Maths Olympiad Prep

Library / /45 of 62

Number theory Difficulty 6.1 National Olympiad Prove it Ukraine

Let P(n)P(n) and Q(n)Q(n) be two polynomials (not constant) with positive integer or zero coefficients. For n1n \ge 1 we determine the sequence xn=2016P(n)+Q(n)x_n = 2016^{P(n)} + Q(n). Prove that there are infinitely many prime numbers pp for which there is a positive integer mm not divisible by the square of any prime, and for which the condition xmpx_m \nmid p holds.
(Oleksiy Klurman)

Solution

Suppose that it is not so. Consider the sequence (xn)(x_n) of all members, the indices of which do not contain in their decomposition on factors the squares of prime numbers. Then in this subsequence there is only a finite number of prime divisors p1,p2,,pkp_1, p_2, \dots, p_k other than 22, 33, 77 (prime factors of the number 20162016). Let n0=i=1kpiαin_0 = \prod_{i=1}^{k} p_i^{\alpha_i}, where αix1\alpha_i \ge x_1. Consider the following sequence:

an=1+nn0ϕ(n0)a_n = 1 + n n_0 \phi(n_0), where ϕ(n)\phi(n) is Euler function, i.e. the function that counts the positive integers up to a given integer nn that are relatively prime to nn. Euler function can be represented in the form of the so-called Euler product: ϕ(n)=npn(11p)\phi(n) = n \prod_{p|n} (1 - \frac{1}{p}), where pp is a prime number.

It is clear that Q(an)Q(1)(modn0)Q(a_n) \equiv Q(1) \pmod{n_0} and P(an)P(1)(modϕ(n0))P(a_n) \equiv P(1) \pmod{\phi(n_0)}. As by the construction (n0,2016)=1(n_0, 2016) = 1, hence xan=2016P(an)+Q(an)x1(modn0)x_{a_n} = 2016^{P(a_n)} + Q(a_n) \equiv x_1 \pmod{n_0}.

Later we will show that for large n1n \ge 1 and for each pip_i, i=1,ki = 1, k the sequence of values ordpi(xan)\text{ord}_{p_i}(x_{a_n}) is limited. We can choose an infinite number of primes from the sequence (an)(a_n), which follows from Dirichlet's theorem or from using simple arguments about the progression with the first member 11. Next, let us consider only those nn for which corresponding ana_n is a prime number. Then we have that there are infinitely many values n1n \ge 1:

2016P(n)+Q(n)=2a3b7cM,(1) 2016^{P(n)} + Q(n) = 2^a \cdot 3^b \cdot 7^c \cdot M, \quad (1)
for a fixed limited MM. But then
M(max{2a;3b;7c})3M2a3b7c=2016P(n)+Q(n)>2016n. M \cdot (\max\{2^a; 3^b; 7^c\})^3 \ge M \cdot 2^a \cdot 3^b \cdot 7^c = 2016^{P(n)} + Q(n) > 2016^n.
Then from one of these prime numbers AA (which is the largest in the previous transition) in equation (1) we have that is 2016P(n)2016^{P(n)} divisible by A13nC0A^{\frac{1}{3}n-C_0}. The right-hand side is also divisible by A13nC0A^{\frac{1}{3}n-C_0}, hence Q(n)Q(n) should be divisible by the same power. However, for sufficiently large values nn it is not possible, because Q(n)Q(n) is a power function, while A13nC0A^{\frac{1}{3}n-C_0} is exponential.

Want a route through all this instead of an archive? The track puts 2,000 problems in a working order, from AMC 10 level to the IMO shortlist.

Source: MathNet, licensed CC-BY-4.0. Statement reproduced verbatim; metadata (topic, difficulty) added by this project.