Diophantine equations in separated variables and polynomial power sums

We consider Diophantine equations of the shape \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ f(x) = g(y) $$\end{document}f(x)=g(y), where the polynomials f and g are elements of power sums. Using a finiteness criterion of Bilu and Tichy, we will prove that under suitable assumptions infinitely many rational solutions (x, y) with a bounded denominator are only possible in trivial cases.

polynomials are polynomials of the shape c 1 x e 1 +· · ·+c l x e l +c l+1 for a fixed number l of nonconstant terms where the c i and e i may vary with the only restriction that the e i must be pairwise distinct. Kreso proved that under some assumptions on the exponents e i and if g is indecomposable, then f (x) = g(y) has infinitely many solutions with a bounded denominator if and only if f = g • μ for a linear polynomial μ.
Dujella and Tichy proved in [6] the finiteness of the number of integer solutions for the situation when f , g are generalized Fibonacci polynomials. Moreover, Dujella and Gusic [5] as well as Stoll [10] considered families of polynomials parametrized by two parameters and a binary recurrence relation. Beyond this the case of truncated binomial polynomials was considered in [4] by Dubickas and Kreso, sums of products of consecutive integers are considered in [1] by Bazso et al., and Bernoulli and Euler polynomial related families in [9] by Pinter and Rakaczki.
In the present paper we are also considering Diophantine equations of the type f (x) = g(y). Here we are going to assume that the polynomials f and g come from polynomial power sums, i.e. simple linear recurrence sequences of polynomials. We remark that polynomial power sums can be seen as a variant of lacunary polynomials since its Binet representation has a fixed number of summands.

Results
as well as constants a 1 , . . . , a d ∈ Q be the n-th polynomial in a linear recurrence sequence of polynomials satisfying the dominant root condition deg α 1 > max i=2,...,d deg α i and having at most one constant characteristic root. Assume furthermore that G n (x) cannot be written in the form a 1 α 1 (x) n + a 2 α 2 n for α 1 (x) ∈ Q[x] a perfect power of a linear polynomial and a 1 , a 2 , α 2 ∈ Q. We will refer to the assumptions in this paragraph by saying G n (x) = a 1 α 1 (x) n + · · · + a d α d (x) n is the n-th polynomial in a linear recurrence sequence of the required shape.
We call a polynomial f of degree deg f ≥ 2 decomposable if it can be written in the form f = g • h for polynomials g, h satisfying deg g ≥ 2 and deg h ≥ 2. Here • denotes the composition of functions. If such a decomposition does not exist, then we call the polynomial f indecomposable.
Furthermore, we say that an equation f (x) = g(y) has infinitely many rational solutions with a bounded denominator if there exists a positive integer z such that f (x) = g(y) has infinitely many solutions (x, y) ∈ Q 2 with zx, zy ∈ Z.
Our main result is the following theorem. In Remark 1 below we give a possibility how to generalize it to arbitrary number fields: be the n-th polynomial in a linear recurrence sequence of the required shape. Analogously, let H m (y) = b 1 β 1 (y) m + · · · + b t β t (y) m be the m-th polynomial in a linear recurrence sequence of the required shape. Moreover, assume that n, m > 2. If G n (x) is indecomposable, then the equation in separated variables

