Maths Olympiad Prep

Library / /58 of 60

Number theory Difficulty 6.0 National olympiad Prove it South Africa

Suppose that aa is an integer, and that n!+an! + a divides (2n)!(2n)! for infinitely many positive integers nn. Prove that a=0a = 0.

Solutions — 2

Solution 1

(2n)!=(2nn)n!2(2nn)(a)2mod(n!+a),(2n)! = \binom{2n}{n} \cdot n!^2 \equiv \binom{2n}{n} \cdot (-a)^2 \mod (n! + a),
so if n!+an! + a divides (2n)!(2n)!, then it also divides a2(2nn)a^2 \binom{2n}{n}. We will show that when nn is large, n!+an! + a is greater than a2(2nn)a^2 \binom{2n}{n} and therefore does not divide it (unless a=0a = 0). Assume in the following that a0a \neq 0 and n>4512a2n > \frac{4^5}{12}a^2. First of all, by the binomial theorem,
(2nn)k=02n(2nk)=22n=4n. \binom{2n}{n} \le \sum_{k=0}^{2n} \binom{2n}{k} = 2^{2n} = 4^n.
If a>0a > 0, we have
0<a2n!+a(2nn)<a2n!(2nn)a2n!4n=(4n556(n1))45a224n<1. 0 < \frac{a^2}{n! + a} \binom{2n}{n} < \frac{a^2}{n!} \binom{2n}{n} \le \frac{a^2}{n!} 4^n = \left( \frac{4^{n-5}}{5 \cdot 6 \cdots (n-1)} \right) \frac{4^5 a^2}{24n} < 1.
If a<0a < 0, then we observe that n!>n>2an! > n > 2|a|, so that
0<a2n!+a(2nn)=a2n!a(2nn)<a212n!(2nn)2a2n!4n=(4n556(n1))45a212n<1. 0 < \frac{a^2}{n! + a} \binom{2n}{n} = \frac{a^2}{n! - |a|} \binom{2n}{n} < \frac{a^2}{\frac{1}{2}n!} \binom{2n}{n} \le \frac{2a^2}{n!} 4^n = \left( \frac{4^{n-5}}{5 \cdot 6 \cdots (n-1)} \right) \frac{4^5 a^2}{12n} < 1.

Thus a2(2nn)n!+a\frac{a^2 \binom{2n}{n}}{n!+a} is not an integer for all such nn, which contradicts our assumption that n!+an! + a divides (2n)!(2n)! (and thus a2(2nn)a^2 \binom{2n}{n}) for infinitely many nn.
So a=0a = 0 is the only possibility (and indeed n!n! divides (2n)!(2n)! for all nn).

Solution 2

We show that n!+an! + a cannot divide (2n)!(2n)! if a0a \neq 0, n2an \ge 2|a| and n9n \ge 9. In this case, n!a\frac{n!}{a} is an integer. If n!+an! + a divides (2n)!(2n)!, then so does n!a+1\frac{n!}{a} + 1. Next note that every prime pnp \le n divides n!a\frac{n!}{a}: if pap \nmid a, this is obvious (since pn!p \nmid n!), and if pap \mid a, then we see that pp divides 2a2|a|, which occurs as a factor of n!a\frac{n!}{a}. In either case, we find that n!a+1\frac{n!}{a} + 1 is not divisible by pp. So n!a+1\frac{n!}{a} + 1 does not have any prime factors n\le n, which means that it is coprime to all numbers n\le n as well as all the even numbers 2n\le 2n.
Therefore n!a+1\frac{n!}{a} + 1 must divide the product of all odd numbers greater than nn and less than 2n2n. By our assumption, n!a+1na(n1)!12(n1)!1\left|\frac{n!}{a} + 1\right| \ge \frac{n}{a} \cdot (n-1)! - 1 \ge 2(n-1)! - 1. We show that the product of the odd numbers between nn and 2n2n is less than 2(n1)!12(n-1)! - 1 for n9n \ge 9, which yields the desired contradiction.
For n=9n = 9 and n=10n = 10, we have the two inequalities
28!1=80639>36465=11131517 2 \cdot 8! - 1 = 80639 > 36465 = 11 \cdot 13 \cdot 15 \cdot 17
and
29!1=725759>692835=1113151719 2 \cdot 9! - 1 = 725759 > 692835 = 11 \cdot 13 \cdot 15 \cdot 17 \cdot 19
respectively. Now we proceed by induction. Assume that
2(n1)!1>n<k<2nk oddk. 2(n-1)! - 1 > \prod_{\substack{n<k<2n \\ k \text{ odd}}} k.
If n>10n > 10 is even, we use the induction hypothesis to obtain
n<k<2nk oddk=(2n3)(2n1)n1n2<k<2n4k oddk<2(2n1)n2<k<2n4k oddk<(4n2)(2(n3)!1)<(n1)(n2)(2(n3)!1)=2(n1)!(n1)(n2)<2(n1)!1, \prod_{\substack{n<k<2n \\ k \text{ odd}}} k = \frac{(2n-3)(2n-1)}{n-1} \prod_{\substack{n-2<k<2n-4 \\ k \text{ odd}}} k < 2(2n-1) \prod_{\substack{n-2<k<2n-4 \\ k \text{ odd}}} k < (4n-2)(2(n-3)!-1) \\ < (n-1)(n-2)(2(n-3)!-1) = 2(n-1)! - (n-1)(n-2) < 2(n-1)!-1,
and if n>10n > 10 is odd, we get analogously
n<k<2nk oddk=(2n3)(2n1)nn2<k<2n4k oddk<2(2n3)n2<k<2n4k oddk<(4n6)(2(n3)!1)<(n1)(n2)(2(n3)!1)=2(n1)!(n1)(n2)<2(n1)!1. \prod_{\substack{n<k<2n \\ k \text{ odd}}} k = \frac{(2n-3)(2n-1)}{n} \prod_{\substack{n-2<k<2n-4 \\ k \text{ odd}}} k < 2(2n-3) \prod_{\substack{n-2<k<2n-4 \\ k \text{ odd}}} k < (4n-6)(2(n-3)!-1) \\ < (n-1)(n-2)(2(n-3)!-1) = 2(n-1)! - (n-1)(n-2) < 2(n-1)!-1.

This completes the induction and thus the proof.

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.