Viete's relations
(All-Russian Mathematical Olympiad 2007 Grade 11 Day 2 P6, AoPS, proposed by N. Agakhanov, I. Bogdanov) Do there exist nonzero reals \(a, b, c\) such that, for any \(n > 3\), there exists a polynomial \(P_{n}(x) = x^{n} + \dots + a x^{2} + b x + c\), which has exactly \(n\) (not necessarily distinct) integral roots?
Click here for the spoiler!
Assume that there exist nonzero reals \(a, b, c\) satisfying the given condition. Let \(n > 3\) be a positive integer, and let \(P_{n}(x) = x^{n} + \dots + a x^{2} + b x + c\) be a polynomial with exactly \(n\) integral roots, which we denote by \(r_{1}, r_{2}, \dots, r_{n}\), counted with multiplicities. Since \(c\) is nonzero, none of the roots \(r_{1}, r_{2}, \dots, r_{n}\) is equal to zero. It follows that \[ \frac{1}{r_1}, \frac{1}{r_2}, \dots, \frac{1}{r_n} \] are the roots of the polynomial \[ Q_n(x) = x^n P\left( \frac{1}{x} \right) = c x^n + b x^{n-1} + a x^{n-2} + \dots + x + 1. \] By Vieta’s formulas, we obtain \[ \frac{1}{r_1} + \dots + \frac{1}{r_n} = - \frac{b}{c} , \quad \sum_{1 \leq i < j \leq n} \frac{1}{r_i r_j} = \frac{a}{c} . \] This implies that \[ \left( \frac{1}{r_1} + \dots + \frac{1}{r_n} \right)^2 - 2 \sum_{1 \leq i < j \leq n} \frac{1}{r_i r_j} = \frac{b^2 - 2 a c}{c^2} , \] and hence \[ \frac{1}{r_1^2} + \dots + \frac{1}{r_n^2} = \frac{b^2 - 2 a c}{c^2} . \] Also note that \[ r_1 r_2 \dots r_n = (-1)^{n} c . \] It follows that \[ b^2 - 2ac \geq \frac{r_1^2 r_2^2 \dots r_n^2}{r_1^2} + \dots + \frac{r_1^2 r_2^2 \dots r_n^2}{r_n^2} \geq n. \] This shows that \(n \leq b^2 - 2ac\) for any integer \(n > 3\), which is a contradiction. Therefore, there do not exist nonzero reals \(a, b, c\) satisfying the given condition.
Lecture notes for Math Olympiad (IOQM, RMO, INMO)
Click on the icons below to download.
Topics Links Algebra Combinatorics Geometry Number Theory INMO Training Camp IMO Training Camp MOPSS Simon Marais Mathematics Competition Resources