Maths Olympiad Prep

Track / Stage 7 / 107 of 300 #1987 of 2444

Problem 1987

National Olympiad second round; IMO P1/P4
Number theory Difficulty 7.3 Prove it Estonian Math Competitions · Estonia

Find all polynomials P(x)P(x) with integral coefficients and the following property: for any pair (u,v)(u, v) of positive integers, gcd(u,v)=1\text{gcd}(u, v) = 1 implies gcd(P(u),P(v))=1\text{gcd}(|P(u)|, |P(v)|) = 1.

This one wants a proof. Work it on paper, then read the official solution and mark yourself. Be honest about it: the record is only any use to you if it is.

Next problem →

Official solutions — 2

Solution 1

Answer: All polynomials of the form P(x)=±xlP(x) = \pm x^l where ll is a non-negative integer.

Firstly, we show that all prime divisors of P(n)P(n) are divisors of nn. Suppose that there is a prime number qq dividing P(n)P(n) but not dividing nn. Obviously P(n+q)P(n)0(modq)P(n+q) \equiv P(n) \equiv 0 \pmod{q}; but then gcd(n,n+q)=1\text{gcd}(n, n+q) = 1 while gcd(P(n),P(n+q))q>1\text{gcd}(|P(n)|, |P(n+q)|) \ge q > 1, contradicting the assumption.

Next we show that only polynomials of the form P(x)=±xlP(x) = \pm x^l where l0l \ge 0 satisfy the conditions of the problem. We proceed by induction on the degree of P(x)P(x). If P(x)P(x) is a constant polynomial then the constant must be coprime with itself in order to satisfy the conditions of the problem. This is possible only if P(x)=±1=±x0P(x) = \pm 1 = \pm x^0. Assume now that P(x)P(x) is a non-constant polynomial of degree ll. Then P(q)>1|P(q)| > 1 for infinitely many prime numbers qq. In each such case, namely qq must divide P(q)P(q). Hence the constant term of P(x)P(x) must be divisible by qq. As the constant term does not depend on qq, it must be divisible by infinitely many prime numbers. Hence the constant term equals 0, i.e., P(x)=xP1(x)P(x) = xP_1(x) where P1(x)P_1(x) is a polynomial with integral coefficients. The polynomial P1(x)P_1(x) satisfies the conditions of the problem, because gcd(P1(u),P1(v))>1\text{gcd}(|P_1(u)|, |P_1(v)|) > 1 for coprime uu and vv would imply gcd(P(u),P(v))=gcd(uP1(u),vP1(v))>1\text{gcd}(|P(u)|, |P(v)|) = \text{gcd}(|uP_1(u)|, |vP_1(v)|) > 1, contradicting the choice of P(x)P(x). By the induction hypothesis, P1(x)=±xl1P_1(x) = \pm x^{l-1}. Thus P(x)=±xlP(x) = \pm x^l.

On the other hand, all such polynomials satisfy the condition of the problem as gcd(u,v)=1\text{gcd}(u, v) = 1 implies gcd(ul,vl)=1\text{gcd}(u^l, v^l) = 1.

Solution 2

As in Solution 1, we show that all prime divisors of P(n)P(n) are divisors of nn. Letting qq be any prime number, we conclude that P(q)=±qkP(q) = \pm q^k for some natural number kk. Let the degree of P(x)P(x) be ll. Then xl+1>P(x)x^{l+1} > |P(x)| for every x>Nx > N where NN is a large enough integer. Thus for infinitely many prime numbers qq, there exists a natural number kk not exceeding ll such that P(q)=±qkP(q) = \pm q^k. As there are finitely many such natural numbers, one can fix kk such that either P(q)=qkP(q) = q^k for infinitely many primes qq or P(q)=qkP(q) = -q^k for infinitely many primes qq. This implies that either P(x)=xkP(x) = x^k or P(x)=xkP(x) = -x^k (as a matter of fact, k=lk = l). All such polynomials satisfy the conditions of the problem as gcd(u,v)=1\text{gcd}(u, v) = 1 implies gcd(ul,vl)=1\text{gcd}(u^l, v^l) = 1.

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