Maths Olympiad Prep

Track / Stage 8 / 25 of 180 #1725 of 1964

Problem 1725

IMO Shortlist mid-range; USAMO P2/P5
Number theory Difficulty 8.0 Prove it China National Team Selection Test · China

Let n>1n > 1 be an integer, kk be the number of distinct prime factors of nn. Prove that there exists an integer aa, 1<a<nk+11 < a < \frac{n}{k} + 1, such that na2an \mid a^2 - a. (posed by Yu Hongbing)

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

Let n=p1a1pkakn = p_1^{a_1} \cdots p_k^{a_k} be the standard factorization of nn. Since p1a1,,pkakp_1^{a_1}, \dots, p_k^{a_k} are pairwise coprime, by the Chinese Remainder Theorem, for each ii, 1ik1 \le i \le k, congruence equations
{x1(modpiai)x0(modpjaj),ji \begin{cases} x \equiv 1 \pmod{p_i^{a_i}} \\ x \equiv 0 \pmod{p_j^{a_j}}, \quad j \neq i \end{cases}
have solution xix_i.
For any solution of x02=x0(modn)x_0^2 = x_0 \pmod{n}, we see that x0(x01)0(modn)x_0(x_0 - 1) \equiv 0 \pmod{n}. Then for each i=1,2,,ki = 1, 2, \dots, k, either x00(modpiai)x_0 \equiv 0 \pmod{p_i^{a_i}} or x01(modpiai)x_0 \equiv 1 \pmod{p_i^{a_i}}. Further, let S(A)S(A) be the sum of elements of subset {x1,x2,,xk}\{x_1, x_2, \dots, x_k\} (particularly, S()=0S(\emptyset) = 0). Obviously, we have
S(A)(S(A)1)0(modn). S(A)(S(A) - 1) \equiv 0 \pmod{n}.
(This is because of the selection of xix_i, such that S(A)(modpiai)S(A) \pmod{p_i^{a_i}} is either 0 or 1.) Moreover if AAA \neq A', then S(A)S(A)(modn)S(A) \neq S(A') \pmod{n}. Therefore, the sum of all subsets of {x1,x2,,xn}\{x_1, x_2, \dots, x_n\} is exactly all solutions of x(x1)0(modn)x(x - 1) \equiv 0 \pmod{n}.
Let S0=nS_0 = n, SrS_r be the least non-negative remainder of x1+x2++xrx_1 + x_2 + \dots + x_r modulo nn, r=1,2,,kr = 1, 2, \dots, k. Thus Sk=1S_k = 1. For all 1rk11 \le r \le k - 1, Sr0S_r \neq 0. Since k+1k+1 numbers S0,S1,,SkS_0, S_1, \dots, S_k are in [1,n][1, n], by Dirichlet's Drawer Principle, there exist 0l<mk0 \le l < m \le k, such that Sl,SmS_l, S_m in the same interval (jnk,(j+1)nk]\left(\frac{jn}{k}, \frac{(j+1)n}{k}\right], (0jk1)(0 \le j \le k-1), where l=0l=0 and m=km=k do not hold simultaneously.
Thus, SlSm<nk|S_l - S_m| < \frac{n}{k}. Denote y1=S1,yr=SrSr1y_1 = S_1, y_r = S_r - S_{r-1} (r=2,3,,kr=2, 3, \dots, k). So any sum of yrxr(modn)y_r \equiv x_r \pmod{n} (r=1,2,,kr=1, 2, \dots, k) meets the requirement.
If SmSl>1S_m - S_l > 1, then a=yl+1+yl+2++ym=SmSl(1,nk)a = y_{l+1} + y_{l+2} + \dots + y_m = S_m - S_l \in (1, \frac{n}{k}) is the solution of the equation x2x0(modn)x^2 - x \equiv 0 \pmod{n}.
If SmSl=1S_m - S_l = 1, then nn (y1+y2++yl+(ym+1+ym+2++yk)y_1 + y_2 + \dots + y_l + (y_{m+1} + y_{m+2} + \dots + y_k), that is, nn (x1+x2++xl+(xm+1+xm+2++xk)x_1 + x_2 + \dots + x_l + (x_{m+1} + x_{m+2} + \dots + x_k). Notice that m>lm > l, which contradicts to the definition of xix_i.
If SmSl=0S_m - S_l = 0, then nyl+1+yl+2++ymn \mid y_{l+1} + y_{l+2} + \cdots + y_m, that is, nxl+1+xl+2++xmn \mid x_{l+1} + x_{l+2} + \cdots + x_m, which contradicts the definition of xix_i.
If SmSl<0S_m - S_l < 0, then
a=(y1+y2++yl)+(ym+1++yk)=Sk(SmSl)=1(SmSl) \begin{aligned} a &= (y_1 + y_2 + \cdots + y_l) + (y_{m+1} + \cdots + y_k) \\ &= S_k - (S_m - S_l) = 1 - (S_m - S_l) \end{aligned}
is the solution of equation x2x0(modn)x^2 - x \equiv 0 \pmod{n}, and 1<a<1+nk1 < a < 1 + \frac{n}{k}.
Summing up, there exists aa satisfying the condition. \square

Source: MathNet, licensed CC-BY-4.0. Statement and solution reproduced as published; topic, difficulty and ordering added by this site.