Let $p(x) = c_0 + c_1x + \cdots + c_d x^d$ be a polynomial with integer coefficients $c_0, \ldots, c_d \in \mathbb{Z}$ and $c_d \neq 0$, so $p(x)$ has degree $d$. Let $a/b$ be a rational root of $p(x)$ written in lowest terms. How do I see that $a \mid c_0$ and $b \mid c_d$?

- 75,673
-
5Is this not precisely the rational root theorem? – Théophile Jul 12 '16 at 18:07
-
it is obvious when you write $p(x) = (x- \frac{a}{b}) \sum_{k=0}^{d-1} u_k x^k $ – reuns Jul 12 '16 at 18:08
-
Do you know anything about polynomial long division? If you do, I will try to write something nice up for you. – The Count Jul 12 '16 at 18:13
-
@user1952009 That's assuming you know Gauss's Lemma or equivalent results, e.g. see here. – Bill Dubuque Jul 12 '16 at 19:02
4 Answers
We know that $$c_0 + c_1\frac{a}{b}+ \cdots c_d(\frac{a}{b})^d = 0$$ is true. Multiplying both sides by $b^d$, we get $$c_0b^d + c_1 a b^{d-1}+ \cdots c_da^d = 0.$$ Right side is divisible by $a$ and $b$ so left side is also divisible by $a$ and $b$. I believe you can finish it from here!

- 3,044
Suppose $\frac ab$ is a root of $P(x)$ and $\gcd(a,b) = 1.$ i.e. $\frac ab$ is in lowest terms.
$P(\frac ab) = c_0 + c_1 \frac ab+c_2 \frac {a^2}{b^2} \cdots + c_d\frac {a^d}{b^d} = 0\\ c_1 \frac ab+c_2 \frac {a^2}{b^2} \cdots + c_d\frac {a^d}{b^d} = -c_0$
Every term on the left side of the equals sign is divisble by $a.$ Therefore $c_0$ is divisible by $a.$
Multiply trough by $b^d$ and rearrange the terms.
$c_d a^d = - c_0 b^d - c_1 a b^{d-1} \dots -c^{d-1} a^{d-1}b$
And similarly, every term on the right is divisible by $b,$ therefore $c_d$ is divisible by $b.$

- 57,877
First note you can assume $a$ and $b$ are coprime ($gcd(a,b)$=1). By substitution $x=\frac{a}{b}$ you get
$c_0+c_1\frac{a}{b}+\dots+c_d\frac{a^d}{b^d}=0$
Then multiply by $b^d$
$c_0b^d+c_1ab^{d-1}+\dots+c_da^d=0$
Now you see that $a$ divides the RHS, so it must divide LHS. But since it divides every summand except possibly $c_0b^d$, it must also divide this one. But $\gcd(a,b^n)=1$, so it must be $a\,|\,c_0$. By the same reasoning (swap $a$ and $b$, $c_0$ and $c_d$) you get $b\,|\,c_d$.
EDIT: I'm sorry for repeating the above posts, they weren't there when I started replying. I believed the website would have informed me before posting if other answer appeared while I was writing.

- 341
Here's a modular proof: $\,\bmod a\!:\ b^{-1}$ exists by $\,(a,b)=1,\,$ so $\,a/b\equiv ab^{-1}\equiv 0,\,$ so
$$\ {\rm mod}\ a\!:\,\ 0 \equiv p(a/b) \equiv p(0)\ \Rightarrow\, a\mid p(0) = c_0$$
The reversed poly $\ \bar p(x) = x^d p(1/x)\ $ has root $\, b/a,\,$ hence $\ b\mid \bar p(0) = c_d\,$ by above.
Remark $\ $ The is usually called the Rational Root Test. The above proof presumes that you know that modular arithmetic (congruences) can be extended to fractions with denominators coprime to the modulus (see the discussion here), a property which is greatly clarified when one studies ring theory (the universal properties of fraction rings or localizations).

- 272,048
-
Not if $,a=0,$ then $,\gcd(a,b)=1,\Rightarrow, b=1,$ so $,b\mid c_d,,$ so we meed only consider the case $,a\neq 0$ in the reversed case with root $,b/a.\ $ – Bill Dubuque Oct 16 '18 at 18:27