’—1=127 is a prime that yields the perfect number 64 - 127=8128; for n=11, 2\—1=2"—1=2047=23 . 89 is compos- ite. The question is, therefore, whether there are infinitely many primes p for which 2’—1 is a prime. Even this is not known. Are there infinitely many odd perfect numbers? I do not even know whether there is a single one. However, I should like to ask the reader not to meditate too long over these two questions; he will meet with many more promising and gratifying problems in his study of this work. The analogous problem of finding all the numbers a>1 which are equal to the product of their factors, i.e., for which (10) Hama a>1 is trivial. For the following simple theorem holds: TueoreM 33: (10) holds if and only if a=p* or a=pipe, Pike. Proof: 1) Asd runs through all the positive divisors of a, so, obviously, does +. It therefore follows from (10) that d- 9 )=a=ar, ae dja a eae T@=4; and hence, by Theorem 23, (h-F1) oe =4 in the canonical decomposition a=p,'- + - pyr, so that either r=1 and 1,=3, or r=2 and h=h=1. 32 Part Ong. IV. Numper-THEorETIC FUNCTIONS 2) Conversely, in these cases, TT d=1- py: p+ pP=p,°=(py? and IT d=1- py Py - PxPa=(PrPs)’, dips appa respectively. Derinition 11: The number-theoretic function u(a) (the Mébius Function) is defined by lifa=t, B(a)= } (—1)" if ais the product of r (21) distinct primes, O otherwise, i.e., if the square of at least one prime divides a. The numbers a1 that are not divisible by the square of any prime (or, equivalently, by any perfect square >1) are also called square-free numbers; this quite customary terminology is just as logical as saying that two numbers are prime to each other when they have exactly one positive common divisor (namely, 1). In this sense of square-free, we say: u(a@)=-+1 if a is square- free, and u(a)=0 otherwise. Examples: u(1)=1, u(2)=—1, 4(3)=—1 (u(p) is always =—1), H(4)=0, pO=—1, 4(6)=1, u(]D=—1, u(8)=0, 4(9)=0, “(10)=1. TueoreM 34: If a>0, b>0, and (a,b)=1, then # (ab) =p (a) 4(6). Proof: 1) Ifaor b is not square-free, then neither is ab, so that (0b) =0=1(a) (8). 2) Ifaand b are square-free, then since (a, b)=1, ab is also square-free. If a=1 or b=1, then the statement is obviously true; otherwise the number of prime factors of ab equals the sum of the number of prime factors of a and of b. 1 for a=, TueoreM 35: Lu(d)=) =—— 7 Zee {i for a>1, Proof: 1) Zu()=#()=1. 7" 2) If a>1 and if a=p,4--+- p,r is the canonical decomposition of a, then obviously Eu (=Fu(d=1+ (1) —Y+($)+ > +(2) ay dja ip, .+ + Py =5()) (-1)'=(1—1)"=0; [Der. 11] THEOREMS 34-37 33 for if s=1, 2, ..., 7, then there are exactly i: divisors of pi... p, which consist of exactly s prime factors, and for these we have u(d)=(—1)*. Tueorem 36: If 21, then Bay (E —l=1, ZH [3] Proof: Let the formula of Theorem 35 be summed over a=1, 2, This gives + [E]. <2 o-feolah for, by Theorem 25, the number of positive multiples of d up to & is [4]. TueoreM 37: If 21, then £4) 21 n=l 1 Remark: The infinite series therefore either converges, or else it oscillates between finite limits. The question as to which of these two alternatives holds does not interest us at the moment ; the reader can learn the answer in Part Seven, Chapter 12, § 1 of my Vorlesungen iiber Zahlentheorie. Gordan used to say something to the effect that “Number Theory is useful since one can, after all, use it to get a doctorate with.” In 1899 I received my doctorate by answering this question. Proof: Wehave FES oe Hence, by Theorem 36, 2 4O_4|—| #n0y(2-[2))| 2 (E-[Z]) se, 254M) <14@-Yaz, nai 1 34 Part One. IV. NumBer-THeoretic Functions Tuerorem 38: Let F(a) be any number-theoretic function whatever. Let G(a) be the number-theoretic function @@=ZFO. Then F@=Zn @a(4). (This is the so-called Mébius Inversion.) Remark: The fact that F(a) is uniquely determined at all, in reverse, by G(a) is clear to start with; for from GQ=F(), GQ=FA+ ++ EB)HFOR)+ oe we can successively compute F(1), F(2), F(3),...