You are on page 1of 49

The abc Conjecture

A Thesis Presented for the Master of Science Degree The University of Tennessee, Knoxville

Jeffrey Paul Wheeler August 2002

ii

Dedication

This thesis is dedicated to

My Parents
whose loving sacrifices ensured me the best possible education.

“...for we were born only yesterday and know nothing, and our days on earth are but a shadow.” Job 8:9 (NIV)

iii

iv

Jamie. Ruth. Thank you also to Carl Wagner for guidance and encouragement. David. He also was able to run his program in parallel. and to Carl Sundberg for showing interest in my topic and for helpful comments and reference sources.my mother. my wife. Pavlos Tzermias. Mike Saum then improved Joel’s initial program (I am certain Joel could have improved it. I also wish to express gratitude to Ben Walker. but his upcoming wedding held priority) and this has led to much exciting data. is to be praised for her love and soft encouragement. the systems administrator. as well as other extended family memebers. Jeffrey. thanks to my committee members. Regarding the computer computations that were done. is to be praised for infinite patience and encouragement. My family deserves recognition for their continual support . The following are but a few of those individuals. Shashikant Mulay and David Anderson. As well I wish to thank my in-laws. As well. and brother. My Major Professor. for permitting us to use so many machines for all the various lengths of time. Jack and Cheryl Stankus. Special thanks to him also for choosing to work with me at a low point in my academic career.Acknowledgements A great number of individuals are owed a debt of gratitude for their contributions to any academic success I have attained. I am blessed to have her in my life. In particular. father. mostly for their daughter. Mike is also to be thanked for solving what I thought were A impossible L TEX problems. v . but as well for their assistance. Joel Mejeur is to be credited for scrapping my slow Matlab program in favor of his quicker C program.

vi .

This is the first time that such brute force computations have been published. Some computational computer work verifying these values is discussed. vii . Other known consequences are stated without proof. The idea of good abc triples is defined and all known good triples are stated. Topics supporting belief that the abc Conjecture is true are discussed.Abstract The problem of the abc Conjecture is stated and various consequences are established.

viii .

. . . . . . . . 5 Good Triples Associated with the abc Conjecture 5. . . . . . .2 Futher Consequences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .1 Preliminaries . . . . . . . . . . 4. . . . . . . . . . . . . . . . .2 The Evidence .2 util. . . . . . . . . . . . . . . . . . . . . . . .2 The Polynomial Analogue of the abc Conjecture . . . . 5. Bibliography Vita . . . . . . . . . . . . . . . . . . . . . . . . . . . .4 Program Listings . . 1 3 3 5 11 11 16 19 19 20 25 25 27 30 30 30 32 35 37 ix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .3 Computations Regarding Good Triples . . . . . . . . .2 Good Triples . . .4. . . . . . 3. . . . 3 Consequences of the abc Conjecture 3. . . 5. . . . . . . . . . . . . .c . 5. . . . . . . . . . . . . . . . . . . . . . . . . . .4. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .1 Specific Consequences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.c . . . . . . .1 abc-mpi. . . . . . 4 Evidence for the abc Conjecture 4. . . . . .Contents 1 Introduction 2 The Problem Stated 2. . . . 5. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2. .1 The abc Conjecture . . . . . . . .1 Preliminaries . . . . . . . . . . . . . .

x .

. . 27 xi . . . . . . . . .1 Known Good abc Triples . . . . . . . . . . . .List of Tables 5. . . . . . .

xii .

Nonetheless.1). As we will see. Pierre de Fermat (1601 − 1665) stated his “Last Theorem” 1 in the margin of his copy of Diophantus’s Arithmetica in 1637. This intriguing chapter in mathematics’ history came to a close in 1993 with the work of Andrew Wiles. A little later.W. or undecidable) could not be established. false. 1 n x + y n = z n has no nontrivial solutions in   for n ≥ 3. this conjecture has many fascinating applications. Oesterl´ e and D. despite intense effort by the best mathematicians. yet the conjecture’s certainty or uncertainty is not in sight. with the pace of science today. it is too early to tell. one of which is a version of one of the subject’s most celebrated problems. the development of Modern Algebra being one of the foremost. though. whether any mathematician could now make a mathematical conjecture whose status (true. is that the abc Conjecture implies a weaker (yet significant) form of Fermat’s Last Theorem (see Conjecture 3. Unlike most other Number Theory problems. In one of the boldest claims by one of the brightest individuals in the history of mathematics. Masser. though. Oesterl´ e was motivated by a conjecture of Szpiro regarding elliptic curves. We will see both Szpiro’s conjecture (Conjecture 3. It is very likely that his proof was incomplete.Chapter 1 Introduction On its own the abc Conjecture merits much admiration. Masser was motivated by considering an analogous statement over Z of Mason’s Theorem for polynomials. As is often the case with some of the more intriguing problems of Number Theory. his innocent enough statement incited hundreds of capable (and not so capable) individuals into feverish work for over three and one-half centuries.4) and Mason’s Theorem (Theorem 2. The significance of this is best summarized in a comment by John Fraleigh regarding Wiles’ proof of Fermat’s Last Theorem: “One wonders. the abc Conjecture is easy to state but yet difficult to verify. The abc Conjecture was posed in 1985 by both J. The interesting connection. for another 350 years”. 1 . Though it may be the case that the abc Conjecture is one such conjecture. Fermat wrote that he had a proof. but that he did not have enough room to write it in the margin. much has been accomplished.1). These individuals made great accomplishments in mathematics.

2 .

b. namely: for each ε > 0 there exists a positive universal constant µ(ε) 1 such that max(|a|. 2 1 3 . We now state two Lemmas that will be repeatedly quite useful. b. c > 0. it is already the case that the elements involved have gcd = 1. c) = 1. We then define the radical of n to be: r (n) = p1 · · · pk with r (1) = 1. The literature commonly refers to this constant as C (ε). c) = log max(|a|. Lemma 2. b. suppose n = p1 e1 · · · pk ek where the pi ’s are distinct prime numbers and the e i ’s are positive integers.|c|) log r (abc) = log c log r (abc) and considered whether the L’s are bounded. We will be considering non-trivial triples of integers (a. Under the hypotheses of the abc Conjecture. In other words.Chapter 2 The Problem Stated 2. any sum of the form a + b = c can be rearranged so that a. c) such that a + b = c and gcd(a. b.|b|.1. The redundancy is for emphasis. By definition of a multiplicative function. We will consider this more extensively in Chapter 5. Now we concern ourselves with the hypotheses. Obviously. Masser refined the statement into its more common form. r (n) is the greatest square-free factor of n. Oesterl´ e originally stated the conjecture in the form L = L(a. hence we will assume all elements of our triples are positive.1. |b|.1 (The radical of a positive integer).1 The abc Conjecture First we begin with a defintion: Definition 2. For n ∈ P. r is a muliplicative function 2 . |c|) = c ≤ µ(ε)r (abc)1+ε .

bn = 1. by induction.1. Proof. and c meeting the hypotheses.3) Since 2k |(32 − 1) and 2|(32 + 1) then 2k+1 |(32 k+1 k k (induction hypothesis) (2.4) (viz. The ε in the abc Conjecture is essential. So 32 k+1 n n − 1 = 32 = (3 k ·2 = (3 )2 − 1 2k 2k −1 (2.2. 32 + 1 is even) k (2. Obvious It is worthwhile to emphasize the importance of the ε in Masser’s version of the abc Conjecture. the claim is established. k Assume true for k .1.2) 2k − 1)(3 + 1) (difference of two squares) (2.Proof. i. 2k |(32 − 1). Note that the values meet the conditions of the hypotheses of the abc Conjecture. 4 .. First. consider an = 32 − 1. n 2n |(32 − 1) Proof. Claim 2.5) − 1). and cn = 32 where n ∈ P. Obvious. For all n ∈ P. r (n) ≤ n. Lemma 2. For an example.6) Hence. We show that there does not exist a µ such that c ≤ µ · r (abc) for all a.e. We will do this by using an example developed by Wojtek Jastrzebowski and Dan Spielman as reported by Serge Lang [8]. 2|(32 − 1). (2. For n = 1.1) (2. b. Proposition 2.

