How do I prove that $\cos{\frac{2\pi}{7}}\notin\mathbb{Q}$?

Should I use some geometrical approach or apagoge?

  • 5
    $\begingroup$ Niven establishes this. See this as well. $\endgroup$ Oct 11 '11 at 12:33
  • $\begingroup$ You may be interested to see this: $\text{Recurrent proofs of the irrationality of certain trigonometric values}$ here at this link: arxiv.org/abs/0911.1933 $\endgroup$
    – user9413
    Oct 11 '11 at 14:36

If $\cos \frac{2 \pi}{7}$ is rational, then $i\sin \frac{2 \pi}{7}=\sqrt{\cos^2 \frac{2 \pi}{7}-1}$ is a quadratic irrational, and hence so is $\cos \frac{2 \pi}{7}+i\sin \frac{2 \pi}{7}$. But $\cos \frac{2 \pi}{7}+i\sin \frac{2 \pi}{7}$ is a primitive $7$th root of $1$, and so has minimal polynomial $x^6+x^5+x^4+x^3+x^2+x+1$, which is not a quadratic.

  • 1
    $\begingroup$ There is a ${}^{2}$ missing under the $\sqrt{}$ ... $\endgroup$
    – martini
    Oct 11 '11 at 12:48
  • $\begingroup$ @martini: yes, thanks $\endgroup$ Oct 11 '11 at 12:51
  • $\begingroup$ Does anyone else see the polynomial $x^6+x^5+\ldots+1$ at the top of the post instead of where it should be? This only seems to happen when I'm using Chrome, not in Opera or Firefox... $\endgroup$
    – user1729
    Oct 11 '11 at 14:48
  • $\begingroup$ @Swlabr: Take a screenshot and please make a meta post. $\endgroup$ Oct 11 '11 at 15:03
  • $\begingroup$ @J. M.: Done... $\endgroup$
    – user1729
    Oct 11 '11 at 15:21

Since Chebyshev polynomials have the property $T_n\left( \cos \theta \right) = \cos(n \theta)$, we can use this property with $n=7$ and $\theta = \frac{2 \pi}{7}$.

Let $z = \cos\left( \frac{2\pi}{7} \right)$. Then $T_7(z) = 1$

Let $p(x) = T_7(x)-1 = 64 x^7 - 112 x^5 +56 x^3 - 7x - 1 = (x-1)( 8 x^3 + 4 x^2 - 4 x -1 )^2.$

So $z$ must be a root of $q(x) = (2 x)^3 + (2 x)^2 - 2 (2 x) - 1$. Observe that $2z$ is a root of a monic polynomial, and thus is an algebraic integer. If $z$ were rational, then $2z$ must be an integer. Since $-2 \le 2 \cos \theta \le 2$, there are only 5 possibilities to consider, which are ruled out because $q(1) = 7$, $q(-1) = q(0) = q(1/2) = -1$, $q(-1/2) = 1$.

  • 1
    $\begingroup$ It looks quite like the argument here. Nice! $\endgroup$ Oct 11 '11 at 13:19
  • $\begingroup$ $T_7(z)=1$, not $-1$. $\endgroup$ Oct 11 '11 at 13:23
  • $\begingroup$ Following Chris, the polynomial considered should then be $T_7(x)-1=(x-1)(8x^3+4x^2-4x-1)^2$... $\endgroup$ Oct 11 '11 at 13:33
  • $\begingroup$ @Chris and J.M. Thanks for catching the typo. I have edited the post to correct it. $\endgroup$
    – Sasha
    Oct 11 '11 at 14:54

Here is a slightly different way of phrasing Chris Eagle's argument:

If $\cos(2\pi/7)$ were rational, $X^2-2\cos(2\pi/7)\,X+1$ would be a factor of $$X^6+X^5+X^4+X^3+X^2+X+1$$ in $\mathbb Q[X]$, contradiction.

EDIT. Let $n$ be an integer $\ge3$. Then the distinct conjugates of $\cos(2\pi/n)$ are the $\cos(2k\pi/n)$, where $k$ runs over the integers prime to $n$ lying between $1$ and $n/2$.

In particular the degree of $\cos(2\pi/n)$ over $\mathbb Q$ is $\varphi(n)/2$, where $\varphi$ is Euler's totient function.

This results immediately from the following well known facts.

If $\zeta$ is a primitive $n$-th root of $1$ in $\mathbb C$, then $\mathbb Q(\zeta)/\mathbb Q$ is Galois of degree $\varphi(n)$.

Let $G$ be the Galois group and $k$ an integer prime to $n$. Then there is a unique $\sigma_k$ in $G$ satisfying $\sigma_k(\zeta)=\zeta^k$.

In particular $k\mapsto\sigma_k$ induces an isomorphism from the multiplicative group of $\mathbb Z/n\mathbb Z$ onto $G$.

  • $\begingroup$ This is a very nice way of putting it. $\endgroup$ Oct 11 '11 at 15:03

Here is another answer which uses some machinery:

Given $n$, consider $\zeta_n =e^{2\pi i/n}$. Then $$\frac{1}{2}\left(\zeta_n +\zeta_n^{n-1}\right)=\cos \left(\frac{2\pi}{n}\right).$$

The rational numbers are those elements of $\mathbb{Q}[\zeta_n]$ which are fixed by the action of the Galois group. The Galois Group of $\mathbb{Q}[\zeta_n]$ is isomorphic to $(\mathbb{Z}/n\mathbb{Z})^\times$ and is given by $\zeta_n\rightarrow \zeta_n^a$ for integers $a$ satisfying $(a,n)=1$. From this, we see that $\frac{1}{2}\left(\zeta_n +\zeta_n^{n-1}\right)$ is fixed by the Galois group only when all of the elements either fix $\zeta_n$ and $\zeta_n^{n-1}$ or swap their places. But if there exists $a\neq -1,1$ and $(a,n)=1$ then $\zeta_n\rightarrow \zeta_n^a$ won't fix this element, so that it must be irrational. This only occurs if $|(\mathbb{Z}/n\mathbb{Z})^\times|>2$, or in other words if $n\neq 3,4,6$. In particular we see that for $n=7$ it is irrational.


As a mirror of Sasha's answer, consider the Chebyshev polynomial of the second kind, $U_6(x)=(8x^3+4x^2-4x-1)(8x^3-4x^2-4x+1)$. That $\cos\dfrac{2\pi}{7}$ is a root is established from expressing $U_n(x)$ as $\dfrac{\sin((n+1)\arccos\,x)}{\sqrt{1-x^2}}$. We can then test for rational roots in the two cubic factors of $U_6(x)$, and find that they have no rational roots at all...


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.