IJMMS 29:4 (2002) 239244 PII. S0161171202004696 http://ijmms.hindawi.com Hindawi Publishing Corp.

ON THE EQUATION x 2 + 2a 3b = y n
FLORIAN LUCA
Received 1 December 1999 and in revised form 15 January 2000

We nd all positive integer solutions (x, y, a, b, n) of x 2 + 2a 3b = y n with n 3 and coprime x and y. 2000 Mathematics Subject Classication: 11D41, 11D61.

1. Introduction. The Diophantine equation x 2 + C = y n , where x and y are positive integers, n 3 and C is a given integer, has received considerable interest. The earliest reference seems to be an assertion by Fermat that he had shown that when C = 2, n = 3, the only solution is given by x = 5, y = 3; a proof was published by Euler in 1770. The rst result for general n is due to Lebesgue [9] who proved that there are no solutions for C = 1. Ljunggren [10] solved this equation for C = 2, Nagell [13, 14] solved it for C = 3, 4, and 5 and Chao [5] proved that it has no solutions for C = 1. For an extensive list of references one should consult Cohns beautiful paper [6] in which he develops a method by which he nds all solutions of the above equation for 77 of the values of C 100. This equation was later solved for two additional values of C 100 (namely, C = 74 and C = 86) by Mignotte and de Weger [12]. It is interesting to mention that the equation x 2 + 7 = y n is still unsolved. In recent years, a dierent form of the above equation has been considered, namely, when C is no longer a xed integer but a power of a xed prime. Le [8] investigated the equation x 2 +2m = y n . Arif and Muriefah solved the equation x 2 +3m = y n when m is odd (see [2]). They also gave partial results in the case when m is even (see [1]) but the general solution in the case m is even was found by Luca in [11]. For any nonzero integer k, let P (k) be the largest prime dividing k. Let C1 be any xed positive constant. It follows, from the work of Bugeaud [4] and Turk [15], that if x2 + z = y n with (x, y) = 1, P (z) < C1 , (1.1)

then max(|x|, |y|, n) is bounded by a constant computable in terms of C1 alone. In this paper, we nd all solutions of (1.1) when C1 = 5 and z > 0. More precisely, we nd all solutions of the equation x 2 + 2a 3b = y n with a, b 0, n 3, (x, y) = 1. (1.2)

The proof uses the new result on the existence of primitive divisors of the Lucas numbers due to Bilu et al. [3] as well as a computational result of de Weger [7].

240 2. The result

FLORIAN LUCA

Theorem 2.1. All positive solutions of the equation x 2 + 2a 3b = y n with a, b 0, n 3, (x, y) = 1 (2.1)

have n = 4 or n = 3. For n = 4, the solutions are (x, y) = (7, 3), (23, 5), (7, 5), (47, 7), (287, 17). For n = 3, the solutions are (x, y) =(5, 3), (11, 5), (10, 7), (17, 7), (46, 13), (35, 13), (595, 73), (955, 97), (2681, 193), (39151, 1153). (2.3) (2.2)

In the statement of the theorem we have listed only the values of x, y, and n as the values of the parameters a and b that can be determined from the prime factor decomposition of x 2 y n once x, y, and n are given. From Lebesgues result, we know that the equation x 2 + 1 = y n has no positive solutions for n 3 and from the work of Arif, Muriefah, and Luca, we know that the only positive solutions of the equation x 2 +3m = y n with (x, y) = 1 are (x, y, m, n) = (10, 7, 5, 3) and (46, 13, 4, 3). From now on, we assume that a > 0. In particular, both x and y are odd. 3. The case n 3 or 4. In this section, we show that it suces to assume that n {3, 4}. Indeed, assume that n 4. We may certainly assume that n is an odd prime. If n 3, it follows that n 5. Write 2a 3b = dz2 where d {1, 2, 3, 6}. Equation (2.1) can be written as x + i dz x i dz = y n . (3.1)

Since x is odd and dz2 is even, it follows that the two ideals [(x + i dz)] and [(x i dz)] are coprime in the ring of integers of Q (i d). Since the class number of Q (i d) is 1 or 2 and n 5 is prime, it follows that there exists an integer u and a root of unity in Q (i d) such that x + i dz = un , x i dz = un . (3.2)

Since is a root of unity belonging to a quadratic extension of Q , it follows that k = 1 for some k {1, 2, 3, 4, 6}. Since n 5 is prime, it follows that up to a substitution one may assume that = 1 in system (3.2). From (3.2) with = 1, it follows that 2i dz = un un . Since certainly un un Z, uu (3.4) (3.3)

ON THE EQUATION x 2 + 2a 3b = y n

241

we have that P un un uu < 5. (3.5)