assume there exists µ such that c n ≤ µ · r (an bn cn ) for the above conditions.2) = µ · 3 · r 2n · ≤µ·3·2· 2n −1 32 − 1 2n (One may regard the fraction in statement (2.1) (by Lemma 2. 2. |cn |) = 32 n (2.1) (by Claim 2. First.12) n = µ · 3 · r (3 2n − 1] · 1 · 3 ) − 1) n n 32 (by Lemma 2. Put 5 .7) (by assumption) 2n ≤ µ · r (an bn cn ) = µ · r ([3 2n (2.8) (2. Remark 2.1. Before we close this section. In the abc Conjecture. For contradiction. Hence we shall state Mason’s Theorem. let us take a look at one of the conjecture’s influences. Recall from the Introduction that Mason’s Theorem inspired Masser. So max(|an |.10) (2. Multiplying both sides by 2n and dividing by 32 : 32 − 1 . 2 ≤µ·6· 3 2n n n n (2.12) as the product of all factors of 3 2 − 1 different from 2). µ(ε) varies inversely with the choice of ε.2 (The radical of a polynomial). |bn |.Proof. we state a remark that the reader may find useful while contemplating later material. a definition: Definition 2. Let p(t) be a polynomial whose coefficients belong to an algebraically closed field of characteristic 0.11) (2.2 The Polynomial Analogue of the abc Conjecture Before we consider some of the consequences of the abc Conjecture.9) (2.13) Letting n → ∞ the inequality fails and we get the contradiction establishing the necessity of the ε.

c(t)} ≤ n0 (a(t) · b(t) · c(t)) − 1.16) (2. we may now state: Theorem 2.n0 (p) = the number of distinct zeros of p(t). we have: (2.15) f +g =1 Differentiating we get: f +g =0 Rewrite as: g f ·f + ·g =0 f g ∴ f g ·g =− ·f g f So −f g f = g f g (2.14) (2. and c(t) be polynomials whose coefficients belong to an algebraically closed field of characteristic 0. Proof.18) (2.1 (Mason’s Theorem). Suppose a(t). 6 . We have a+b=c Dividing both sides by c: a b + =1 c c Putting f = a c b and g = c . 3 Let a(t). In other words. Then max deg{a(t). b(t).21) It is essential that the reader realizes the similarities between Mason’s Theorem and Masser’s version of the abc Conjecture. b(t). b(t). and c(t) are relatively prime and that a(t) + b(t) = c(t). n0 (p) counts the zeros of p(t) by giving them each a multiplicity of one.19) (2.17) (2. With this definition. hence g b = a f 3 (2.20) Observe that a = f · c and b = g · c.

and c are relatively prime) D (t) := where deg(D (t)) = n0 (abc) Now we make the observation that deg f f = −∞.20) into (2.26). if a ≡ 0.29) i j k .23) Notice: qi is the multiplicity of the root ρi where qi if t − ρi is in the numerator if t − ρi is in the denominator (2. (2.22) and (2.24) Thus R (t) = Hence R (t) = R(t) qi t − ρi (2.28) A common denominator for the numerator and denominator of (2. b(t) = j (t − βj )nj . and c(t) = k (t − γ k ) r k (2. Then R(t) = i (t − ρi )qi where the qi ∈ Z >0 <0 (2. if b ≡ 0. b. deg 7 g g = −∞. b f =−g =− a g f mi i t−αi nj j t−βj − − rk k t−γk rk k t−γk (2.Substituting (2.30) (t − αi ) · (t − βj ) · (t − γk ) (2. Now suppose a(t) = i (t − αi )mi .25) i i The advantage of (2.27) Then by (2.26) qi · R(t) t − ρi (2.28) is (since a.21) −f b f = g a g (2.26) is that the multiplicity of each distinct root is now exactly one.31) (2.22) Now suppose R(t) is a rational function with ρ i representing the distinct roots of the numerator and denominator.

35) (2.31). (2.40) (2. (2.39) (2. and (2.38).32). (2.) By (2. deg(b)} So.33) f f = deg g g = −1 if a nor b ≡ 0.39).b) = 1 a| Thus by (2.38). This will also be used in (2.36) f f =b· D· g g (2.37). and (2. (2.33) deg(a) ≤ n0 (abc) − 1 A similar argument yields deg(b) ≤ n0 (abc) − 1 As well deg(c) ≤ max {deg(a).34) Having established Mason’s Theorem we get 8 . b(t).and deg Hence deg D · f f and deg D · g g ≤ n0 (abc) − 1.37) D· g g (2. and (2.32) (Note that in (2. c(t)} ≤ n0 (a(t) · b(t) · c(t)) − 1.39): max deg{a(t). by (2.38) (2.33) we needed the hypothesis that the polynomials have coefficients in a field of characteristic 0. (2. (2.37).22) we get −D · f b f = g a D· g Hence −a · D · Since (a.

and h(x) = 1 be polynomials whose coefficients are in a field of char p > 0. and z (t) be relatively prime polynomials whose coefficients belong to an algebraically closed field of characteristic 0 such that at least one of them has degree > 0. By successively replacing the x(t) on the LHS with y (t) and z (t) and summing we get n[deg(x(t) + deg(y (t)) + deg(z (t))] ≤ 3[deg(x(t)) + deg(y (t)) + deg(z (t))] − 3 This is an obvious contradiction for n ≥ 3. Then x(t)n + y (t)n = z (t)n has no solution for n ≥ 3. Proof.2. By Mason’s Theorem we have deg(x(t)n ) = n · deg(x(t)) ≤ deg(x(t)) + deg(y (t)) + deg(z (t)) − 1. 9 . Remark 2. Then f (x)p = g (x)p + h(x)p . Fermat’s theorem for polynomials fails if char p > 0. Let x(t). y (t). g (x) = x.Corollary 2.1 (Fermat’s theorem for polynomials). let f (x) = x + 1. For an example.

10 .

y.1) (3. As stated in the Introduction.1. has only trivial solutions in the integers. We will use the symbol to mean the following: For functions x(t) and y (t) x(t) y (t) means ∃ C ∈ R.1 (The Asymptotic Fermat Problem). z ) = 1. The abc Conjecture implies the Asymptotic Fermat Problem. 1 Another way to state this is that in big oh notation x(t) Proof of Theorem 3. 1 11 .2) In the wake of Andrew Wiles’ accomplishment. Theorem 3. Notation (x(t) y (t)). We begin with one of the more interesting ones.Chapter 3 Consequences of the abc Conjecture 3.1 Specific Consequences This chapter states some of the conjecture’s fascinating implications. an academic corollary. more aptly. C > 0 such that x(t) ≤ C · y (t) for all t.1. Conjecture 3. xn + y n = z n . Then there exists N ∈ Z such that for n > N . this conjecture may be labeled a corollary or. we may make the appropriate rearrangements in the sum so that all integers are positive. By the abc Conjecture: | xn |≤ µ ε ε · r (xyz )1+ 3 3 ε ε n · r (xyz )1+ 3 | y |≤ µ 3 | xyz |1+ 3 | xyz |1+ 3 ε ε (3. This is due to the µ(ε). Again. y (t) means x(t) = O (y (t)). the abc Conjecture implies a weaker form of Fermat’s Last Theorem. where gcd(x.

Theorem 3. So 2p−1 = (2d )m ≡ 1m mod p = 1 mod p.and | z n |≤ µ Hence | xn | · | y n | · | z n |= | xyz |n (| xyz |1+ 3 )3 = | xyz |3+ε . 2d ≡ 1 mod p. ε ε ε · r (xyz )1+ 3 3 | xyz |1+ 3 . Otherwise. ε (3. Since d = ord(2) in U(Z/pZ).4) Thus for | xyz |> 1 we get n bounded. Since 2d = 1modp2 . The abc Conjecture implies that an infinity of primes satisfy the Wieferich Condition. It will be helpful to employ the set S := {p | p is prime and 2 p−1 ≡ 1 mod p2 }.3) (3.1 (Wieferich Condition). Conjecture 3. p prime ⇒ (m. our choice of ε determines the value of the N . But 2 p−1 =2 dm = (2 ) d m = (1 + pk ) m = m m−1 m m 1 (pk )1 + 1 (pk )0 + 1 0 m i=2 m m−i 1 (pk )i i divisible by p2 ≡ 1 + mpk (modp2 ) ≡ 1(modp2 ) ∴ p ∈ S and the claim is established. we establish the necessary definition. p) = 1. it follows that p k . Before we state the conjecture.2. Also. So 2n ≡ 1 mod p2 ⇒ 2d ≡ 1 mod p2 . Hence ∃ k ∈ Z such that pk = 2d − 1 ⇔ 2d = 1 + pk .2 (Infinity of Primes Satisfying the Wieferich Condition). Now write p − 1 = dm for some m ∈ Z. Let n ∈ P and p be a prime such that 2n ≡ 1 mod p but 2n ≡ 1 mod p2 . The abc Conjecture also implies the following classical conjecture. | xyz |≤ 1 and at least one of the integers must be 0.1.1. we first establish the following claim: Claim 3. d ≤ p − 1. Proof of Claim 3. But before we prove the theorem. Definition 3. A prime p ∈ Z satisfies the Wieferich Condition iff 2 p−1 ≡ 1 mod p2 . Write n = dr for some r ∈ Z. Then p ∈ S. 12 (since p k and p m) . Put d = ord(2) in U(Z/pZ) where | U(Z/pZ) | = p − 1 Hence d | (p − 1) and d | n. ∴ m < p − 1 < p. In particular. There exist infinitely many primes p satisfying the Wieferich Condition. It is worthwhile to note the role of the µ(ε) in the previous proof.

