Let $n\geq 4$ be a positive integer and $x_{1},x_{2},\ldots ,x_{n},x_{n+1},x_{n+2}$ be real numbers such that $x_{n+1}=x_{1}$ and $x_{n+2}=x_{2}$. If there exists an $a>0$ such that \[x_{i}^2=a+x_{i+1}x_{i+2}\quad\forall 1\leq i\leq n\]then prove that at least $2$ of the numbers $x_{1},x_{2},\ldots ,x_{n}$ are negative.
Problem
Source: Bulgaria NMO 2022 P4
Tags: algebra, system of equations, parameter, Bulgaria
17.04.2022 17:57
Here is a sketch of my approach for the first part (i.e. that all numbers can't be positive): Note that $x_{i+1}=\frac {x_i^2-a}{x_{i+2}}=\frac {x_{i-1}^2-a}{x_i}$, so $x_i^3-ax_i=x_{i-1}^2x_{i+2}-ax_{i+2}$ and summing cyclically, we obtain $\sum x_i^3 = \sum x_{i-1}^2x_{i+2}$ and using AM-GM, we have that $\frac {2x_{i-1}^3+x_{i+2}^3}{3} \geq x_{i-1}^2x_{i+2}$, and summing cyclically, we have that equality in the above inequality holds, so $x_{i-1}=x_{i+2}$ for all $i$. Now either the sequence is constant (if $n$ is not divisible by 3 - but constant sequence is obviously impossible), or there are exactly three values the sequence has, so now it is easy to finish considering three equations from the given.
17.04.2022 18:14
First assume that $x_{i}>0\forall i$. Then notice that $x_{i}^2=a+x_{i+1}x_{i+2}>x_{i+1}x_{i+2}$ and if we multiply all such inequalities for $1\leq i\leq n$ we get that $\prod\limits_{i=1}^{n}x_{i}^2>\prod\limits_{i=1}^{n} x_{i}x_{i+1}=\prod\limits_{i=1}^{n}x_{i}^2$, contradiction. This means that there is at least one negative number among the $x_{i}$-s. WLOG $x_{1}<0$ as the equations are cyclic. Assume that $x_{i}\geq 0\forall i\neq 1$. As the equations are homogenous, we can assume that $x_{1}=-1$ (we divide every $x_{i}$ by $|x_{1}|$ and $a$ by $x_{1}^2\neq 0$). Now we have that $1=x_{1}^2=a+x_{2}x_{3}>a\Longrightarrow a\in(0,1)$ and $x_{2}^2=a+x_{3}x_{4}>a\Longrightarrow x_{2}\geq \sqrt{a}$. However, notice that now we have a contradiction because: $$0\leq x_{n}^2=a+x_{1}x_{2}=a-x_{2}<a-\sqrt{a}<0$$
17.04.2022 20:37
Quickest approach perhaps? If all numbers are positive and without loss of generality $x_1$ is minimal, we get $x_1^2 = a + x_2x_3 \geq a + x_1^2 > x_1^2$, contradiction! Now suppose without loss of generality that $x_1 < 0$ and all others are non-negative. Then $x_1^2 = x_2x_3 + a \geq a$, so $x_1 \leq -\sqrt{a}$; also $x_2^2 = x_3x_4 + a \geq a$, i.e. $x_2 \geq \sqrt{a}$. This implies $x_n^2 = x_1x_2 + a \leq 0$ and hence $x_n = 0$, $x_1 = -\sqrt{a}$, $x_2 = \sqrt{a}$. But then $a = x_1^2 = x_2x_3 + a$, giving $x_3 = 0$ and $0 = x_3^2 = x_4x_5 + a \geq a$ when $n\geq 5$ (so that $x_5 \neq x_1 < 0$), and also for $n=4$ (from $x_n = 0$), contradiction! Remark: I invite the ones who are interested to look for significantly better lower bounds for the number of negative terms in the sequence.
23.04.2022 15:48
https://dgrozev.wordpress.com/2022/04/23/bulgarian-national-math-olympiad-2022-part-3/
11.11.2022 19:32
Here is my solution: https://calimath.org/pdf/BulgariaNMO2022-4.pdf And I uploaded the solution with motivation to: https://youtu.be/3f7e8sW52og