8

Let $F$ be a field and $r$ and $s$ positive integers. Prove that, in $F[x]$,

$\gcd(x^r-1,x^s-1)=x^{\gcd(r,s)}-1$.

If $r$ and $s$ were known numbers, I could be able to attempt the problem, but I don't know how to do about this.

gebruiker
  • 6,154
YYG
  • 471
  • 3
  • 17
  • Hint: Try to do it for a couple of "known numbers" (e.g. take $r=2, s=3$, or $r=15,s=9$). Then try to do exactly the same procedure for "abstract entities", i.e. leaving the $r$ and $s$ as abstract integers. – Daniel Robert-Nicoud Dec 13 '15 at 10:30

2 Answers2

7

We can assume $r \geq s \geq 1$. Look at

$$x^r-1 - x^{r-s}(x^s-1)=x^{r-s}-1.$$

Hence $\operatorname{gcd}(x^r-1,x^s-1)=\operatorname{gcd}(x^{r-s}-1,x^s-1)$.

By induction on $r+s$ we can use the assertion for the right hand side and thus deduce

$$\operatorname{gcd}(x^r-1,x^s-1)=\operatorname{gcd}(x^{r-s}-1,x^s-1)=x^{\operatorname{gcd}(r-s,s)}-1=x^{\operatorname{gcd}(r,s)}-1.$$


Some notes on that proof: Of course the induction makes this look very graceful and you might not see why this works. Then have a look at $$x^r-1 - x^{r-s}(x^s-1)=x^{r-s}-1$$ again. This computation actually shows you that polynomial division of polynomials of the type $x^r-1$ is nothing but the division with remainders applied to the exponents. Since the greatest common divisor is calculated by iterated use of division with remainder, this gives you the result. The induction is just the technial tool to make this explaination rigid.

MooS
  • 31,390
  • MooS. This is very useful to me. But what I don't understand is where you deduced $\operatorname{gcd}(x^r-1,x^s-1)=\operatorname{gcd}(x^{r-s}-1,x^s-1)$ – YYG Dec 13 '15 at 10:47
  • I just used the well-known $\operatorname{gcd}(a,b)=\operatorname{gcd}(a-cb,b)$. – MooS Dec 13 '15 at 11:14
2

Lemma. Let $I$ be any ideal of $F[X]$, and let $H$ be the set of $n \in \mathbb{N}$ such that $X^n - 1 \in I$. Then $H$ is the intersection with $\mathbb{N}$ of a subgroup of $\mathbb{Z}$.

Proof of Lemma. Let $A = F[X]/I$, and let $a = \bar{X} \in A$. If $a$ is not invertible in $A$, then $H = \{0\}$. Otherwise, $H$ is the intersection with $\mathbb{N}$ of the kernel of the morphism $n \mapsto a^n$ from $\mathbb{Z}$ to $A^{*}$.

Applying the lemma to the ideal $(X^r-1,X^s-1)$, we find (by Bezout's lemma) that $X^{\gcd(r,s)} - 1 \in (X^r-1,X^s-1)$. Conversely, applying the lemma to $(X^{\gcd(r,s)} - 1)$, we see that $X^r - 1, X^s - 1 \in (X^{\gcd(r,s)} - 1)$. Thus $(X^r-1,X^s-1) = (X^{\gcd(r,s)} - 1)$.

David
  • 6,306