S finite ⇒ un is bounded. Let u. by the Lang-Trotter conjectures. 2n ≡ 1 mod p2 .8) (3. Originally the assumption that u and v be relatively prime was not made. Proof of Theorem 3. Moreover.10) · un −1 vn 1−ε 2 un −1 .3 (Hall’s Original Conjecture). i.9) 1+ε 1+ε = un 1+ε · vn vn Therefore un vn Multiplying both sides by vn −1−ε 2 1+ε 2 . v be relatively prime2 nonzero integers such that u3 − v 2 = 0.13) (3. there are only two known exceptions.12) (3. pi ∈ S and each pk | vn .2. the probibilty that 2 p−1 ≡ 1 + pk (modp2 ) for a fixed residue class k mod p should be O ( 1 p ).7) (3.11) Hence vn un ε−1 ⇒ a finite number of vn contradiction as n → ∞. Conjecture 3. vn 1+ε 2 .6) (3. But p un . pk ∈ / S.5) (3. Suppose S is finite. ∴ p2 | vn .e. by the abc Conjecture | 2n − 1 |= un vn ≤ µ(ε) · r (un vn )1+ε ≤ µ(ε) · (un vn ) (un vn ) 1+ε 2 1 2 1 2 (3. the number of primes p ≤ x such that 1 ) = O (log log x).3. Then | u3 − v 2 | | u | 2 −ε . (3. The following proof is due to Lang.14) Regarding the Wieferich Condition. The abc Conjecture implies Hall’s Original Conjecture. (3. ∴ p2 | un vn (since 2n − 1 = un vn ).The following proof is due to Silverman. 1 Theorem 3. p2 | (2n − 1). 2 (3. By the claim.e. most primes should satisfy 2p−1 ≡ 1 + pk (modp2 ) should be O ( p<x p the Wieferich Condition. 2 13 . for fixed x. Since (2n − 1) + 1 = 2n . Write 2n − 1 = un vn where ∀ pi | un . This is remedied by removing any common factor and then proceeding as dictated in the proof. i. Hence. Suppose p | vn .

) By (3. if true. namely: r (k ) ≤ k . By the abc Conjecture |u|m A similar argument yields |v |n Without loss of generality. has a bound for t. m(1+ε ) k mn−(m+n)(1+ε ) . Note that we could equivalently state that v 2 = u3 + t.3 In particular. (3.17) (3. n n (3. 14 .19). (3. b ∈ Z and let m. n ∈ P be such that mn > m + n.Proof.26) = r (k ) mn−(m+n)(1+ε ) .20) ) r (k )1+ε . Put a · um + b · v n = k. (3. the abc Conjecture would imply that |u| |t| 2+ε . now suppose |a · um | ≤ |b · v n |. 3 Note that the abc Conjecture.19) |v 1+ m · r (k )|1+ε = |v |(1+ m )(1+ε ) · r (k )1+ε . t ∈ Z.23) (Note that we needed Lemma 2. (3. m(1+ε ) |u| r (k ) mn−(m+n)(1+ε ) n(1+ε ) n ·m .17) |v |n Hence |v |n−( Thus m(1+ε ) m(1+ε ) m+n )(1+ε m (3.15) |uv · r (k )|1+ε .25) (3.24) (3. We prove a more general statement: Fix nonzero a. (3. Fix ε > 0.16) |uv · r (k )|1+ε . improves Baker’s bound for this situation.21) |v | r (k ) mn−(m+n)(1+ε ) k mn−(m+n)(1+ε ) . Then |u| Subsituting into (3.18) |v | m . n (3.2.22) (3.

Given this we identify the disciminant of E.) 15 .31) Conjecture 3. Put m = 3 and n = 2. First.e. |u| Thus |u| 2 Substituting for ε |u| 2 −ε 1 1 ε − 112 −5ε k 1−5ε = k 2+2ε ε 2+ 112 −5ε . ε ε Pick ε such that ε = 112 −5ε . Before we continue. 3 2 (3. (Note that Szpiro did not include the notion of r (D ) in his Conjecture. By (3.Having established the general case. (3.4 (Original Szpiro Conjecture). i. (3. then |D | r (D )6+ε c(E )6+ε . v ∈ Z.29) Recall from the Introduction that Oesterl` e was inspired by a conjecture of Szpiro. ε = 12+5ε . Since we are considering fields of characteristic 0 we may assume that our elliptic curves have Weierstrass equations of the form E : y 2 = x3 − ux + v where u. namely ∆ = 16(4u3 − 27v 2 ) and D := 4u3 − 27v 2 is the discrimanant of the cubic polynomial.28) k (1− 2 ·ε−ε ) < k. (3. we may establish the implication of Hall’s Conjecture. Assuming a Weierstrass equation with D the discriminant of the cubic polynomial and c(E ) the conductor of the equation.30) p fp if the reduction of E is non-singular if the reduction of E is multiplicative if the reduction of E is additive and δp is a bounded constant independent of the curve with δ p = 0 if p ≥ 5. Hence we shall consider Szpiro’s Conjecture. it is important to observe that r (D ) ≤ c(E ). Also. we will want to indentify the conductor of E.27) k ε ε (2+ 112 )( 1 ) − 112 2 −5ε −5ε =k 3 12ε ε [1− 2 ) ] · 1−5ε −( 112 −5ε 2 . some preliminaries.26). namely for prime p ∈ Z c(E ) := where   0 fp = 1   2 + δp p (3.

2 Futher Consequences In this section we list further consequences of the abc Conjecture without proof.4.5 (Erdo ¨s − M ollin − W alsh Conjecture).1. For the interested reader. see [9]. Theorem 3. Pairs of integers satisfying Brocard’s Problem n! + 1 = m 2 are called Brown Pairs. Definition 3.2 (Brown Pairs). the above problem has been generalized to the number of integer solutions of the equations (x!)n + 1 = y m (see [10]) and x! + B 2 = y 2 for arbitrary B (see [3]). For the proof of the opposite implication. 3. For n ∈ P.33) [(r (D ))2+ε ]3 by (3. Proof. For further information. The abc Conjecture is equivalent to Szpiro’s Original Conjecture. 1 ε. see [8]. 16 . By the abc Conjecture (in particular. Erdo ¨s refers to theses numbers as k-ful numbers where the k plays the role of the 2 in the above definition. The abc Conjecture implies that there exist only finitely many Brown Pairs.Theorem 3. We have Fix ε > 0 and put ε = 3 4u3 − 27v 2 = D. n is said to be a powerful number if for every prime p dividing n. p 2 divides n. Conjecture 3.5.27) and noting that r (D ) < D (3. our proof of Hall’s Conjecture) |u|3 and |v |2 Hence |D | r (D )6+ε c(E )6+ε by (3. There do no exist three consecutive powerful integers. The proof of this is in [11].31) (3.34) [(r (D )3+ε ]2 by (3. Definition 3.3 (Powerful Numbers). The abc Conjecture implies the Original Szpiro Conjecture.32) Remark 3.22) (3.

