5
$\begingroup$

Let $P(x)$ be a polynomial with integer coefficients of degree $d>0$.

  1. If $\alpha $ and $\beta $ are two integers such that $P(\alpha)=1$ and $P(\beta)=-1$, then prove that $|\beta -\alpha | $ divides $2$.
  2. Prove that the number of distinct integer roots of $(P(x))^2-1$ is at most $d+2$.

First one is very easy. But I cannot understand how to prove the second one. I would appreciate any help.

|cite|improve this question
$\endgroup$
  • 2
    $\begingroup$ Just a clarification, does $P^2(x)$ mean $P(P(x))$ or $(P(x)) ^2$? $\endgroup$ – Aritra Das May 5 '16 at 7:23
  • 2
    $\begingroup$ I remember this is an ISI entrance problem. Can you tell me which year? Then, I might be able to link you to the solution. $\endgroup$ – SinTan1729 May 5 '16 at 7:30
  • 2
    $\begingroup$ @Perth This might help. ISI B.Math 2007 Question number 5 $\endgroup$ – SinTan1729 May 5 '16 at 7:35
  • 1
    $\begingroup$ @Perth The second part does not look difficult. Suppose $P(x)=1$ has 6 or more distinct integer roots. Then we cannot have an integer root of $P(x)=-1$ because it would be more than a distance 2 from at least one of the roots of $P(x)=1$. That basic idea can easily be elaborated to a complete solution. $\endgroup$ – almagest May 5 '16 at 7:43
  • 1
    $\begingroup$ Since the solutions over at AoPS involve case checking, I would like to see an elegant solution for this, especially because the result is nice. $\endgroup$ – Aritra Das May 5 '16 at 8:05
4
$\begingroup$

The first part is obvious. Since $\alpha,\beta, P(\alpha)$ and $P(\beta)$ are all integers, $\mid \beta - \alpha \mid$ divides $\mid P(\beta) - P(\alpha) \mid = 2$.

For the second part, notice that $(P(x))^2-1=(P(x)+1)(P(x)-1)$. So, all the roots of those two polynomials would be a root of this polynomial. There are $2d$ such (not necessarily distinct) roots. Now, it's true for $d=1,2$ as $d+2 \geq 2d$. For $d \geq 3$, suppose there are more than $d+2$ such roots. So, there has to be at least $3$ roots of each of those two polynomials. It's not possible to have $6$ distinct integers such that any two of them are apart by $1$ or $2$. Hence, we cannot have more than $d+2$ distinct integer solutions.(thanks to @almagest) Imagine a numberline. WLOG, Let $P(a)=1$. Now, the $3$ integer roots of $P(x)=-1$ would have to be of the form $a \pm r, r=1 \text{or}\, 2$. It's easy to see that whatever integer we choose next, it'd differ by more than $2$ with at least one of these roots. Hence, we cannot have more than $d+2$ solutions.

|cite|improve this answer
$\endgroup$
  • 1
    $\begingroup$ +1 A minor quibble. You don't need as much as 6 integers such that any two of them are apart by 1 or 2. You need two sets of 3 integers such that the distance between any integer in the first set and any integer in the second set is 1 or 2. Of course, that is still impossible. $\endgroup$ – almagest May 5 '16 at 8:39
  • $\begingroup$ @almagest Oops. You're right. I'll correct that. $\endgroup$ – SinTan1729 May 5 '16 at 8:42

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.