Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

Let $l$ be an odd prime number and $\zeta$ be an $l$-th primitive root of unity in $\mathbb{C}$. Let $\mathbb{Q}(\zeta)$ be the cyclotomic field and $\alpha$ be a non-zero element of $\mathbb{Q}(\zeta)$.

There exists a polynomial $f(X) \in \mathbb{Q}[X]$ such that $\alpha = f(\zeta)$. Let $N(\alpha) = f(\zeta)f(\zeta^2)...f(\zeta^{l-1})$.

From $\bar\zeta = \zeta^{-1}$ it follows that $\bar f(\zeta) = f(\zeta^{-1})$. Likewise, $\bar f(\zeta^i) = f(\zeta^{-i})$ for $i = 1,2,\cdots,l - 1$. Since $f(\zeta^i)\bar f(\zeta^i) = |f(\zeta^i)|^2 > 0$, it follows that $N(\alpha) > 0$.

We used the fact that the field of complex numbers $\mathbb{C}$ has an $l$-th primitive root of unity. It seems to me that this fact can only be proved by using some (elementary) analysis. My question is:

Can we prove $N(\alpha) > 0$ purely algebraically?

In other words, can we prove $N(\alpha) > 0$ without using the field of real numbers? Please note that $\mathbb{Q}(\zeta) \cong \mathbb{Q}[X]/(1 + X + ... + X^{l-1})$ can be constructed purely algebraically from $\mathbb{Q}$.

share|cite|improve this question
I edited your question so that it is a little easier to read. Feel free to undo the edit if you prefer it as it was before. – Ragib Zaman Jul 22 '12 at 5:12
That's nice. Thanks. – Makoto Kato Jul 22 '12 at 5:14
$\mathbb{C}$ is a red herring; it has nothing to do with the argument, and was presumably only mentioned out of habit. You could replace $\mathbb{C}$ with $\overline{\mathbb{Q}}$, but it would still be unrelated. But as long as I'm on the unrelated issue, you may be interested in the notion of a "real closed field".... – Hurkyl Jul 22 '12 at 5:39
@Hurkyl I'm just asking a proof using only elementary properties of $\mathbb{Q}$. – Makoto Kato Jul 22 '12 at 12:14
What part of the proof do you think isn't an elementary property of $\mathbb{Q}(\zeta)$? I'm honestly not sure what you think this proof uses that you want to avoid. – Hurkyl Jul 22 '12 at 14:37

Yes, this can be proved using "only elementary properties of $\mathbb{Q}$", since the original argument can be rewritten in this manner. Let $K=\mathbb{Q}[X]/(1+X+X^2+\dots+X^{\ell-1})$, and let $\zeta$ be the image of $X$ under the natural map $\mathbb{Q}[X]\to K$. Let $L=\mathbb{Q}(\zeta+1/\zeta)$. All that needs to be done is to define a total ordering on $L$ which makes $L$ into an ordered field, and for which $-2\le\zeta+1/\zeta\le 2$. For, if this is accomplished then the classical proof works: any element of $K$ is $a+b\zeta$ with $a,b\in L$, and $$ N_{K/L}(a+b\zeta)=(a+b\zeta)(a+b/\zeta)=a^2+b^2+ab(\zeta+1/\zeta)\ge a^2+b^2-2ab = (a-b)^2 \ge 0, $$ so for any $h(X)\in\mathbb{Q}[X]$ it follows that $$ N_{K/\mathbb{Q}}(h(\zeta)) = \prod_{i=1}^{(l-1)/2} h(\zeta^i) h(1/\zeta^i) = \prod_{i=1}^{(l-1)/2} N_{K/L}(h(\zeta)) \ge 0. $$

It remains only to define a total ordering on $L$ which is compatible with the field operations, and for which $-2\le\zeta+1/\zeta\le 2$. If $\ell=3$ then $\mathbb{Q}(\zeta+1/\zeta)=\mathbb{Q}$, and there is nothing to prove. So assume $\ell>3$. Let $f(X)$ be the minimal polynomial of $\zeta+1/\zeta$ over $\mathbb{Q}$. For any rational number $\beta$, define $\zeta+1/\zeta$ to be less than $\beta$ if $\beta$ is greater than all rational numbers which are smaller than all positive rational numbers $\gamma$ for which $f(\gamma)\cdot f(0)<0$. If $\zeta+1/\zeta$ is not less than $\beta$, then define $\zeta+1/\zeta$ to be greater than $\beta$. Exercise for the OP: verify that this ordering has a unique extension to a total ordering on $L$ which is compatible with the field operations. QED

share|cite|improve this answer
I think you mean "with $a, b \in L$". (Then define $|a|$ by means of the assigned total ordering on $L$.) – user43208 Aug 23 '13 at 13:33
Thanks, you're right of course. I edited to fix that. – Michael Zieve Aug 23 '13 at 14:40
I can imagine there being other proofs which are more satisfying than the one above. For instance, if we write $f(X)=\sum_{i=0}^{l-2} a_i X^i$ with $a_i\in\mathbb{Q}$, then the norm from $\mathbb{Q}(\zeta)$ to $\mathbb{Q}$ of $f(\zeta)$ equals $H(a_0,a_1,\dots,a_{l-2})$ for some $H(x_0,\dots,x_{l-2})\in\mathbb{Z}[x_0,\dots,x_{l-2}]$. The question is to show that this polynomial $H$ takes nonnegative values at all points of $\mathbb{Q}^{l-1}$. Perhaps $H$ is a sum of squares in $\mathbb{Z}[x_0,\dots,x_{l-2}]$? – Michael Zieve Aug 23 '13 at 15:22
Dear Michael: can you give some hint on how to solve the exercise in the end? – Dan Petersen Aug 23 '13 at 20:38
(Obviously you can prove that $L$ is formally real and that there's an ordering with $(\zeta + \zeta^{-1})^2 < 4$ by using the obvious embedding of $L$ in $\mathbb R$. But it should be clear that this is not the sought for solution.) – Dan Petersen Aug 23 '13 at 21:24

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

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