Theorem 3. no repeated roots. for the conditions of Roth’s Theorem. p 1 |α − | > 2+k q q −2 for all but a finite number of fractions p · q in reduced form. 1 For the above.6 (Mordell’s Conjecture). and content = 1. The following is due to Granville [6]: Theorem 3.10 (Squarefree Values of Polynomials). The abc Conjectue implies that.7. In [5] it is established that the abc Conjecture with an explict constant µ(ε) would give explicit bounds on the heights of rational points in Mordell’s Conjecture. 17 . where k = C (α) · (log q ) − 1 (log log q ) for some constant C (α) depending only upon α. Theorem 3. E. 4 Any curve of genus larger than 1 defined over a number field K has only finitely many rational points in K. the reader may also see [5]. The abc Conjecture implies that the set of triples of consecutive powerful integers is finite.9. In closing. Conjecture 3. Fix ε > 0. The abc Conjecture for number fields implies the Mordell Conjecture over an arbitrary number field. and gives bounds for the order of the Tate-Shafarevich group. Many more implications are given in [9]. 4 This is now a theorem after the work of G.6. In 1994. Bombieri [1] proved that the abc Conjecture implies a stronger version of Roth’s Theorem: Theorem 3. implies that the Dirichlet L-function has no Siegel zeros. we mention that the abc Conjecture also gives a way of counting squarefree values of polynomials. Theorem 3. For every algebraic number α. The following is due to Elkies [4]. F (n) is squarefree for infinitely many integers n. the diophantine inequality 1 p |α − | < 2+ε q q has only finitely many solutions. The abc Conjecture implies that for a polynomial F (x) with integer coefficients. Faltings (1984).8 (Roth’s Theorem).

18 .

Lemma 4. . . Put K = Q(α0 ) and let D = Ω ∩ K . D is the ring of algebraic integers in K (Note: since K is a cyclotomic field. . if p > 2 if p = 2.Chapter 4 Evidence for the abc Conjecture In this chapter. Let f℘ be the residue class degree of ℘. |c| = x2 + y 2 .1 Preliminaries For the following. . . .2) where ζm has the usual meaning e m for m ∈ P. For c = x + iy ∈ C. .e. then ord℘ θ < (c1 n)n p2 · log B · log log A · log A1 · · · · · log An where c1 is an effectively computable postive number. For α ∈ K \{0}. 4. ord℘ αj = 0 for j = 1. a theorem of C. . αn 1/q ) : K ] = q n+1 . (α)D = ℘ 1 1 · · · · · ℘g g . . put Θ = α1 b1 · · · αn bn − 1. let p be a prime number and put q= As well put α0 = 2πi 2 3 ζ4 ζ6 if p > 2 if p = 2. Lastly. n. . i. (4.L.1. in Z) such that |b i | ≤ B where B is a fixed integer ≥ 3. Define ord℘i α = e℘i . i. 19 . and Θ = 0. Let α1 . Given the above.e. If [K (α0 1/q . D ≡ Z[ζ 0 ]). This is the ramification index of ℘i . . since D is a Dedekind domain the fractional ideal (α)D can be written e℘ e℘ as a unique product of prime ideals in D . the curious reader may see [12]. Put A = max Ai .1) (4. . αn ∈ D such that |αi | ≤ Ai for 1 ≤ i ≤ n where each Ai ≥ 4. 1≤i≤n Let b1 . . . Stewart and Kunrui Yu establishing a weak form of the abc Conjecture is discussed. These are stated without proof. bn be rational integers (i. we now state some essential preliminaries. .e. .

Lemma 4. . Then for any prime p max{ordp a. v ≥ 1. . log 2 (4. gcd(a. c5 . Since a + b = c.Lemma 4. . Then [K (α0 q . .2 The Evidence Theorem 4. αn be prime numbers with α1 < α2 < · · · < αn . · · · . u ≥ 1. Write a = p1 e1 · · · pt et . . g ∈ P.3. . . and α1 = 2 and in this case [K (α0 2 .2. . where c2 is an effectively computable positive number. q1 . be the sequence of prime numbers in increasing order. . α1 q . . p2 . For α1 . . . . There exists an effectively computable constant k such that for all a. s1 . Similarly denote the largest primes dividing b and c by p b and pc respectively. 3 1 1 1 1 1 1 1 1 1 1 4. Without loss of generality suppose a ≤ b. qu . . α2 . . pt . c > 2. . . . b = q1 f1 · · · qu fu . where p1 . f. . α2 2 . if [Q(α1 1/2 . Let c4 . and e. sv are distinct primes with t ≥ 0. and a + b = c log c < r (abc) 3 The following proof is due to Stewart and Yu. αn 2 ) : Q] = 2n . denote effectively computable positive constants. ordp b. Let p1 = 2. α2 2 .. (1 + i) 2 . . Denote the largest prime dividing a by p a except when a = 1 and in this situation simply put pa = 1. . Lemma 4. .1. . .4. αn 1/2 ) : Q] = 2n and b1 · log α1 + · · · + bn · log αn = 0. c) = 1. and c ∈ P with (a. αn q ) : K ] = q n+1 except when q = 2. . . αn ∈ P. and c = s1 g1 · · · sv gv . b. c) = 1. Let q = 2 and α0 = ζ4 or q = 3 and α0 = ζ6 as well put K = Q(α0 ). and c ≥ 2. αn 2 ) : K ] = 2n+1 . .. it follows that a < b < c and that r (abc) ≥ 6. . Proof. ordp c} ≤ 20 log c . b. then |b1 · log α1 + · · · + bn · log αn | > exp(−c2 n)n log B (log log A)2 log A1 · · · log An . Let α1 . . . . . Then [Q(α1 2 . α0 = ζ4 . . b.3) 2 + log logkr (abc) . Then ∃ an effectively computable constant c3 > 0 such that for every positive integer r we have pj r j =1 log pj +3 r +3 > ( rc ) .

Since p|c and (a.3.1 ordp c ≤ ord℘ Θ < (c4 · (t + u))t+u · p2 · log log c · log log r (abc) · c 4 Similarly if p|b then.6) that log z < (c5 · (t + u))t+u · pc 2 · log log c 21 p|bc log p. c) = (b.7) ordp a < (c6 · (u + v ))u+v · p2 · log log c · log log r (abc) · It follows from (4. . . αn be the primes p1 . . Define q and α0 as in statements (4. . . . we put K = Q(ζ6 ). .5) − 1) = ordp (p1 4e1 · · · pt 4et · q1 −4f1 · · · qu −4fu − 1) for each prime p dividing c. α1 q . by (4. 4 Put Θ = ( a b ) − 1. . . 8fu }. c) = 1 we have ordp αi = 0 for i = 1. log 2 log p. we have p|ab 1 1 1 log c . . . for each prime p dividing c. (4.4) Since (a. while if p > 2 we put K = Q(ζ 4 ). take α1 = 1+ i instead of α1 = 2 and note that 24 = (1+ i)8 . . except in the case when p > 2 and α 1 = 2. t + u. by considering ordp (( b p|ac log p (4. put n = t + u. (4.8) p|ab log p · (log r (abc))2 . c) = (b. . Now let ℘ be a prime ideal of the ring of algebraic integers of K lying above the prime p. ordp c = ordp We now estimate ordp ( a b 4 c −b = ordp a −1 −b ≤ ordp ( a b 4 − 1). b) = (a. .3). . 8et . αt+u q ) : K ] = q t+u+1 . by considering ord p (( a ) − 1). . Clearly Θ = 0. by Lemma 4. . c) = 1. q1 . p|c (4. [K (α0 q . .1.4) and (4. we also have and if p|a then. . In this situation.Observe that log c = p|c (ordp c · log p) ≤ (max{ordp c}) · log r (abc). . So. . For n in Lemma 4. Now put B = max{8e1 . (4. If p = 2. Thus we have ordp Θ ≤ ord℘ Θ. 8f1 . pt . .1. .2). .1) and (4. qu arranged in increasing order. We do this by employing Lemma 4. Thus.6) ordp b < (c5 · (t + v ))t+v · p2 · log log c · log log r (abc) · c 4 ) − 1). (4.9) . . . As well let α 1 . . . B ≤8· Hence by Lemma 4.