From (3.5), we nd that the Lucas number given by formula (3.4) has no primitive divisor. From [3], it follows that there are at most 10 pairs (u, n) satisfying inequality (3.5) and all of them appear in [3, Table 1]. A quick investigation reveals that none of the us from [3, Table 1] belongs to Q (i d) for some d {1, 2, 3, 6}, which is the desired contradiction. 4. The case n = 4. Let S = {k | P (k) < 5}. Then, we have the following preliminary result. Lemma 4.1. All solutions of the equation x2 = k l are (x, k, l) =(1, 2, 1), (2, 3, 1), (3, 8, 1), (5, 24, 1), (7, 48, 1), (17, 288, 1), (1, 4, 3), (1, 9, 8), (5, 16, 9), (5, 27, 2), (7, 81, 32). (4.2) with k, l > 0, k, l S, (k, l) = 1 (4.1)

Proof of Lemma 4.1. This lemma is a particular case of a result of de Weger [7, Chapter 7]. The proof of the theorem for n = 4. Rewrite (2.1) as y2 x y 2 + x = 2a 3b . (4.3)

Since a > 0 and (x, y) = 1, it follows that (y 2 x, y 2 + x) = 2. Thus, y 2 x = k, Hence, y2 = k l + , 2 2 (4.5) y 2 + x = l, with k, l > 0, k, l S, (k, l) = 2. (4.4)

where k/2, l/2 S are positive and coprime. By Lemma 4.1, we obtain that (4.5) has only 6 solutions. Five of them lead to solutions (2.2) of (2.1). One of the solutions of (4.5) leads to 22 + 2 2 3 = 2 4 , (4.6)

which is not a convenient solution of (2.1) because x = 2 and y = 2 are not coprime. The case n = 4 is therefore settled. 5. The case n = 3. We begin with another lemma. Lemma 5.1. The only solutions of the equation 3x 2 = k l with k, l > 0, k, l S, (k, l) {1, 3} (5.1)

242 are

FLORIAN LUCA

(x, k, l) =(1, 2, 1), (1, 4, 1), (1, 6, 3), (2, 9, 3), (3, 24, 3), (5, 72, 3), (7, 144, 3), (17, 864, 3), (1, 12, 9), (1, 27, 24), (5, 48, 27), (5, 81, 6), (7, 243, 96). Proof of Lemma 5.1. This lemma too is a particular instance of the more general computation of de Weger [7, Chapter 7]. The proof of the theorem for n = 3. Write again 2a 3b = dz2 where d {1, 2, 3, 6}. From arguments employed in Section 3, we know that there exist u and in Q (i d) such that y = |u|2 , is a root of unity and x + i dz = u3 , Clearly, 2i dz = u3 u3 . We distinguish two cases. Case 1 ( = 1). Equation (5.4) reads 2i dz = u3 u3 . Assume rst that u = a + ib d with a and b integers. Then, we get 2i dz = a + ib d or 2i dz = 2i db 3a2 db2 . Hence, b | z and 3a2 = db2 z . b (5.8) (5.7)
3

(5.2)

x i dz = u3 .

(5.3)

(5.4)

(5.5)

a ib d

(5.6)

Let k = db2 and l = z/b. Notice that k, l S. Moreover, notice that (k, l) {1, 3}. Indeed, if (k, l) {1, 3}, it follows that there exists a prime p such that p | (k, l, a). In particular, p | db2 and p | a, therefore p | a2 + db2 = y. Since p | z and 2a 3b = dz2 , we come to p | 2a 3b . It follows now that p | (y 3 2a 3b ) = x 2 and therefore p | x. This contradicts the fact that x and y are coprime. Now all solutions of (5.8) are given by Lemma 5.1. For example, the solution 3 12 = 21 + 1 (5.9)

gives either a = 1, d = 2, b = 1, and z = 1 or a = 1, d = 1, b = 1, and z = 2. The rst possibility yields y = a2 + db2 = 1 + 2 = 3 and dz2 = 2, which leads to the solution 33 = 2 + 52 of (2.1). The second possibility gives y = a2 + db2 = 2 and dz2 = 4, which leads to the solution 23 = 22 + 22 of (2.1). This is not an acceptable solution, since x = 2 and y = 2 are not coprime.

