Let a, b, c be fixed coprime positive integers with min{a, b, c} > 1. In this paper, by analyzing the gap rule for solutions of the ternary purely exponential diophantine equation a x + b y = c z , we prove that if max{a, b, c} ≥ 10 62 , then the equation has at most two positive integer solutions (x, y, z).(x, y, z) of (1.1). As a straightforward consequence of an upper bound for the number of solutions of binary S−unit equations due to F. Beukers and H. P. Schlickewei [2], we have N(a, b, c) ≤ 2 36 . In nearly two decades, many papers investigated the exact values of N(a, b, c). The known results showed that (1.1) has only a few solutions for some special cases(see the references of [5] and [6] ).Recently, Y.-Z. Hu and M.-H. Le [5,6] successively proved that (i) if a, b, c satisfy certain divisibility conditions and max{a, b, c} is large enough, then (1.1) has at most one solution (x, y, z) with min{x, y, z} > 1. (ii) If max{a, b, c} > 5 × 10 27 , then N(a, b, c) ≤ 3. R. Scott and R. Styer [9] proved that if 2 ∤ c, then N(a, b, c) ≤ 2. The proofs of the first two results are using the Gel'fond-Baker method with an elementary approach, and the proof of the last result is using some elementary algebraic number theory methods. In this paper, by analyzing the gap rule for solutions of (1.1) along the approach given in [6], we prove a general result as follows:Notice that, for any positive integer k with k > 1, if (a, b, c) = (2, 2 k − 1, 2 k + 1), then (1.1) has solutions (x, y, z) = (1, 1, 1) and (k + 2, 2, 2). It implies that there exist infinitely many triples (a, b, c) which make N(a, b, c) = 2. Therefore, in general, N(a, b, c) ≤ 2 should be the best upper bound for N(a, b, c).
PreliminariesLemma 2.1 Let t be a real number. If t ≥ 10 62 , then t > 6500 6 (log t) 18 .Proof. Let F (t) = t − 6500 6 (log t) 18 . Then we have F ′ (t) = 1 − 18 × 6500 6 (log t) 17 /t and F ′′ (t) = 18×6500 6 (log t) 16 (log t−17)/t 2 , where F ′ (t) and F ′′ (t) are the derivative and the second derivative of F (t). Since F ′ (10 62 ) > 0 and F ′′ (t) > 0 for t ≥ 10 62 , we get F ′ (t) > 0 for t ≥ 10 62 . Further, since F (10 62 ) > 0, we obtain F (t) > 0 for t ≥ 10 62 . The lemma is proved.Let α be a fixed positive irrational number, and let α = [a 0 , a 1 , . . . ] denote the simple continuous fraction of α. For any nonnegative integer i, let p i /q i be the i−th convergent of α. By Chapter 10 of [7], we obtain the following two lemmas immediately.Lemma 2.2 (i) The convergents p i /q i (i = 0, 1, . . . ) satisfy p −1 = 1, p 0 = a 0 , p i+1 = a i+1 p i + p i−1 , q −1 = 0, q 0 = 1, q i+1 = a i+1 q i + q i−1 , i ≥ 0.