I think it's true that in a field $F$ of characteristic zero, there are exactly $n$ distinct $n$th roots of unity (in some algebraic closure $\bar{F}$), that is, roots of the polynomial $x^n-1$.

I know there can be at most $n$ roots (since $x^n-1$ is of degree $n$), but how can we show that all $n$ roots are distinct?

I also know that in fields of characteristic zero, all irreducible polynomials have zeros of multiplicity 1. But $x^n-1$ is reducible, so I think there must be some additional arguments that are escaping me.

  • 6
    $\begingroup$ If $x^n-1$ had repeated roots then they would have to be common roots of $x^n-1$ and its derivative... $\endgroup$ – Daniel Schepler Sep 14 '18 at 23:13
  • $\begingroup$ Does using derivatives in this context require invoking theorems from calculus/analysis? Because don't those presuppose that we are working in $\mathbb{R}$ as opposed to generic fields? $\endgroup$ – WillG Sep 14 '18 at 23:57
  • $\begingroup$ Or is there a way to use formal derivatives for polynomials to prove this without relying on analysis concepts? $\endgroup$ – WillG Sep 14 '18 at 23:58
  • $\begingroup$ Yes, use the formal derivative $Df$ of a polynomial $f$. You can prove that $f$ and $Df$ have a common root $\alpha$ iff $f$ has a repeated root at $\alpha$ (that is, $(x-\alpha)^2$ divides $f$). $\endgroup$ – Daniel Mroz Sep 15 '18 at 0:00
  • $\begingroup$ Right, the formal derivative works, i.e. it satisfies the Leibniz product rule, so if you had $f(x) = g(x) (x-a)^n$ then $f'(x) = g'(x) (x-a)^n + g(x) \cdot n (x-a)^{n-1} = (x-a)^{n-1} (g'(x) (x-a) + n g(x))$. $\endgroup$ – Daniel Schepler Sep 15 '18 at 0:01

Suppose we have a primitive $n$th root of unity, $\omega$. Then for all $1 < a \le n$, $(\omega^a)^n = (\omega^n)^a = 1^a = 1$. If $1 \le a < b \le n$ then $\omega^a = \omega^b \implies \omega^{b-a} = 1$, contradicting the assumption that $\omega$ is a primitive $n$th root of unity. Therefore the existence of a primitive $n$th root of unity implies that there are $n$ distinct $n$th roots of unity, independently of the characteristic.

Now, let $\omega$ be a repeated root of $x^n - 1$ in the splitting field of $x^n - 1$ over $F$. It must be a primitive $m$th root of unity for some $m$ which is a proper factor of $n$. Let $k = \frac nm$. Then $$\gcd\left(\frac{x^{km}-1}{x^m-1}, x^m - 1\right) \neq 1$$ Now, $\gcd(p(x), q(x)) = \gcd(p(x) - q(x)r(x), q(x))$. Taking $r(x) = \sum_{i=0}^{k-2} (k-1-i)x^{im}$ we have $$\frac{x^{km} - 1}{x^m-1} - (x^m-1) \sum_{i=0}^{k-2} (k-1-i)x^{im} = \frac{(x^{km} - 1) - (x^m-1)^2 \sum_{i=0}^{k-2} (k-1-i)x^{im}}{x^m-1}$$ But that $r$ was carefully chosen to telescope: $$(x^m-1)^2 \sum_{i=0}^{k-2} (k-1-i)x^{im} = x^{km} -k x^{m} + (k-1)$$ so that $$\gcd\left(\frac{x^{km}-1}{x^m-1}, x^m - 1\right) = \gcd(k, x^m - 1)$$ This is non-constant iff $k$ is a multiple of the characteristic of $F$.


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.