has infinitely many rational solutions with a bounded denominator if and only if there exists a polynomial P(y) ∈ Q[y] such that H m (y) = G n (P(y)) holds identically. If in addition H m (y) is also indecomposable, then in the above statement we can restrict P(y) to be linear.
We exclude the case when G n or H m has exactly one constant and one nonconstant characteristic root, where the nonconstant one is (a perfect power of) a linear polynomial, since the conclusion is not true in general in this situation. Consider We leave it up to the reader to verify that all assumptions of the theorem are satisfied. One can check that the identity H m (y) = G n (P(y)) holds for the polynomial P(y) = y 2 − 1. Therefore, by Theorem 1, we have infinitely many rational solutions with a bounded denominator. If we consider G 3 (x) from above and then we get deg G 3 = 6 as well as deg H 7 = 14. Hence deg G 3 does not divide deg H 7 and therefore there is no polynomial P such that H m (y) = G n (P(y)). By Theorem 1 we cannot have infinitely many rational solutions with a bounded denominator.
Note that we can check whether there exists a polynomial P(y) such that H m (y) = G n (P(y)) holds a priori. To do so we first determine deg P by the equality deg H m = deg G n · deg P. If this equation has no solution in positive integers, then there is no such polynomial P. Otherwise we start with a polynomial P of the given degree and unknown coefficients. By a comparison of coefficients we determine step by step (starting with the leading coefficient) the values for the coefficients of P. If we end up in a contradiction, then there is no such polynomial P. In the case that there is no contradiction we have found a polynomial with the sought property. We remark that in the case that there are only finitely many solutions our result is ineffective in the sense that we do not find all the solutions (for a given common denominator).
Note that G n and H m can be elements of different linear recurrence sequences, but they could also be elements of the same linear recurrence sequence. We do neither require nor exclude the situation G n (x) = G m (y) for n = m if the assumptions of our theorem are satisfied.
Furthermore, we remark that the polynomial of the second linear recurrence sequence H m can be replaced by an arbitrary fixed polynomial h(y) ∈ Q[y]. If we replace all assumptions about H m by the two assumptions that deg h > 4 and that h is not of the shape h(y) = a(cy + d) k + b for rational numbers a, b, c, d, then the same result as in Theorem 1 holds. The proof is completely analogous to the below given proof of Theorem 1.

Preliminaries
The proof of our theorem uses a criterion of Bilu and Tichy [3], for which the following terminology of so-called standard pairs is needed.
In our notation, k and l are positive integers, a and b are non-zero rational numbers and p(x) is a non-zero polynomial with coefficients in Q. We denote by D k (x, a) the k-th Dickson polynomial which is defined by the equation Using this notation we have the following five kinds of standard pairs (over Q); in each of them the two coordinates can be switched: A standard pair of the Our main tool is now the following theorem which is proven as Theorem 1.1 in [3] by Bilu and Tichy:

Theorem 2 Let f (x), g(x) ∈ Q[x] be non-constant polynomials. Then the following two assertions are equivalent: (a) The equation f (x) = g(y) has infinitely many rational solutions with a bounded
denominator. and ( f 1 (x), g 1 (x)) is a standard pair over Q such that the equation f 1 (x) = g 1 (y) has infinitely many rational solutions with a bounded denominator.

Proof
All necessary preparations that are needed for the proof of our theorem are finished. So we can start with the proof: Proof of Theorem 1 First note that by the dominant root condition we have the bounds deg α 1 ≥ 1 and deg G n = n deg α 1 > 2. Analogously, the bound deg H m = m deg β 1 > 2 holds. The next important observation is that we can neither have deg α 1 = 1 nor deg β 1 = 1. Otherwise, if deg α 1 = 1, then G n (x) would have exactly two characteristic roots and one of them would be constant. This shape is forbidden by the conditions of the theorem. The argument for deg β 1 is the same. Now assume that Eq. (1) has infinitely many rational solutions with a bounded denominator. Thus, by Theorem 2, we have and a standard pair (g(x), h(x)).
From here on we distinguish between two cases. In the first case we assume that deg ϕ = 1.
Then (g(x), h(x)) cannot be a standard pair of the first kind. Otherwise we would either have Moreover, (g(x), h(x)) cannot be a standard pair of the second kind since we have deg G n > 2 and deg H m > 2.
If (g(x), h(x)) is a standard pair of the third kind, then we get Since G n (x) is indecomposable and Dickson polynomials have the composition property (D l (x, a), a l ) the index p in (2) must be a prime. Hence together with n > 2 implies deg α 1 = 1. As shown above this is a contradiction. Therefore (g(x), h(x)) cannot be a standard pair of the third kind. Also, (g(x), h(x)) cannot be a standard pair of the fourth kind since otherwise with an even k would contradict the fact that G n (x) is indecomposable. If H m (y) is indecomposable, then q(y) is linear. Thus by construction P(y) is linear, too.
Conversely, if we assume the identity G n (P(y)) = H m (y), then Eq. (1) obviously has infinitely many rational solutions with a bounded denominator.

Remark 1
We remark that if we utilize Theorem 10.5 in [3] instead of Theorem 1.1, then we can replace Q by an arbitrary number field K and get for a finite set S of places of K , containing all archimedean ones, the analogous result as above for infinitely many solutions with a bounded O S -denominator.
Funding Open access funding provided by Paris Lodron University of Salzburg.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.