- Proof: For every positive dja we have a(S )= a =F Oy HG (4)= cre ze@a(G)=Z oe ZER@FO | "ale (for b only runs through positive divisors of a, and to every such b there corresponds exactly to those d for which dja, and in fact for which dbja, that is, for which d/$ ) =2F@) Zu@O=FO, ba at since, by Theorem 35, oo 1 for b=a, aff 0 for b/a, b0 that divide a enter into consideration. To each d/a let there belong the n=kd for which (kd, a) =d, ie. (by Pei! 13 and 14), (%, $)=1 and moreover for which 01 let a=pe o “pt be its canonical decomposition. Then by Theorem 40 we have g@= Ya aeitl On ait (1), n=l as is seen by calculating the 2’ terms of the product. THEorEM 42: For a>1 we have, in the canonical notation, Proof: By Theorem 41, r=. pn ts it = 36 Part Ong. IV. Numser-THEoRETIC FuNcTIoNS TueoreM 43: For l>0 we have o@p)=P*(p—))- Two proofs: 1) Special case of Theorem 42. 2) (Direct proof.) Of the numbers 1,2,..., p/, those not relatively prime to p! are precisely all the multiples of ; their number is p!~?; hence ep)=r—F. All of Theorem 42 itself can be proven directly by counting the numbers n which are not relatively prime to a and for which 1SnSc; but this is somewhat more laborious and is a good exercise for the reader. (The solu- tion of this exercise is, however, not essential for the remainder of this book.) Tueorem 44: If a>0, b>0, and (a, b)=1, then 7(a)=9(@)9(0). Proof: Without loss of generality let (canonically) a = It Pn n> 1 and b= mn" >1, From Theorem 42 it follows that m= ?@)= Hp —*py—1), 90) = Ht a Fn q—I). Since (a, b)=1, ab= TE pin a ae nai is the canonical decomposition of ab; hence, by Theorem 42, 9(@) =I a (Pl) TT an™®—* On l= 910) 0) The reader will find another proof of Theorem 44, one based directly on the definition of y, in Theorem 74. CHAPTER V CONGRUENCES In this chapter m will always be >0. Derinition 13: a is said to be congruent to b modulo m, written a==b (mod m), if m|(a—b). ais called incongruent to b modulo m, written a==b (modm), if Examples: a==b (mod 1) for arbitrary a and b. Any concept such as “congruent,” “equivalent,” “equal,” or “similar,” in mathematics must satisfy three properties (the so-called reflexivity, sym- metry, and transitivity), which are expressed here by means of the following three theorems. THEOREM 45 (Reflexivity): We always have a==a (mod m). Proof: m/|0, m|(a—a). Tueorem 46 (Symmetry): If a=b (mod m) then b=a (mod m). 37 38 Part OnE. V. CoNGRUENCES Proof: m|(a—b), and hence, by Theorem 1, m/(b—a). Tueorem 47 (Transitivity) : If a=b (mod m), b=c (mod m) then a= (mod m). Proof: mja—b, m/b—c, m|(a—b)+-(b—c), ml/a—e. Thus, just like equations, congruences (with the same modulus) can be written in sequence as a congruence with more than two terms; for example, a=b=c (mod m). The following theorem (which, incidentally, makes Theorems 45-47 self- evident) provides a useful necessary and sufficient condition for the validity of a congruence. THEOREM 48: According to Theorem 7, given the numbers c and m, there is a uniquely determined number r such that c=qm-+r, 0Sr 0, then a=b°. Proof: Follows from Theorem 53. TueoreM 55: Let famqtaat-taar—Z es (nz0) be any rational integral function with integer coefficients. If a=b, then f(a) =f (0). The solutions (if any exist) of the congruence thus fall into complete residue classes mod m. Proof: By Theorem 54 it follows from the hypothesis that a’=b" for 00. If a=b (mod m), /a—b. then ac=be (mod cm), and conversely. Proof: Since c>0, it follows from Theorem 3 that the relations m/(a—b) and cm|c(a—b) are equivalent. THEOREM 59: If a=b (mod m), n>0, n/m, then a=b (mod n). Proof: m|(a—b) and n/m; hence n|(a—b). 