(4.18) .9). So from (4.4) with c replaced by a together with (4. (4. log p (4. c (4. 4 (4.4 we have c10 ρ < (r (abc)) log log r(abc) . log c 4 log log c Again by Lemma 4. log b > log c − log 2 > log c . . 3 2 < (c8 · ρ)2ρ · (pa pb pc )2 · p|abc log p · (r (abc))6 .7) log c < (c5 · (t + v ))t+v · pb x · 4 log log z Now either a > √ b or a ≤ √ b.pc p . by(4. (4. (4.4) holds if we replace c by b.17) (4. Put ρ = u + t + v .3 we may use Lemma 4. √ or a ≤ b.pb .11) √ for a > b.11). b log( a+ b ) = log(1 · log b > log c 8 a + b) < log 1 + 1 √ b < 1 √ b < √ √2 . (4. (4. we use (4.14) p|bc log p · (log r (abc))2 . From (4.8) to conclude that log c < (c6 · (u + v ))u+v · pa 2 · 8 log log c In the latter case.17). Thus.13) c By Lemma 4. (4.2 to obtain a lower bound for log b .16) with the usual convention that the empty product is 1.15) ρ c9 ρ < ρ−3 i=1 pi <2· log pi p|abc p=pa . 0 < log c = log b a+b b = g1 · log s1 + . 22 c11 3 < c10 ρ · (r (abc))2 · (log r (abc))12 .Since b > c 2 and c ≥ 3. and the result now follows from (4.4.15). Hence log a ≥ 1 2 p|ac log p · (log r (abc))2 . .13). gv · log sv − f1 · log q1 − · · · − fu · log qu . we deduce that log c 4 log log c By Lemma 4.12) In the former case. Comparing this with the upper bound given by (4.12) we again obtain (4.10) But (4.13) with c 6 replaced by c7 .

16}.Recently the authors improved this estimate. A second estimate is also established. is an ultrametric analog of an Archimedean measure due to E. This result. pb . The method employed to improve the estimate is a p−adic linear independence measure for logarithms of algebraic numbers. Matveev. In particular. b. then c < exp(p · (r (abc)) c12 · log log log r∗ (abc) log log r (abc) 1 . where c12 is an effectively computable constant. In [13]. pc }. c are relatively prime positive integers such that a + b = c and c > 2. if a.M. due to Yu. r ∗ (abc) = max{r (abc). 23 . the better estimate c < exp(c11 · (r (abc)) 3 · (log r (abc))3 where c11 is an effectively computatble positive constant is established. and p = min{pa .

24 .

|b|.1. Theorem 5. The abc Conjecture holds iff lim sup{L} ≤ 1. b.|b|. We will also state the known good triples. we consider the notion of good triples.2) (5. Put k = µ(ε). So L(abc) = ≤ log max{|a|. Proof. under the appropriate hypotheses. |c|} log r (abc) (5. 5.|c|) log r (abc) = log c log r (abc) and asked if these L’s have a bound. he considered L = L(a.1) (5.Chapter 5 Good Triples Associated with the abc Conjecture In this chapter.3) log[µ(ε) · r (abc)1+ε ] by the abc Conjecture log r (abc) log µ(ε) + 1 + ε. 25 . (⇒) Assume the abc Conjecture. c) = log max(|a|. It is this form of the abc Conjecture that we will be using for the topics of this chapter. = log r (abc) Fix ε > 0. namely.1 Preliminaries Recall Oesterl´ e’s version of the abc Conjecture.

and cn = 32 . b. bn = 1.8) log cn ≤ 1 + ε for n large.12) (5. by the hypotheses of the abc Conjecture. and cn .4) (5.6) This holds since. log r (an bn cn ) Then for n > N for some N ∈ Z: cn ≤ r (an bn cn )1+ε . Choose constants µ1 (ε). 2n n (5. . b.9) (5. c) ≤ M ..5) (5.1 our choices for a n .13) (5. µN (ε) such that ci ≤ µi (ε) · r (abc)1+ε for all i Let µ(ε) = max {µi (ε)} 1≤i≤N (5. there exist only finitely many (a.16) . namely: an = 32 − 1. c)’s such that r (a.We want log k ≤ ε for all but finitely many triples (a.14) (5. So..11) Thus cn ≤ µ(ε) · r (an bn cn )1+ε for all n. (5. Recall from Proposition 2.. c) log r (abc) log k ⇔ log r (abc) ≥ ε log k ⇔ r (abc) ≥ M := e ε .15) since r (32 ) ≤ 2 · n 2n log 3 log 3 + log 2 + log (32n − 1) − log 2n 26 32 − 1 . µ2 (ε). for these values Ln = log 32 log r (32n − 1 · 1 · 32n ) n log 32 = log 3 + log r (32n − 1) n log 32 ≥ 2n −1 ) log 3 + log 2 · r ( 3 2n ≥ n n n (5. b. bn . (⇐) Suppose lim sup{L} ≤ 1.10) (5.7) (5. This is true iff lim sup( ⇒ log cn )≤1 log r (an bn cn ) (5.

B. Je. 24.W. b. A.R..1 (Good Triple).N. 1. 9. 11. A. & P.B. & Ju.19) In particular.625991 1.M. L3 > 1.547075 1. If the abc Conjecture holds.R.481322 1. 18. H.491590 1.580756 1.B.N. It is easy to see that the fraction on the RHS of inequality (5. 20. A. 5. We have just shown Theorem 5. & P.W. 2002: Table 5.N. Je.. B.W.b. 5.. Je.482910 1. 10. A. The abc Conjecture holds iff lim sup{L n } = 1. there are only finitely many good triples.W. A. Je. 12.2 Good Triples Definition 5.W. 22. 2.461924 a 2 112 19 · 1307 283 1 73 72 · 412 · 3113 53 13 · 196 318 · 23 · 2269 239 52 · 7937 22 · 11 73 224 112 37 514 · 19 1 72 4 3 · 199 174 · 67 712 2 7 · 52 b 310 · 109 2 3 · 56 · 73 7 · 292 · 318 511 · 132 2 · 37 310 1116 · 132 · 79 29 · 317 · 132 230 · 5 173 · 29 · 318 58 · 173 713 32 · 1310 · 17 · 151 · 4423 213 · 77 · 9412 117 · 19 · 292 39 · 13 215 25 · 3 · 713 316 · 7 210 · 11 · 532 118 219 · 1374 214 · 673 · 461 76 · 41 c 235 2 · 23 28 · 322 · 54 8 8 2 · 3 · 173 54 · 7 211 · 29 2 · 33 · 523 · 953 115 · 17 · 313 · 137 313 · 112 · 31 210 · 52 · 715 210 · 374 218 · 37 · 132 59 · 1396 316 · 1032 · 127 311 · 53 · 73 · 41 211 · 53 38 · 5 117 · 372 · 353 23 · 11 · 23 · 533 34 · 5 8 23 · 5 7 · 7 3 315 · 53 · 13 · 892 313 · 11 · 194 136 21 Discoverer(s) E. cn ) such that Ln > 1.N. 15.B.1: Known Good abc Triples No.17) increases as n gets large. A.255203.18) (5.So Ln ≥ 2n log 3 .465676 1.N.R. & Ju.2 we get Corollary 5. 3.1) of known good abc triples as of January 2. B.2.N. B.. H.B. c) is a good triple if L > 1. & Ju. 13. L 1. B.. B.623490 1.465520 1.N. A. 19.4.N. & Ju.502839 1.17) Thus for n = 3: L3 ≥ 8 · log 3 log 3 + log (38 − 1) − 2 · log 2 ≈ 1.N. So.B.489245 1.N..622912 1. A.W. Table 5.1.B.488865 1.522160 1. we say that a triple (a. B.N.474450 1. 7. by Theorem 5. 23.474137 1. bn . 8. B. 6. A. (5. 21. A. A.536714 1. 14.M.W.. 16.522699 1.N. Hence there are infinitely many triples (a n . The following is the list (Table 5.B.1 continued on next page 27 . & Ju. A.492432 1. For the abc Conjecture. log 3 + log (3 − 1) − (n − 1) · log 2 2n (5.567887 1. A.497621 1.544434 1. Je. 4.471298 1.B. 17.N.

