On a Conjecture of Erdos on 3-powerful Numbers
نویسنده
چکیده
A conjecture of P. Erdos says that the diophantine equation x+y = z has infinitely many solutions with (x,y) — 1 and such that if a prime p divides xyz, then p divides xyz. In this paper, we give a proof of this conjecture. Let k ^ 2 be an integer. A /c-powerful number is a positive integer x with the property that if the prime p divides x, then p divides x (2-powerful numbers were called initially powerful numbers by S. W. Golomb [2]). In [1], P. Erdos studied kpowerful numbers and conjectured for k = 3 (see also [3, B 16]. CONJECTURE 1 (Erdos, 1975). The diophantine equation x + y = z has infinitely many solutions in relatively prime 3-powerful numbers. In this paper, we give a proof of this conjecture. In fact, we construct explicitly three infinite families of solutions. Let a be a positive integer. Let x0, y0 and z0 be relatively prime integers such that = az\. Define the sequence of triples (xn,yn,zn) with the recurrences for n ^ 0. By induction, one may show that xn+y 3 n = az\ for all n ^ 0. If we choose a, x0, y0 and z0 such that ( a is cubic-free, xoyo= I(mod2), ' x0 = 2 (mod 3), y0 = 1 (mod 3), a\zl then we have the following. THEOREM. Let a, JC0, y0 andz0 be as above. For n^ I, let the triples (xn, yn, zn) of integers be defined by recurrences (1). Then for all n^ 1, (xn,yn) = 1, a\z\ and Received 1 February 1994. 1991 Mathematics Subject Classification 11D25. Bull. London Math. Soc. 27 (1995) 317-318 318 ABDERRAHMANE NITAJ Proof. It is obvious that for all n ^ 1, xn and yn satisfy xnyn = 0(mod2), xn = 2 (mod 3) and yn = 1 (mod 3). Suppose that (xn,yn) = 1, and let/? be a prime such that p divides (xn+1,yn+1) = (xl + 2yl2xl+yl). Thenp divides 2 (x 3 +2/ n ) (2x 3 + ; ; 3 ) = 3>> and hence p = 3, which contradicts the congruence xl + 2yn = 1 (mod 3). Then (n+iJn+i) = 1 3 zn # 0. Now, since a\z\, we have a |z '+ 1 . Finally, for all n ^ 1, 2 | zn. Hence l im, ,^ \zn\ = + oo. If we start with integers a, JC0, y0 and z0 satisfying (2), and if we put for all n = mm(\xn\\\yn\\a\zf), W,|jn | 3 ,fl |zn | ), (3) .Yn = Zn-Xn, then Xn, Yn and Zn are relatively prime 3-powerful numbers. Thus each of the triplets (Xn, Yn, Zn) gives a solution of Erdos' conjecture. Numerical examples (i) Let a = 6, x0 = 1,805,723, y0 = -2,237,723 and z0 = -960,540. Then (ii) Let a = 9, JC0 = 2 7 1 , y0 = 919 and z0 = 438. Then x 3 + ^ = 2-3-73. (iii) Let a = 12, xQ = 63,963,143, y0 = -26,919,143 and z0 = 27,226,290. Then In these three examples, the integers a, x0, yQ and z0 satisfy (2), and hence, using the recurrence relations (1), we obtain infinitely many solutions for the conjecture of Erdos. We see that in all these solutions, two terms are perfect cubes. For the remaining cases, we propose the following. CONJECTURE 2. The diophantine equationx+y = 2has infinitely many solutions in relatively prime 3-powerful numbers none of which is aperfect cube or only one of which is a perfect cube. References1. P. ERDOS, 'Problems and results on consecutive integers', Eureka 38 (1975-76) 3-8.2. S. W. GOLOMB, 'Powerful numbers', Amer. Math. Monthly 11 (1970) 848-852.3. R. K. GUY, Unsolved problems in number theory (Springer, New York, 1981). Departement de MathematiquesUniversite de Caen14032, Caen CedexFrance
منابع مشابه
On a Combinatorial Problem in Number Theory
1. Introduction One of the problems mentioned injl] (on p. 187) concerns the estimation of the magnitude of the largest number g(N) such that, from every sequence of N distinct natural numbers, one can always select a subsequence of g{N) integers with the property that no sum of two distinct integers of this subsequence belongs to the original sequence. It should be mentioned that the function ...
متن کاملThe Pure Numbers Generated by the Collatz Sequence
The Collatz Conjecture has been vexing mathematicians for over seventy years. It is the only mathematics problem that is comprehensible to a fourth-grader, yet has caused Paul Erdos to say, “Mathematics is not yet ready for such a problem.” This paper divides the integers into “pure” and “impure” numbers, where the “impure” numbers occur in the trajectories of smaller numbers. It develops an in...
متن کاملCircular Averages and Falconer/erdos Distance Conjecture in the Plane for Random Metrics
We study a variant of the Falconer distance problem for perturbations of the Euclidean and related metrics. We prove that Mattila’s criterion, expressed in terms of circular averages, which would imply the Falconer conjecture, holds on average. We also use a diophantine conversion mechanism to prove that well-distributed case of the Erdos Distance Conjecture holds for almost every appropriate p...
متن کاملErdos Distance Problem in Vector Spaces over Finite Fields
The classical Erdos distance conjecture says that the number of distance determined by N points in Rd, d ≥ 2, is at least CN 2 d − . We shall discuss this problem in vector spaces over finite fields where some interesting number theory comes into play. A connection with the continuous analog of the Erdos distance conjecture, the Falconer distance conjecture will be also be established. Universi...
متن کاملErdos Conjecture I
An old conjecture of Paul Erdos [6] states that there exist only 7 integers A. = 4,7,15.21,45,75 and 105 such that the difference A 28 is a prime for all B for which it is at least two. It is known that the conjecture is true for all A < 277 , as Cchiyama and Yorinaga have verified in 1977 ([21]), and in this short paper I show how it is related to other famous unsolved problems in prime number...
متن کامل