42 Part One. V. ConGRUENCES Tueorem 60: If a=b (mod my) for n=1, 2,...,0 (v=2) then, if m is the smallest common positive multiple of my,..., My, we have az==b (mod m). Proof: a—b is divisible by m, ..., m», and hence, according to Theorem 30, by m. Tueorem 61: If a=b (mod m), then (a, m)=(b, m). In particular: If (a,m)=1, then (b,m)=1. Consequently the num- bers in a residue class are either all relatively prime to m, or none of them is. Proof: From b=a-+mg it follows that (a, m)/b, so that (a, m)/(b, m) ; similarly, (b, m)/(a, m). Derrnition 15: Bya complete set of residues mod m is meant a set of m numbers each of which is congruent to exactly one of the numbers 0,1,..., m—I1 (mod m), that is, which represents the m classes into which all the integers mod m fall. It suffices, of course, to require that at least one of the m numbers belong toeach class. “If m objects are put into m pigeon-holes and each pigeon-hole contains at least one object, then each pigeon-hole contains exactly one object.” Alternatively: It suffices to require that each pair of m numbers be incongruent. “If m objects are put into m pigeon-holes and each pigeon-hole contains at most one object, then each pigeon-hole contains exactly one object.” Examples: Any m consecutive numbers, for example 1,..., #t, or the integers of the interval -F (exclusive) to z (inclusive) constitute a com- plete set of residues, since they are incongruent to each other. Our old Definition 14 can now be expressed as follows: The number of solutions of {(#)=0 (mod m) is the number of its solutions taken from any complete set of residues. Derrnition 16: By a reduced set of residues mod m is meant a set of v(m) numbers exactly one of which belongs to each of the classes all of whose numbers are relatively prime to m. [Ders. 15-16] THEoREMS 60-65 43 Once again it suffices, given @(m) numbers, to require either that at least one belong to each of the above-mentioned (mm) classes or that each of the ¢(m) numbers be relatively prime to m and that each pair of them be incongruent. THEOREM 62: If (k, m)=1, then the numbers 0-k,1-k,2-k,..., (m—1) -k constitute a complete set of residues mod m. More generally: If (k,m)=1 and a,..., dm is any complete set of residues, then so is ak,..., mk. Proof: From a,k==a,k (mod m), l0 and djm, then a residue class mod 3 breaks up into d residue classes mod m. THEOREM 66: Let n>1 and let at least one of the numbers ay,..., dn be different from 0; set Gy «+ On) =". We claim that the diophantine equation (i.e., equation with integral coefficients and unknowns) yh +++ 4+Ontn=C is solvable if and only if dlc. Hence, in particular: If (a, b)=1, then (13) ax+by=1 is solvable. THEOREM 66 45 Proof: 1) If exactly one coefficient does not vanish, say a1, then 4 %,+0-%, +--+ +0-a,=6 is obviously solvable if a/c, that is, if (a, 0, ..., O)/e. 2) If at least two coefficients do not vanish, then we may assume without loss of generality that no coefficient vanishes ; for otherwise we simply omit those terms mm for which a,=0, and this does not alter the value of the greatest common divisor of the coefficients ; the number of terms that remain is then still =2. Without loss of generality we may even take all the coefficients to be >0; for we merely have to replace each negative am by —am (which does not alter the greatest common divisor) and the corresponding %m by —#mn. We may therefore assume that n>1, G>0, ..., da>0. 21) If our diophantine equation is solvable, then obviously d[ dye +++ Anta dle. 22) Let dlc. 221) If n=2, then we merely have to show that @,2,=c (mod a,) is solvable for +1. This follows from Theorem 65, since (Gy 4) /—6 222) Let n>2, and assume the assertion proved for 2,... »n—1; if we set (ay «+ +) Ona) =a then, by (9), (a, Oq,)=d. From what we showed in 221), it follows that @L+OnLp=C- for suitably chosen 7, +». By our induction hypothesis for n—1, it follows in addition, since (hy «+7 Ona) 42, that 46 Part ONE, V. CONGRUENCES OB, On—1 Spa = aE for suitably chosen %1,..., %n—1, 80 that, finally, yy ++ On —1 nat Ann =C- TueoreM 67: If (a,b)=d and dlc, then ax-+-by=c is solvable, by Theorem 66, then, given any solution xo, Yo, all the solutions are of the form b a t=atha, y=y—hs, where his arbitrary. Proof: 1) The fact that such a pair x, y satisfies the equation follows from the relation a(,+45)+6(y—0$) ax +by=0. 