ON THE EQUATION x 2 + 2a 3b = y n

243

All the solutions of (2.1) for the case n = 3 except for (x, y) = (10, 7) are obtained in this way by identifying a, b, d, and z from (5.8) via Lemma 5.1. When d = 3, we also need to investigate the case in which a + i 3b u= 2 for some odd integers a and b. From (5.5), we simply get that 16i 3z = a + i 3b or 16i 3z = 2i 3b 3a2 3b2 . It follows that b divides z and 3a2 = 3b2 8z . b (5.13) (5.12)
3

(5.10)

a i 3b

(5.11)

From Lemma 5.1, we derive that (5.13) has only two convenient solutions, namely, 312 = 332 83 and 372 = 334 812. These lead to the solutions (x, y) = (10, 7) and (595, 73) of (2.1). Case 2 ( 1). It is easy to see that the only case in which one may not be able to set = 1 in system (5.4) is when d = 3. In this case, one may assume that = (1 + i 3)/2 and that u = (a + i 3b)/2 for some integers a and b such that a b (mod 2). Then (5.4) becomes 2i 3z = a + i 3b 1+i 3 2 2
3

a i 3b 1i 3 2 2

(5.14)

This equation is equivalent to 16z = a3 + 3a2 b 9ab2 3b3 . Assume rst that both a and b are odd. Then, from (5.15), it follows that 16z = a3 ab2 + 3a2 b 3b3 8ab2 = a2 b2 (a + 3b) 8ab2 . (5.16) (5.15)

Since a and b are both odd, we obtain that 16 | (a2 b2 )(a + 3b). Equation (5.16) forces 16 | 8ab2 , which is impossible. Assume now that both a and b are even. Since y = (a/2)2 +3(b/2)2 is odd, it follows that exactly one of the numbers a/2 and b/2 is even. Equation (5.15) now implies that 2z = a 2
3

+3

a 2

a b 9 2 2

b 2

b 2

(5.17)

However, (5.17) is now impossible, because precisely one of the numbers a/2 and b/2 is even and the other one is odd. Hence, this case can never occur. The theorem is therefore completely proved.

244

FLORIAN LUCA

Acknowledgements. This work was supported in part by the Alexander von Humboldt Foundation and in part by the programme KONTAKT ME 148 of the Czech Republic. This paper was written when I visited the Mathematical Institute of the Czech Academy of Sciences in April, 1999. I would like to thank Michal Kek both for the invitarz tion and for helpful conversation on the topic as well as the institute for its hospitality. Michal also wrote a computer program which pointed out two solutions of (2.1) that I had originally missed. I thank him for that too. Finally, I would also like to thank Yuri Bilu for sending me a copy of the preprint [3], an anonymous referee for making me aware of de Wegers work [7] and Benne de Weger for allowing me to consult his personal copy of [7]. References
[1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] F. S. Abu Muriefah and S. A. Arif, On a Diophantine equation, Bull. Austral. Math. Soc. 57 (1998), no. 2, 189198. S. A. Arif and F. S. Abu Muriefah, The Diophantine equation x 2 + 3m = y n , Int. J. Math. Math. Sci. 21 (1998), no. 3, 619620. Y. Bilu, G. Hanrot, and P. M. Voutier, Existence of primitive divisors of Lucas and Lehmer numbers, preprint, 1999. Y. Bugeaud, Sur le plus grand facteur premier de ax m + by n , C. R. Acad. Sci. Paris Sr. I Math. 326 (1998), no. 6, 661665 (French). K. Chao, On the Diophantine equation x 2 = y n +1, xy = 0, Sci. Sinica 14 (1965), 457460. J. H. E. Cohn, The Diophantine equation x 2 + C = y n , Acta Arith. 65 (1993), no. 4, 367 381. B. M. M. de Weger, Algorithms for Diophantine Equations, Stichting Mathematisch Centrum Centrum voor Wiskunde en Informatica, Amsterdam, 1989. M. Le, Diophantine equation x 2 + 2m = y n , Chinese Sci. Bull. 42 (1997), no. 18, 1515 1517. V. A. Lebesgue, Sur limpossibilit en nombres entiers de lquation x m = y 2 + 1, Nouv. Ann. Math. 9 9 (1850), 178181 (French). W. Ljunggren, ber einige Arcustangensgleichungen die auf interessante unbestimmte Gleichungen fhren, Ark. Mat. 29A (1943), no. 13, 111 (German). F. Luca, On a Diophantine equation, Bull. Austral. Math. Soc. 61 (2000), no. 2, 241246. M. Mignotte and B. M. M. de Weger, On the Diophantine equations x 2 + 74 = y 5 and x 2 + 86 = y 5 , Glasgow Math. J. 38 (1996), no. 1, 7785. T. Nagell, Sur limpossibilit de quelques quations a deux indetermines, Norsk. Mat. Forensings Sknifter 13 (1923), 6582 (French). , Contributions to the theory of a category of Diophantine equations of the second degree with two unknowns, Nova Acta Soc. Sci. Upsal. ser (4) 16 (1955), no. 2, 38. J. Turk, On the dierence between perfect powers, Acta Arith. 45 (1986), no. 4, 289307.

Florian Luca: Mathematical Institute, UNAM, Ap. Postal 61-3 (Xangari), CP 58 089 Morelia, Michoacn, Mexico E-mail address: fuca@matmor.unam.mx