455126 1.B.N. & Ju.N.428908 1. 96. 99.420036 1. 42.426565 1.N.N. 83. 47. 28. 65. 76. 70.B.N. & Ju. & Ju..436180 1. T.421828 1. 56.455673 1. 100. T. A. T. Je. A.416793 a 511 · 31 · 191 2 5 · 17 · 312 · 1699 36 · 512 3 · 109 · 1314 3 2 · 52 225 · 19 1 32 · 116 232 · 315 78 · 2707 136 219 · 13 · 103 35 · 7 3 5 · 73 1 1 2 11 · 43 89 317 4094 32 · 57 · 79 2 · 132 311 · 58 · 4229 219 · 263 22 · 114 · 17 1 32 · 193 25 · 112 · 199 316 · 232 73 · 295 · 1512 313 34 · 232 235 · 72 · 172 · 19 1 1 2 · 135 10 2 ·7 119 · 43 5 2 · 318 53 · 8111 312 221 317 · 67 614 · 149 174 · 793 · 211 227 · 75 211 29 · 192 193 36 · 72 · 13 · 1272 9 2 · 37 · 975 39 · 29 321 732 514 · 11 11 73 614 310 31 3 52 · 11 17 · 194 24 · 59 57 67 · 2633 29 · 373 · 89 78 · 19 313 214 · 310 · 43 · 461 233 194 · 372 72 13 · 3499 56 · 1609 39 · 433 12 b 28 · 713 · 89 · 8592 14 23 · 29 216 · 13 · 594 522 · 89 24 · 173 · 314 3 · 515 · 1033 2 5 · 3 · 52 235 225 · 7 · 1093 210 · 510 · 293 2 · 34 · 74 · 119 · 23 711 56 · 67 213 · 233 · 59 33 · 53 · 77 · 23 3 · 55 · 472 59 · 72 · 134 · 97 7 · 118 221 · 56 · 23 · 7993 221 · 115 · 17 · 19 · 397 229 · 13 58 175 · 235 · 313 83 · 1675 517 · 13577 212 · 53 511 515 · 372 · 47 213 · 292 · 373 24 · 516 · 97 · 919 2 · 17 · 415 315 327 · 1072 24 · 37 · 547 19 · 5093 76 · 1732 57 24 · 236 · 47 · 2772 56 · 710 · 232 1912 · 29 35 · 5 9 76 · 17 · 82092 77 · 113 · 2272 · 547 223 · 13 · 295 229 · 23 · 292 326 · 11 · 19 · 139 39 · 73 · 113 · 19 596 · 73 2 · 56 · 192 · 11932 238 · 61 · 137 55 · 7 · 897 76 · 432 72 · 116 · 199 211 · 114 · 133 36 · 75 · 132 · 251 73 · 1672 115 · 1572 220 · 413 · 832 78 · 23 25 · 510 · 192 53 133 · 14832 33 · 510 · 72 · 293 512 · 19 115 · 132 310 · 59 · 233 39 · 59 · 31 215 · 52 · 372 221 · 54 · 1992 115 · 294 · 83 · 3972 39 · 57 · 31 34 · 514 · 79 217 · 1812 239 29 · 314 · 133 513 · 5323 c 330 · 134 · 277 2 2 · 3 · 11 · 1310 · 47 711 · 47 · 113 23 · 112 · 195 · 974 710 · 257 117 · 133 · 472 74 195 · 13883 319 · 52 · 192 · 29 318 · 114 · 43 57 · 1034 · 2399 311 · 53 · 112 220 53 · 196 213 · 114 · 13 · 41 218 · 79 23 · 3 · 737 220 · 33 · 53 472 · 3075 35 · 75 · 139 117 · 192 3 · 194 232 · 72 · 1093 54 · 297 34 · 239 · 71 35 · 72 · 43 217 · 373 37 · 711 · 743 59 · 114 · 13 327 · 134 3 · 5 7 · 75 215 · 53 · 7 515 · 372 · 2311 5 8 · 72 219 · 34 · 59 313 · 472 38 · 13 514 · 72 · 134 119 · 691 · 1433 219 · 33 · 174 · 2332 25 · 234 · 53 512 · 7432 214 · 57 · 176 38 · 5 · 74 · 734 519 52 · 136 · 434 · 179 29 · 2774 33 · 57 · 72 · 313 39 · 138 511 · 196 320 · 174 · 3323 224 · 13 2 · 138 · 17 311 · 55 · 7 · 17 221 · 234 2 · 314 22 · 310 · 75 322 · 5 · 19 · 167 29 · 5092 3 · 75 · 113 · 412 27 229 · 32 213 · 137 · 613 33 · 112 · 175 215 · 72 · 17 210 · 76 · 132 · 413 1036 3 · 177 78 · 832 · 1307 526 27 · 73 · 13 · 174 28 · 315 · 732 38 · 8092 34 · 511 · 139 15234 27 · 73 · 236 19 Discoverer(s) K.428323 1..B. 79.B. Je.444199 1.N.N.459425 1.421371 1. A.B. & Ju. A. & A. A. A.N.S. 36.446246 1. 41. 64. & Ju. & Ju.S.B. A. T.S.432904 1.. B.443307 1. A. 88. A.426753 1. G. 48. B. Je.W.431815 1.. A.N. 66. & Ju.B.N.V.435006 1. A. A.S. 89. 34.B.B. 32.N. B. 87. 69. Je.450858 1.B. Je.B.433452 1.B.B. A.430176 1. 55. A. 25. 77.421575 1. Je. Je. 82. 60. & Ju.N. A. 91.453343 1.446873 1. Table 5. & Ju. 35. A. 54. Je. 49.N. & P. 97.441519 1.418919 1.N.V. 73.447977 1.443284 1. 94. 74. 62..N. T.V.418233 1.417633 1. 45.B.N. 33.431183 1. 53.. 40.N. Je.N.N. T.B.440264 1.S.427566 1.R.431092 1.B. A.433956 1. T.456203 1. K. 44. A. T.427115 1. B.420232 1.M.431260 1. 92. 58. A.432143 1. 80.B. Je.S.N.455024 1.W. A.454435 1. A.B. Je. A.W. 27.B.442014 1.B.B.N. A. 81. Je. A. 46.N.450026 1.428402 1. 75. A..N.N.N. 30.R. A. H.S.420320 1.F.457790 1.N.B. & A.443502 1. T. A.N. 71. 86. T. A. K. & Ju. A. 90.S.449651 1. & Ju. 59.B.433043 1.429007 1. B. & Ju.B. 61.445064 1.N. A.continued from previous page No.N. 57. Je.N.N. & Ju.B.457066 1.R. A. Je..N.N.N.N. A. A.B.429552 1. 31. A. 85. B. A. 95. & Ju. A.B.439063 1. 39.441814 1. 78. 51. A. A. A. L 1.S.N.. A.457794 1.N.W..431623 1.441441 1. A.N. 26.430418 1. & Ju.B.1 continued on next page 28 .B.N.N.447591 1.438357 1. 84. 50.429873 1.B. A.451344 1.S. & Ju.440969 1.427488 1. 63. 72.447743 1. Je.433464 1.W. A.452613 1.444596 1. Je. 68.N.423381 1. 93.W.N.B. 67. & Ju..N.N.N.B. 52.422083 1. & Ju.B. A.B.B. 37. 98. 43.N. A. 29.B.457482 1.421008 1.N.420437 1.419292 1.N. A. A. Je. A. 38. Je.

