Maths Olympiad Prep

Track / Stage 6 / 107 of 400 #1107 of 1964

Problem 1107

National olympiad, first round
Number theory Difficulty 6.1 Prove it

Let τ(n)\tau(n) denote the number of positive integer divisors of a positive integer nn (for example, τ(2022)=8\tau(2022)=8 ). Given a polynomial P(X)P(X) with integer coefficients, we define a sequence a1,a2,a_{1}, a_{2}, \ldots of nonnegative integers by setting
an={gcd(P(n),τ(P(n))) if P(n)>00 if P(n)0 a_{n}=\left\{\begin{array}{ll} \operatorname{gcd}(P(n), \tau(P(n))) & \text { if } P(n)>0 \\ 0 & \text { if } P(n) \leq 0 \end{array}\right.
for each positive integer nn. We then say the sequence has limit infinity if every integer occurs in this sequence only finitely many times (possibly not at all).

Does there exist a choice of P(X)P(X) for which the sequence a1,a2,a_{1}, a_{2}, \ldots has limit infinity?

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.

Official solution

§3.5a Solution
We claim the answer is no, such PP does not exist.
Clearly we may assume PP is nonconstant with positive leading coefficient. Fix PP and fix constants n0,c>0n_{0}, c>0 such that c=P(n0)>0c=P\left(n_{0}\right)>0. We are going to prove that infinitely many terms of the sequence are at most cc.
We start with the following lemma.
Claim - For each integer n2n \geq 2, there exists an integer r=r(n)r=r(n) such that
- For any prime pp which is at most nn, we have νp(P(r))=νp(c)\nu_{p}(P(r))=\nu_{p}(c).
- We have
 c. prime pnr2cprime pnp \text { c. } \prod_{\text {prime } p \leq n} \leq r \leq 2 c \cdot \prod_{\text {prime } p \leq n} p \text {. }

Proof. This follows by the Chinese remainder theorem: for each pnp \leq n we require rn0r \equiv n_{0} (modpνp(c)+1)\left(\bmod p^{\nu_{p}(c)+1}\right), which guarantees νp(P(r))=νp(P(n0))=νp(c)\nu_{p}(P(r))=\nu_{p}\left(P\left(n_{0}\right)\right)=\nu_{p}(c). Then there exists such an rr modulo pnpνp(c)+1\prod_{p \leq n} p^{\nu_{p}(c)+1} as needed.

Assume for contradiction that all aia_{i} are eventually larger than cc. Take nn large enough that n>cn>c and r=r(n)r=r(n) has ar>ca_{r}>c. Then consider the term ara_{r} :
- Using the conditions in the lemma it follows there exists a prime pn>np_{n}>n which divides ar=gcd(P(r),τ(P(r))a_{r}=\operatorname{gcd}\left(P(r), \tau(P(r))\right. ) (otherwise ara_{r}, which divides P(r)P(r), is at most cc ).
- As pnp_{n} divides τ(P(r))\tau(P(r)), this forces P(r)P(r) to be divisible by (at least) qnpn1q_{n}^{p_{n}-1} for some prime qnq_{n}.
- For the small primes pp at most nn, we have νp(P(r))=νp(c)n\nu_{p}(P(r))=\nu_{p}(c)n.
- Ergo,
P(r)qnpn1>nn. P(r) \geq q_{n}^{p_{n}-1}>n^{n} .

In other words, for large enough nn, we have the asymptotic estimate
n n 0 define\text{n n 0 define}
δ(M):=pM(11p)\delta(M):=\prod_{p \leq M}\left(1-\frac{1}{p}\right) .

Then π(n)<δ(M)n+pMp\pi(n)<\delta(M) n+\prod_{p \leq M} p, so it suffices to check that limMδ(M)=0\lim _{M \rightarrow \infty} \delta(M)=0. But
1δ(M)=pM(11p)1=pM(1+1p+1p2+)1+12++1M \frac{1}{\delta(M)}=\prod_{p \leq M}\left(1-\frac{1}{p}\right)^{-1}=\prod_{p \leq M}\left(1+\frac{1}{p}+\frac{1}{p^{2}}+\ldots\right) \geq 1+\frac{1}{2}+\cdots+\frac{1}{M}
which diverges for large MM.

Source: NuminaMath-1.5, licensed Apache-2.0. Statement and solution reproduced as published; topic, difficulty and ordering added by this site.