Maths Olympiad Prep

Library / /200 of 299

Algebra Difficulty 6.8 National Olympiad Prove it Iran

Let n>1n > 1 be an integer and a1,a2,,an+1a_1, a_2, \dots, a_{n+1} be distinct positive integers. Does there exist a polynomial p(x)p(x) with integer coefficients of degree at most nn that satisfies the following conditions?
* 1i<jn+1:gcd(p(ai),p(aj))>1\forall 1 \le i < j \le n+1 : \gcd(p(a_i), p(a_j)) > 1
* 1i<j<kn+1:gcd(p(ai),p(aj),p(ak))=1\forall 1 \le i < j < k \le n+1 : \gcd(p(a_i), p(a_j), p(a_k)) = 1

Solution

Let b(i,j)b_{(i,j)} for 1i,jn+11 \le i, j \le n+1 be positive integers in which for every i,j,k,li, j, k, l where {i,j}{k,l}\{i, j\} \ne \{k, l\}, we have (b(i,j),b(k,l))=1(b_{(i,j)}, b_{(k,l)}) = 1 and b(i,j)=b(j,i)b_{(i,j)} = b_{(j,i)}. According to Lagrange interpolation formula, there exists a unique polynomial P(x)P(x) with rational coefficients of degree at most nn that satisfies 1in+1,P(ai)=j=1n+1b(i,j)\forall 1 \le i \le n+1, P(a_i) = \prod_{j=1}^{n+1} b_{(i,j)}. Then we'll have:
(P(ai),P(aj))=b(i,j)>1,(P(ai),P(aj),P(ak))=(b(i,j),P(ak))=1 (P(a_i), P(a_j)) = b_{(i,j)} > 1, \quad (P(a_i), P(a_j), P(a_k)) = (b_{(i,j)}, P(a_k)) = 1
Therefore, it suffices to prove that there is a sequence b(i,j)b_{(i,j)} in which the resulting polynomial PP has integer coefficients in lieu of rational. According to Lagrange interpolation formula,
P(x)=i=1nP(ai)j=1,jin(xaj)j=1,jin(aiaj) P(x) = \sum_{i=1}^{n} P(a_i) \frac{\prod_{j=1, j \neq i}^{n} (x - a_j)}{\prod_{j=1, j \neq i}^{n} (a_i - a_j)}
Define cc to be c=i=1n+1j=1,jin+1(aiaj)c = \prod_{i=1}^{n+1} \prod_{j=1, j \neq i}^{n+1} (a_i - a_j). Redefine b(i,j)b_{(i,j)} so that in addition to former conditions, for every (i,j)(i, j), b(i,j)1(modc)b_{(i,j)} \equiv 1 \pmod{c}. Obviously such sequence b(i,j)b_{(i,j)} exists, as it's equivalent to existence of n(n+1)2\frac{n(n+1)}{2} pairwise distinct positive integers all being 11 modulo cc, and the sequence
c1=1,2in(n+1)2:ci=cj=1i1cj+1 c_1 = 1, \quad \forall 2 \le i \le \frac{n(n+1)}{2} : c_i = c \prod_{j=1}^{i-1} c_j + 1
works well. Furthermore, P(ai)=j=1n+1b(i,j)1(modc)P(a_i) = \prod_{j=1}^{n+1} b_{(i,j)} \equiv 1 \pmod{c}.
It remains to prove that for such P(ai)P(a_i), PP would have integer coefficients. Define kjk_j to be kj=P(aj)1ck_j = \frac{P(a_j)-1}{c}. According to Lagrange interpolation formula for constant polynomial 11, we have
1=i=1n1j=1,jin(xaj)j=1,jin(aiaj)=i=1n(P(ai)cki)j=1,jin(xaj)j=1,jin(aiaj)=i=1n(P(ai))j=1,jin(xaj)j=1,jin(aiaj)i=1nckij=1,jin(xaj)j=1,jin(aiaj)=P(x)i=1nkij=1,jin(xaj)cj=1,jin(aiaj) \begin{align*} 1 &= \sum_{i=1}^{n} \frac{1 \cdot \prod_{j=1, j \neq i}^{n} (x - a_j)}{\prod_{j=1, j \neq i}^{n} (a_i - a_j)} = \sum_{i=1}^{n} \frac{(P(a_i) - c k_i) \prod_{j=1, j \neq i}^{n} (x - a_j)}{\prod_{j=1, j \neq i}^{n} (a_i - a_j)} \\ &= \sum_{i=1}^{n} \frac{(P(a_i)) \prod_{j=1, j \neq i}^{n} (x - a_j)}{\prod_{j=1, j \neq i}^{n} (a_i - a_j)} - \sum_{i=1}^{n} \frac{c k_i \prod_{j=1, j \neq i}^{n} (x - a_j)}{\prod_{j=1, j \neq i}^{n} (a_i - a_j)} \\ &= P(x) - \sum_{i=1}^{n} k_i \prod_{j=1, j \neq i}^{n} (x - a_j) \frac{c}{\prod_{j=1, j \neq i}^{n} (a_i - a_j)} \end{align*}
But by definition j=1,jin(aiaj)c\prod_{j=1, j \neq i}^{n} (a_i - a_j) \mid c. Therefore,
i=1nkij=1,jin(xaj)cj=1,jin(aiaj) \sum_{i=1}^{n} k_i \prod_{j=1, j \neq i}^{n} (x - a_j) \frac{c}{\prod_{j=1, j \neq i}^{n} (a_i - a_j)}
has integer coefficients, and so does P(x)P(x).

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 reproduced verbatim; metadata (topic, difficulty) added by this project.