2) The fact that no other solutions exist is seen as follows. Without loss of generality, let b+-0. (Otherwise interchange a and b, and observe that as h runs through all the integers, so does —h.) Since ax-+-by=c=ax+ by, it follows that 0 (mod (6), 0 (mod |b), ax, —c= and hence by Theorem 65 (with a= —c, m=|b|), we have t=, (moa ), nan ho, 6 ha ha a by=c—ar=c a(x +h2) (c—ax,) bs by, bet =b(y—b5), a Y= hs. THEoREMsS 67-69 47 TueEorem 68: If (a,b)=1 and if xo, yo is any solution of (13), then all the solutions are of the form r= %+hb, y=Y—ha, where h is arbitrary. Proof: This follows from Theorem 67, with d=c=1. Tueorem 69: 1) The congruences ay x=a, (mod m,), (15) x=a, (mod m,) have a common solution if and only if (16) (m4, ™,) 44-4. In particular, therefore, they always do if (st, m,2)=1. 2) If condition (16) is satisfied and if m represents the smallest common multiple of m, and mz, then the common solutions of (14) and (15) consist of all the numbers in a certain residue class mod m. Proof: 11) If we set (m, m2)=d, then it follows from (14) and (15) that d/ a,—a,. 12) If @ | a—4,, then from among all solutions of (14) of the form w=a,+ym, (y arbitrary) we can certainly choose one for which (15) holds. For we need a,+ym,=a, (mod m,); this is equivalent to (7) m,y-+ (a,—a@,)=0 (mod m,) which, by Theorem 65, 1), is solvable. 2) If (16) is satisfied, and therefore (14) and (15) along with it, then congruence (17) is satisfied for suitably chosen yo precisely by 48 Part One. V. CoNGRUENCES Y=Y (moa ™) by virtue of Theorem 65, 2). Therefore, since =m (by Theorem 11), all the numbers x satisfying (14) and (15) are given by the formulas My My m, . a=a,t (u | 8) my 4,-+m, ¥yt nay 2 =a, +m, Yy+ hm, h arbitrary, but these constitute a certain residue class mod m. Tueorem 70: Let r>1, and let every pair from among the numbers my,..., mM, be relatively prime. Then the congruences (18) x=a, (modm,), n=l, ..,7 are consistent, and their common solutions consist of all the numbers in a certain residue class mod mye . . . My. Proof: 1) For r=2 this follows from Theorem 69, since m=m me in that case. 2) Let r>2, and assume the theorem proved for r—1. Then the first r—1 congruences (18) are covered by wa (mod m, --- m,). for a suitably chosen a. Hence, by Theorem 69, the conclusion follows, since ‘M, +++ M,—1 is relatively prime to my. Tueorem 71: Let r>1, and let each pair of numbers from among My, ..., My be relatively prime. Then the number of solutions of (19) f(*)=0 (mod mm, ---m,) equals the product of the numbers of solutions of (20) {(%)=0 (mod my), ..., {(2)=0 (mod m,). In particular: If m>1and m= IT p,'* is its canonical decomposition n= then, if r>1, the number of solutions of {(x)=0 (mod m) equals the product of the numbers of solutions of {(x)=0 (mod p,/n), THEoreMs 70-73 49 Proof: First of all, it is clear that (19) 1s satisfied if and only if the r congruences in (20) are simultaneously satisfied. Hence if one of these has no solution, then neither does (19). If the congruences in (20) are all solvable then, by Theorem 70, to each residue class mod m, that satisfies the first congruence in (20) there corresponds one-to-one a residue class mod m,...m, that satisfies (19) ; similarly for mo,..., mr. TuHeEorEM 72: If IDRC H yet bene", Pt Cm then the congruence (21) f(@)=0 (mod p) has at most n solutions. Proof: 1) For n=0 this is obvious, since for every +, =0 (mod 7), &% so that (21) has no root. 2) Let n>0, and assume the theorem true for n—1. If (21) had at least the n+1 (incongruent) roots 0, 71,..., ¥n, then if we note that HAH) = Zea e240) Seat ay afb 5) =(®—a)9(2) (2) =by Hb, Ef On— 129}, bn-1= Cm Pt On—1 it would follow that (i —%p) 9 (12) =f (&x) f (7) =0—0=0 (mod p), .