Maths Olympiad Prep

Library / /25 of 27

Number theory Difficulty 7.2 National olympiad, round 2 Prove it Romania

Determine all polynomials PZ[X]P \in \mathbb{Z}[X] such that there exists kNk \in \mathbb{N}^*, so that for all primes qq, P(q)P(q) has at most kk distinct prime divisors.

Solution

Clearly a constant non-null polynomial PP fulfills the requirements, so we shall assume in the sequel degP1\deg P \ge 1. Before further proceeding with the solution, we state the following "folklore" preliminaries.
(1) Dirichlet's Theorem. The arithmetic progression (an+b)nN(an + b)_{n \in \mathbb{N}} with aNa \in \mathbb{N}^*, bZb \in \mathbb{Z}^*, (a,b)=1(a, b) = 1, contains infinitely many primes.
(2) "Fundamental" property of integer polynomials. For any fZ[X]f \in \mathbb{Z}[X] and any a,bZa, b \in \mathbb{Z} we have abf(a)f(b)a - b \mid f(a) - f(b).
Claim. P(0)=0P(0) = 0. We proceed by contradiction, thus assume P(0)0P(0) \ne 0.
We will prove by induction on kk the following statement
*For any kNk \in \mathbb{N}^* there exists a prime p>P(0)p > |P(0)| such that P(p)P(p) has at least kk distinct prime factors.*
For k=1k=1, there exists a prime p1>P(0)p_1 > |P(0)| such that P(p1)1|P(p_1)| \ne 1, otherwise for all such primes pp we would have P(p)2=1P(p)^2 = 1 and thus the polynomial P(X)21P(X)^2 - 1 has infinitely many roots, thus it is the null polynomial, hence P(X)1(modp)P(X) \equiv 1 \pmod{p} or P(X)1(modp)P(X) \equiv -1 \pmod{p}, contradicting degP1\deg P \ge 1. Therefore P(p1)P(p_1) has at least one prime factor, so the statement is proven for k=1k=1.
Let pkp_k be a prime with the property pk>P(0)p_k > |P(0)| and P(pk)P(p_k) has at least kk distinct prime factors. Let us notice that (pk,P(pk))=1(p_k, P(p_k)) = 1, since from the "fundamental" property we have P(pk)P(0)(modpk)P(p_k) \equiv P(0) \pmod{p_k} and since pk>P(0)p_k > |P(0)| and it is a prime, the conclusion follows. The arithmetic progression (nP(pk)2+pk)nN(nP(p_k)^2 + p_k)_{n \in \mathbb{N}} (according to Dirichlet's Theorem) contains infinitely many primes. Let pk+1=sP(pk)2+pkp_{k+1} = sP(p_k)^2 + p_k be such a prime. Then P(pk+1)P(pk)(modP(pk)2)P(p_{k+1}) \equiv P(p_k) \pmod{P(p_k)^2}, thus there exists tt such that P(pk+1)=P(pk)(1+tP(pk))P(p_{k+1}) = P(p_k)(1 + tP(p_k)). From this it is obvious that P(pk+1)P(p_{k+1}) has at least one more prime factor than P(pk)P(p_k). Since P(pk)P(p_k) had at least kk distinct prime factors, it follows that P(pk+1)P(p_{k+1}) has at least k+1k+1 prime factors.
Thus the statement is proven and we get that the assumption P(0)0P(0) \ne 0 is false. From P(0)=0P(0) = 0 we deduce there exists mm so that P(X)=XmQ(X)P(X) = X^m Q(X) with Q(0)0Q(0) \ne 0. If QQ would be nonconstant, arguing the same as in the above, we again obtain a contradiction.
We conclude the only solutions are P(X)=cXmP(X) = cX^m, for some mNm \in \mathbb{N} and cZc \in \mathbb{Z}^*.

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 and solution reproduced as published; topic and difficulty added by this site.