Maths Olympiad Prep

Library / /230 of 383

, 2011

Number theory Difficulty 8.7 Shortlist Prove it IMO

Let P(x)P(x) and Q(x)Q(x) be two polynomials with integer coefficients such that no nonconstant polynomial with rational coefficients divides both P(x)P(x) and Q(x)Q(x). Suppose that for every positive integer nn the integers P(n)P(n) and Q(n)Q(n) are positive, and 2Q(n)12^{Q(n)}-1 divides 3P(n)13^{P(n)}-1. Prove that Q(x)Q(x) is a constant polynomial.

Solution

First we show that there exists an integer dd such that for all positive integers nn we have gcd(P(n),Q(n))d\operatorname{gcd}(P(n), Q(n)) \leq d.
Since P(x)P(x) and Q(x)Q(x) are coprime (over the polynomials with rational coefficients), Euclid's algorithm provides some polynomials R0(x),S0(x)R_{0}(x), S_{0}(x) with rational coefficients such that P(x)R0(x)Q(x)S0(x)=1P(x) R_{0}(x)- Q(x) S_{0}(x)=1. Multiplying by a suitable positive integer dd, we obtain polynomials R(x)=dR0(x)R(x)= d \cdot R_{0}(x) and S(x)=dS0(x)S(x)=d \cdot S_{0}(x) with integer coefficients for which P(x)R(x)Q(x)S(x)=dP(x) R(x)-Q(x) S(x)=d. Then we have gcd(P(n),Q(n))d\operatorname{gcd}(P(n), Q(n)) \leq d for any integer nn.

To prove the problem statement, suppose that Q(x)Q(x) is not constant. Then the sequence Q(n)Q(n) is not bounded and we can choose a positive integer mm for which
M=2Q(m)13max{P(1),P(2),,P(d)}. \begin{equation*} M=2^{Q(m)}-1 \geq 3^{\max \{P(1), P(2), \ldots, P(d)\}} . \tag{1} \end{equation*}
Since M=2Q(n)13P(n)1M=2^{Q(n)}-1 \mid 3^{P(n)}-1, we have 2,3M2,3 \nmid M. Let aa and bb be the multiplicative orders of 22 and 33 modulo MM, respectively. Obviously, a=Q(m)a=Q(m) since the lower powers of 22 do not reach MM. Since MM divides 3P(m)13^{P(m)}-1, we have bP(m)b \mid P(m). Then gcd(a,b)gcd(P(m),Q(m))d\operatorname{gcd}(a, b) \leq \operatorname{gcd}(P(m), Q(m)) \leq d. Since the expression axbya x-b y attains all integer values divisible by gcd(a,b)\operatorname{gcd}(a, b) when xx and yy run over all nonnegative integer values, there exist some nonnegative integers x,yx, y such that 1m+axbyd1 \leq m+a x-b y \leq d.

By Q(m+ax)Q(m) (mod a)Q(m+a x) \equiv Q(m)\ (\bmod\ a) we have
2Q(m+ax)2Q(m)1(mod M) 2^{Q(m+a x)} \equiv 2^{Q(m)} \equiv 1 \quad(\bmod\ M)
and therefore
M2Q(m+ax)13P(m+ax)1. M\left|2^{Q(m+a x)}-1\right| 3^{P(m+a x)}-1 .
Then, by P(m+axby)P(m+ax)(mod b)P(m+a x-b y) \equiv P(m+a x)(\bmod\ b) we have
3P(m+axby)3P(m+ax)1(mod M). 3^{P(m+a x-b y)} \equiv 3^{P(m+a x)} \equiv 1 \quad(\bmod\ M) .
Since P(m+axby)>0P(m+a x-b y)>0 this implies M3P(m+axby)1M \leq 3^{P(m+a x-b y)}-1. But P(m+axby)P(m+a x-b y) is listed among P(1),P(2),,P(d)P(1), P(2), \ldots, P(d), so
M<3P(m+axby)3max{P(1),P(2),,P(d)} M<3^{P(m+a x-b y)} \leq 3^{\max \{P(1), P(2), \ldots, P(d)\}}
which contradicts (1).

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.