+,M, $0 that and for k=1, g(x) =0 (mod 7), contrary to the induction hypothesis for n—1. Turorem 73: Let a>0, b>0, and (a,b)=1. Let x range over a complete set of residues mod b and y over a complete set of residues mod a. Then ax-+-by ranges over a complete set of residues mod ab. Proof: Of the ab numbers ar--by, any two are incongruent mod ab. For if ax,+by,=a2,+by, (mod ab), ax,+by,=ax,+by, (mod ), az,=ax, (mod 6), 2,=a, (mod 6), then 50 Part One. V. CoNGRUENCES and similarly, by symmetry, n=, (mod a). TueoreM 74: Let a>0, b>0, and (a,b)=1. Let x and y range over reduced sets of residues mod b and mod a, respectively. Then ax+-by ranges over a reduced set of residues mod ab. Remark: This is the direct proof of Theorem 44 which we announced earlier. Since Theorem 43 was also proved directly, there thus results a new, direct proof of Theorem 42, and consequently of Theorems 41 and 40; up to this point, everything had been obtained from the Mébius Inversion formula. Proof: If (x,b)>1, then certainly (ax-++by, ab) >1; for (+,b) divides ax+by and ab, and hence divides (ax+by, ab). If (y,a)>1, then, by symmetry, (ax-+-by, ab) >1 as well. What remains to be shown, by Theorem 73, is that if (x,b)=1 and (y,a)=1, then (ax-+-by, ab)=1. In fact, let p/(ax-+-by, ab). Then we would have jab, so that, without loss of generality, pla; moreover, p/(ax+by), so that p/by, and consequently (since (a,b)=1) p/y, contrary to the assumption that (y,a)=1. THEorEM 75 (The so-called Little Fermat Theorem): If (a,m)=1, then a?™=1 (mod m). Remark: It is not known whether the so-called Last Theorem of Fermat, which is discussed in Parts 12 and 13 of my Vorlesungen iiber Zahlentheorie, is true or not. I would therefore rather refer to it as the Fermat Conjecture, and to Theorem 75 simply as Fermat’s Theorem. Proof: Let Gy...) Gym) be a reduced set of residues mod m. Then, by Theorem 63, 4,..., @@gim) is also such a set. Hence the numbers a, are congruent to the numbers aa, (n=1,...,q(m)), apart from their order. Hence the product of the a, is congruent to the product of the aa,, or (m) —plm) — pian) tn n=l (ain) =a ire Bn (mod m), n=l n= so that, by Theorem 56, 1=a?™ (mod m). THEOREMS 74-77 51 TueoreM 76: If p+a, then a?—t=1 (mod ); for any at all we have ‘y a?=a (mod p). Proof: The first statement follows from Theorem 75, since v(p)=p—1; the second follows from the first by Theorem 51 if p+a; and if pla, it is trivial, since a?=0=a (mod p). TuroreM 77 (The so-called Theorem of Wilson): (p—1)!= (mod p). Two Proofs: 1) For p=2 and p=3, the statement is obvious. For p>3, I consider the p—3 numbers (22) 2, 3,..., p—3, p—2 For each r in this sequence, p+r, and hence, by Theorem 64, there is exactly one s in the sequence 0, 1,..., P—1 for which (23) rs=1 (mod p). s=0 does not obtain here; nor do s=1 and s=p—1, since otherwise r would be ==+1. The s therefore occurs in the sequence (22) as well. Moreover, 8 i fi 7 1=1 (mod 7) would give ple) C+, r=21 (mod 9). Hence to each r in (22) there corresponds exactly one s--r in (22) for which (23) holds. Since rs=sr, it follows, conversely, that r is uniquely —3 determined by s. The p—3 numbers in (22) thus break up into P zy Pairs in such a way that the product of the numbers in each pair is =1. Hence (p—2)!==2-3---(p—2) > =1 (mod p), (@@—1l=@—}) (9-2 52 Part OnE. V. CONGRUENCES 2) Ifwe set sh {()=2°——1-(z—m), m=) then clearly {(@)=e+0,2+---+cp—aaP—*. By Theorem 76, the congruence f(~)==0 (mod p) has at least the p—1 roots +=1,2,...,p—1. Hence, by 7 C)==C)=- --==cp_2=0 (mod p). Our result then follows from the fact that o=—1-(— IPI)!

You might also like

pFad - Phonifier reborn

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.

Note: This service is not intended for secure transactions such as banking, social media, email, or purchasing. Use at your own risk. We assume no liability whatsoever for broken pages.


Alternative Proxies:

Alternative Proxy

pFad Proxy

pFad v3 Proxy

pFad v4 Proxy