& Ju.K. A. A.412893 1. 105.407787 1.416438 1. K.N. 116. 123. A.400588 1. Je.W.402183 1..N. L 1. 104. N.M. A. N. 134. A.413166 1. A. Je. A.414352 1.N. 140. T. T. A. A.N.M.G.B.415633 1.N.N. 142.B. H. Je.K. A. A.N.N.B. 149.401261 1. A.N.N. A.N.403958 1.B. 147.408866 1.413698 1. 133.411682 1. A.R.N.S.E.404484 1..406097 1.406080 1.N.V.405443 1.400262 a 414 · 33941 3 · 37 · 204749 3 · 54 · 599 246 · 23 73 3 · 234 26 · 52 · 713 · 132 · 463 311 · 54 37 · 514 · 72 26 · 5 · 137 52 36 · 1573 · 283 13 · 733 5 793 2 3 · 13 · 1049 13 · 294 672 · 2399 213 · 313 · 113 512 72 212 2 · 7 · 11 · 136 22 · 13 32 · 233 241 39 · 163 79 219 · 3673 216 · 41 · 71 135 · 193 5 · 72 133 224 · 35 631 1 12 5 · 227 9 3 · 103 3 3 · 13 5 · 673 · 1272 · 19219 34 · 19 · 61 · 1732 312 · 56 3 · 514 · 199 233 · 5 310 · 54 · 401 222 36 · 112 · 47 · 3592 229 · 7 234 · 712 134 214 · 313 · 5 518 · 6359 22 b 312 · 197 8 2 · 527 11 · 238 39 · 55 · 117 · 312 · 43 513 · 181 513 · 31 34 · 4312 7 · 116 · 43 251 · 112 314 37 · 133 2310 39 · 55 · 895 311 36 · 7 · 11 · 135 239 · 292 · 107 3 · 710 · 194 313 · 1073 13 · 29 · 436 · 673 22 · 321 · 432 · 52859 835 315 · 192 · 732 · 3343 23 · 434 · 4494 73 · 415 · 181 237 · 2932 212 · 34 · 56 · 1181 23 · 116 · 17 32 · 57 · 133 517 · 197 · 281 315 · 72 2 · 1112 · 1123 · 76081 132 · 433 29 · 372 5 · 195 · 592 226 · 5 · 292 39 · 72 · 197 28 · 3 · 73 · 237 · 41 28 · 112 · 135 · 412 · 47 25 · 11 · 192 · 733 1318 · 37 · 277 244 · 710 79 · 312 72 · 115 · 174 · 41 39 · 76 · 312 · 97 136 · 473 7 · 677 · 137 1713 32 · 312 · 734 · 349 714 · 1231 176 · 463 7 · 296 · 712 32 · 476 · 733 c 223 · 59 · 29 6 13 · 31 · 1034 · 113 222 · 593 1911 · 59 · 7207 24 · 3 · 11 · 132 · 195 2 · 74 · 1993 1112 · 3892 · 6841 217 · 173 295 · 73 · 4192 · 1039 136 28 · 1372 230 · 52 · 112 · 13 219 · 72 · 315 · 467 210 · 173 218 · 433 193 · 1396 25 · 5 · 432 · 1394 26 · 515 520 · 17 710 · 133 · 172 · 1512 22 · 312 · 17 · 109 5 · 413 · 1935 316 · 534 · 972 314 · 5 · 673 215 · 52 · 135 · 312 118 · 134 512 216 · 192 · 67 132 · 2516 197 338 · 397 211 · 38 32 · 5 7 710 · 167 33 · 710 · 37 27 · 57 · 19 11 · 195 · 675 514 · 533 52 · 711 2 · 315 · 72 · 3110 52 · 1494 · 503 · 9293 29 · 115 · 571 230 · 134 112 · 193 · 1274 229 · 312 3 · 54 · 135 · 3532 221 · 54 · 27492 515 · 532 24 · 52 · 112 · 297 221 · 732 119 · 132 · 53 27 · 1910 · 79 2 Discoverer(s) Je.413279 1. & P.B.B.S.407051 1. A.N. 126.R.406420 1. 120.N.. 117. 111. A.N.N. A. A. K.N.B. A.409742 1.V.411615 1.B. A. & A. T. 135.406524 1. 148.R.B.N. Je.N.B.B.407404 1.405785 1.B. 151.402737 1.V.B.N. & Ju. & Ju. G. X. 125..400317 1.401419 1.R...F.415561 1.V. 129. A. A. 113. 110. 139. 141. 118. B. K. & Ju. & Ju. Je.402864 1.N. 106. X.N.401156 1. & P. 152.N. & Ju. 108. 107.401993 1. 144. 115. K. A. 103.410830 1. Je. 150.N.410683 1.N.410044 1.403980 1. 121. A. 146.B.continued from previous page No.415273 1. 101. 130. A.E. A. Discovers of the Known Good abc Triples Initials Je.415090 1. A. A.401291 1.412681 1. A.B.R.R. T.W.N.N. & J. & A.408973 1. A. 127.N.S. A. A.. A.414503 1.B.N. Name(s) Jerzy Browkin and Juliusz Brzezinski Gerhard Frey Noam Elkies and Joe Kanapka Abderrahmane Nitaj Herman te Riele and Peter Montgomery Eric Reyssat Traugott Schulmeiss and Andrej Rosenheinrich Benne M. E.N.B.M.N. 132. A.407208 1. A.403482 1.401979 1.416051 1.N.B.B. A.B. A. A. Je. 122.N. B. A.406079 1. Je.G. & J. & Ju.B. & Ju. 119. & Ju. Je.S.400812 1.404264 1. & Ju. Je. 143. & Ju. 138.. 128. 114. 124. 145. & A.B. & Ju. 131.408577 1.B.N. de Weger Kees Visser Xiao Gang 29 . A.B.416078 1. 137.N.N. 136. 109. H.N. 102. 112.N.N.

6) (No. L = 1.26) (5.h> <sys/time. L = 1. b = 2400. 61) (5. a = 37.5. a = 121. 17) (No. 000. a = 1.20) (5.29) (5. b = 177147. b = 512000.1 were discovered by means of various algorithms. b = 125. b = 4374. 133) 1 (5. 16) (No.21) (5.h> This is a work in process.22) (5. 85) (No. b = 390625. b ≤ 1.4 5. a = 343. Running time for this case was approximately four and one-half days. 50) (No.28) (5. b = 78125. 000 (Further Run)). b = 700928.1.000. ·) refers to the number in Table 5.445064 L = 1.1 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Program Listings abc-mpi. b = 32768. a = 2197.547075 L = 1.412681 L = 1. Saum (July 2002) */ /*--------------------------------------------------------------------------------*/ #include #include #include #include #include #include <stdlib. The interested reader may also see [7] regarding similar unpublished work by Joe Kanapka.3 Computations Regarding Good Triples It seems that the values of the good triples displayed in Table 5. a = 5.h> <mpi.23) (5.435006 (No. a = 338. Initially a program was written in Matlab but was found to be too inefficient. Results (Good Triples for 1 ≤ a.30) 5.488865 L = 1.426565 L = 1.4. 000 and a ≤ b ≤ 100. 1 30 . 31) (No. Knoxville). a = 3.24) (5. 000. b = 255879.25) (No. b ≤ 100.443307 L = 1. Hence a brute force approach was taken to confirm that the stated values were indeed all possible good triples over a particular interval. 5) (No. Note that runs covering even larger intervals are underway and results will be summarized in a future paper. b = 59049.567887 L = 1. a = 1.h> <math. With the aid of Joel Mejeur (now with the Department of Defense) and Michael Saum (University of Tennessee. It is worthwhile to point out that (5.c /*--------------------------------------------------------------------------------*/ /* abc-mpi. 46) (No.h> <stdio.405785 (No. a program was written in C and then run in parallel (using MPI) on a cluster of between 24 and 30 Intel 450 MHz Pentium III computers. 114) (No. a = 7168. With some improvements we have hope of extending the ranges to 10. a = 1.455673 L = 1.h> <getopt. 000 (Initial Run)).22) is the good abc triple with the smallest c value.27) (5. Mejeur (May 2002) */ /* Revised by M. Results (Further Good Triples for 1 ≤ a.482910 L = 1.000.c */ /* */ /* Written by J. Initially the program checked for good triples over the intervals 1 ≤ a ≤ 100.

world.i++) rs[i-1] = log(get_r(i. 1."rank(%d): working on a=%d\n". MPI_Init(&argc.&t_stop)). fprintf(fp. world). double diff_time(struct timeval *. MPI_INT.a<=max. struct timeval *). inline unsigned int get_r(unsigned int n. struct timezone tz_dummy. */ for(a=rank+1. fp=fopen(filename. world = MPI_COMM_WORLD. fflush(stdout). double *rs.&tz_dummy). /* Get command line options */ min=0.i<=2*max+1. 101."Time to gen primes and send = %g\n".&tz_dummy).%d".\n". increments */ /* by number of processors each time. for(i=1. */ /* Primes calculated only on master (rank=0) */ /* and sent to all slave processors via MPI. rank). 100.num_primes)). } /* rs array contains log of radical for each number 1. unsigned int max=10000.out. size. diff_time(&t_start. int num_primes).rank. MPI_STATUS_IGNORE). } /* Main loop. &argv). primes). MPI_STATUS_IGNORE). MPI_Comm_size(world. char **argv) { unsigned int a. break. unsigned int num_primes. fflush(fp).sizeof(double)).a<size. while(1) { c=getopt(argc. case ’m’: max=(unsigned int) atoi(optarg). A good assumption /* is that all processors participating in the MPI VM are of /* same order of magnitude speed wise. primes=calloc(num_primes. No need to send via /* MPI. a. fflush(stdout).. min. 0. "l:m:"). */ */ */ */ */ if(rank==0) { sprintf(filename. MPI_INT. world). char filename[256]. unsigned int *primes). switch(c) { case ’l’: min=(unsigned int) atoi(optarg). argv. c. &size). unsigned int b). i. Each processor starts with different a. MPI_Recv(primes. int main(int argc. */ if(rank==0) { gettimeofday(&t_start. b. "w").2*max+1 /* rs array calculated on each processor. t_stop. sizeof(unsigned int)). printf("Generating list of primes\n"). unsigned int gcd(unsigned int a. unsigned int *primes. num_primes.primes. if(c==-1) break. 1. printf("Found %d primes. /* Set up MPI communication */ MPI_Comm world. 0. world. int rank=0. a). int get_primes(int n. primes=calloc(2*max+1. MPI_Send(primes. rs = calloc(2*max+1. gettimeofday(&t_stop. a. } } else { MPI_Recv(&num_primes. &rank). 100. FILE *fp=NULL.a++) { MPI_Send(&num_primes.15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 struct timeval t_start. unsigned int *primes. num_primes). } } /* Prime numbers calculated up to 2*max + 1 */ /* to ensure that prime factorization can */ /* occur with c (=a+b). MPI_Comm_rank(world.a+=size) { if(rank==0) { fprintf(fp. MPI_UNSIGNED. 31 . MPI_UNSIGNED. double L. num_primes. for(a=1. int count=0. break. "abc. sizeof(unsigned int)). as send traffic could be very large. num_primes=get_primes(2*max+1. 101.

