Olympiad Maths Prep

Track / Stage 7 / 292 of 300 #1692 of 2000

Problem 1692

National olympiad second round; IMO P1/P4
Number theory Difficulty 7.9 Prove it

Example 9 Let pp be a prime, and given p+1p+1 distinct positive integers. Prove that one can select a pair of these numbers such that when the larger of the two is divided by their greatest common divisor, the quotient is at least p+1p+1.

This one wants a proof. Work it on paper, read the official solution, then mark yourself honestly — the ladder only means something if the record is true.

Official solution

Prove that dividing the given p+1p+1 numbers by their greatest common divisor obviously does not affect the conclusion of this problem, so we can assume that these p+1p+1 numbers are coprime. In particular, there must be a number that is not divisible by pp. Let these p+1p+1 numbers be
x1,,xk,xk+1=plk+1yk+1,,xp+1=plp+1yp+1x_{1}, \cdots, x_{k}, x_{k+1}=p^{l_{k+1}} y_{k+1}, \cdots, x_{p+1}=p^{l_{p+1}} y_{p+1}

Here, x˙1,,xk\dot{x}_{1}, \cdots, x_{k} are distinct and coprime with pp (k1)(k \geqslant 1), lk+1,,lp+1l_{k+1}, \cdots, l_{p+1} are positive integers, and yk+1,,yp+1y_{k+1}, \cdots, y_{p+1} are positive integers not divisible by pp.

Among the p+1p+1 numbers
x1,,xk,yk+1,,yp+1x_{1}, \cdots, x_{k}, y_{k+1}, \cdots, y_{p+1}

there must be two that are congruent modulo pp. We discuss three cases.

(1) At least three numbers in (1) are equal. In this case, the conclusion is easy to prove. If yr=ys=yty_{r}=y_{s}=y_{t}, then plr,pls,pltp^{l_{r}}, p^{l_{s}}, p^{l_{t}} are distinct, and the largest number is at least p2p^{2} times the smallest. Without loss of generality, assume plrp2pltp^{l_{r}} \geqslant p^{2} \cdot p^{l_{t}}, then xrx_{r} and xtx_{t} satisfy the condition; if yr=ys=xt(1tk)y_{r}=y_{s}=x_{t}(1 \leqslant t \leqslant k), without loss of generality, assume lr>lsl_{r}>l_{s}, then lr2l_{r} \geqslant 2, so xrx_{r} and xtx_{t} satisfy the condition.

(2) Two pairs of numbers in (1) are equal. If yi=yj,yr=ysy_{i}=y_{j}, y_{r}=y_{s}, then when lilj2\left|l_{i}-l_{j}\right| \geqslant 2 or lrls2\left|l_{r}-l_{s}\right| \geqslant 2, the conclusion holds as above; when lilj1\left|l_{i}-l_{j}\right| \leqslant 1 and lrls1\left|l_{r}-l_{s}\right| \leqslant 1, we can rename xi,xj,xr,xsx_{i}, x_{j}, x_{r}, x_{s} as a,ap,b,bpa, a p, b, b p, and apap

Thus, the integer bp(a,bp)p+1\frac{b p}{(a, b p)} \geqslant p+1.
If xi=yr,xj=ys(1i,jk)x_{i}=y_{r}, x_{j}=y_{s}(1 \leqslant i, j \leqslant k), the conclusion can also be proven similarly.

(3) Exactly two numbers in (1) are equal. This can only be yr=ysy_{r}=y_{s}, or xi=yr(1ik)x_{i}=y_{r}(1 \leqslant i \leqslant k). In this case, we can remove yry_{r} from (1), leaving pp numbers that are distinct but still have two that are congruent modulo pp. Now there are three possibilities:
(i) Suppose yrys(modp)y_{r} \equiv y_{s}(\bmod p). Without loss of generality, assume yr>ysy_{r}>y_{s}. If lr>lsl_{r}>l_{s}, the conclusion is obvious. If lrlsl_{r} \leqslant l_{s}, let yr=ys+ny_{r}=y_{s}+n, then n>0n>0, and pnp \mid n. Let (yr,ys)=d\left(y_{r}, y_{s}\right)=d, then pdp \nmid d, so (xr,xs)=pld\left(x_{r}, x_{s}\right)=p^{l} d. We have (note dn,pnd|n, p| n, and pdp \nmid d)
xr(xr,xs)=yrd=ysd+nd1+p\frac{x_{r}}{\left(x_{r}, x_{s}\right)}=\frac{y_{r}}{d}=\frac{y_{s}}{d}+\frac{n}{d} \geqslant 1+p

Thus, the larger of xrx_{r} and xsx_{s}, when divided by their greatest common divisor, yields a quotient of at least p+1p+1.
(ii) Suppose xrxs(modp)(1r<sk)x_{r} \equiv x_{s}(\bmod p)(1 \leqslant r < s \leqslant k). Without loss of generality, assume xr>xsx_{r}>x_{s}. If ys<yry_{s} < y_{r}, the conclusion is obvious. If ys>yry_{s} > y_{r}, let ys=yr+ny_{s}=y_{r}+n, then n>0n>0, and pnp \mid n. Let (xr,ys)=d\left(x_{r}, y_{s}\right)=d, then pdp \nmid d, so (xr,xs)=(xr,plsys)=d\left(x_{r}, x_{s}\right)=\left(x_{r}, p^{l_{s}} y_{s}\right)=d, thus
xr(xr,xs)=ysd+nd1+p\frac{x_{r}}{\left(x_{r}, x_{s}\right)}=\frac{y_{s}}{d}+\frac{n}{d} \geqslant 1+p

Thus, the larger of xrx_{r} and xsx_{s}, when divided by their greatest common divisor, yields a quotient of at least p+1p+1. This completes the proof of the problem.

We note that if the p+1p+1 integers in Example 9 are replaced by pp integers, the conclusion may not hold. For example, the pp numbers 1,2,,p1,2, \cdots, p clearly do not have two numbers that satisfy the condition.

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