Saum (July 2002) */ /*--------------------------------------------------------------------------------*/ #include <stdlib. fflush(stdout). b. b.4. unsigned int b) { /* recursive routine to calculate gcd(a. d.m */ unsigned int count. } 5.&t_stop)).c /*--------------------------------------------------------------------------------*/ /* util. Mejeur (May 2002) */ /* Revised by M. if(L > 1. fprintf(fp.b++) { /* No need to calculate gcd if both a.b<max. fflush(stdout)."\nTotal Time = %g\n".c */ /* */ /* Written by J. L=log((double)c) / (rs[a-1]+rs[b-1]+rs[c-1]). } } } } } /* Ensure all slave processors are done computing */ MPI_Barrier(world).a.b) */ int r.2 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 util. if(rank==0) { gettimeofday(&t_stop. } /* MPI Cleanup and shutdown */ MPI_Finalize(). rank.b divisible by same small prime */ if( !( ( !(a%2) && !(b%2) ) || ( !(a%3) && !(b%3) ) || ( !(a%5) && !(b%5) ) || ( !(a%7) && !(b%7) ) ) ) { if( gcd(a. L=log((double)c) / (rs[a-1]+rs[b-1]+rs[c-1]). return 0.&tz_dummy). 32 .b++) { /* No need to calculate gcd if both a.h> unsigned int gcd(unsigned int a. if (n==2) { primes[0]=2. diff_time(&t_start. p. if(r==0) return a.h> #include <math. a). if(a<=min) for(b=min. printf("\n\n"). is_unknown. b) == 1) { c=a+b.%f)\n".4) { printf("([%d]:%d. fclose(fp).b divisible by same small prime */ if( !( ( !(a%2) && !(b%2) ) || ( !(a%3) && !(b%3) ) || ( !(a%5) && !(b%5) ) || ( !(a%7) && !(b%7) ) ) ) { if(gcd(a.%f)\n". } } } } if(a>min) for(b=a.108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 fflush(fp). int a. L). r = (b%a). rank. int i.4) { printf("([%d]:%d. } count++.h> #include <stdio. } unsigned int get_primes(int n. return gcd(r.b<=max.%d. unsigned int *primes) { /* Routine taken and altered from Octave’s list_primes.%d. b) == 1) { c = a+b. if(L > 1. L).a. int is_prime.

33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 realloc(primes. unsigned int k.(t_1->tv_sec+t_1->tv_usec/1. therefore set k to just be the number */ int i. i=3. count*sizeof(unsigned int)). d = 3. count=2. } if ( (a*d) == p) { is_prime = 0. } 33 . return (diff). } p+=2.i<num_primes. } inline /* * * unsigned int get_r(unsigned int n. } /*----------------------------------------------------------------------*/ /* FUNCTION diff_time(t1. If k still equals 1. return count. sizeof(unsigned int)). if ( a <= d ) { is_unknown = 0. return 1. k=1. unsigned int *primes. struct timeval * t_2) { double diff.t2) */ /* Returns a decimal value (in secs) of elapsed time between t1 and t2 */ /*----------------------------------------------------------------------*/ double diff_time(struct timeval * t_1. int num_primes) { Check the list of primes to see if it is good. } d = d + 2.0E6)).i++) { if(primes[i] > n) break. } return k. } realloc( primes. count++. primes[1]=3. } if(k==1) { k=n. while(p<n) { is_prime = 1. while(is_unknown) { a = floor ( p / (float)d). } primes[0]=2.0E6) . else if ( (n%primes[i])==0 ) k=k*primes[i]. p=5. is_unknown = 0. } if (is_prime) { primes[count] = p. diff = (double) ((t_2->tv_sec+t_2->tv_usec/1. for(i=0. is_unknown = 1. the number must have been prime.

34 .

Bibliography 35 .

Machiel The abc Conjecture implies Roth’s Theorem and Mordell’s Conjecture. Wiskkd. Andrew ABC allows us to count squarefrees. Further Reading: Joseph H. 99-109 (1991). Math. Silverman. 108 (2001). Soc. Enseign. and Brzezinski. Math. Int. Kunrui On the abc Conjecture. Unsolved Problems in Number Theory. 36 . 25. [7] Guy. No. Res. Ser. E.. no. Serge Old and New Conjectured Diophantine Equations. Mat. Preprint (1994). C. 931-939. and Yu. 1991. Roth’s Theorem and the abc Conjecture. Contemporanea. Andrzej On the Diophantine Equation x! + A = y 2 . Res. No. 1. July 1990. Springer-Verlag. Bulletin of the American Mathematical Society Volume 23. 2. 42. [11] Overholt. Bull. Math. [12] Stewart. IV. Number 206. 7. 37-75. Lond.3-24. 45-72 (1999). Int. C.math. Not. J. abc Implies Mordell. 1998. Math. 3. No. April 1994. [2] Browkin. 76. Richard K. 19. Ser. 104 (1993). Abderrahmane La conjecture abc.L. 225-230(1991).1-2. Marius The Diophantine Equation n! + 1 = m2 . [5] van Frankenhuysen. and Yu.html [10] Nitaj. Ann. 169-181. [9] Nitaj. No.fr/˜ nitaj/abc. 321-324 [4] Elkies. Second Edition. [13] Stewart. Number 1.[1] Bombieri. Mathematics of Computation Volume 62. 991-1009 (1998). [3] Dabrowski. 14. J. J. 291. Duke Math. Kunrui On the abc Conjecture II. Nieuw Arch. http://www. [6] Granville.unicaen. II. New York.L. Math. Noam D. Springer-Verlag.. The Arithmetic of Elliptic Curves. Abderrahmane The abc Conjecture Homepage. Not. No. 1994 [8] Lang.. Some Remarks on the abc Conjecture.

In each of those four years. earned her MBA from the University of Tennessee. 1968 in Wheeling. Tennessee where Jamie is employed by FedEx and Jeffrey is pursuing a doctorate degree in Mathematics at the University of Memphis. 37 . West Virginia. Jeffrey was a finalist for the Dorthea and Edgar Eaves Teaching Award and was recipient of the award in the academic year 2001. In May 1990.Vita Jeffrey Paul Wheeler was born April 22. During his four years. Jeffrey’s wife. Jeffrey took a position as a lecturer at Belmont Technical College in St. Jeffrey graduated from Miami University in Oxford. Jeffrey held the position of graduate teaching associate. Jeffrey entered the graduate program at the University of Tennessee. Ohio and resided in Knoxville. Tennessee. Knoxville. 2000 in Chagrin Falls. He graduated from Linsly School in May 1986. In August 1998. Ohio earning a Bachelor of Arts degree in Mathematics with a minor in Social Work. Knoxville in May 2002 while Jeffrey earned his master’s degree in August 2002. The couple married June 3. Clairsville. After graduation. Currently the couple resides in Memphis. Upon a return to Miami Universtiy. Jamie. He also served in an administrative position. Ohio. Jeffrey met the woman who was to be his wife.