I'm working on a special kind of graphs. To prove some uniqueness, I need to prove that the polynomial \begin{equation} x^{8}-7x^{6}+14x^{4}-8x^{2}+1 \end{equation} does not have any root of the form \begin{equation} 2\cos\frac{(2k+1)\pi}{2n} \quad k\in \lbrace 0,1,\cdots , n-1 \rbrace , n \in \mathbb{E}. \end{equation} Can anybody help me?


  • $\begingroup$ Note that all roots of the above polynomial are real. $\endgroup$ – A. Mpi Dec 6 '17 at 12:53
  • 3
    $\begingroup$ This polynomial is solvable, because it is equal to $g(x^2)$ where $g(x) = x^4 - 7x^3 + 14x^2 - 8x + 1$; and quartic polynomials are always solvable. Thus you can find the roots explicitly and therefore determine whether they are or are not of the form that you ask. $\endgroup$ – Stanley Yao Xiao Dec 6 '17 at 13:00
  • $\begingroup$ In fact the polynomial $g(x) = x^4 - 7x^3 + 14x^2 - 8x + 1$ has Galois group isomorphic to $C_4$, and its roots can be found explicitly using the the identity $g(x) = x^2(x-2)^2 - 3x(x-1)(x-2) + (x-1)^2$. Since this polynomial is so special, I am very curious as to how it arose. What graph does this originate from, say? $\endgroup$ – Stanley Yao Xiao Dec 6 '17 at 13:40

The degree of the minimal polynomial of your cos is bounded below by phi(n)/2 since it generates a real field inside the cyclotomic field. If phi(n)/2 is bigger than 8 you are done. For small n you can check one by one with computer (or write the minimal polynomial and see it doesn’t divide that polynomial or compute discriminant, which is an easy thing for cosine as the real part of a root of unity)


The claim is simply not true. A simple check (e.g. following the suggestion of Stanley Yao Xiao, or by substituting $ x = cos \alpha $ and using trigonometric identities) shows that this polynomial has 8 real roots and they are precisely of this form

$$ 2 cos\frac{\pi}{30}, 2 cos\frac{7\pi}{30}, 2 cos\frac{11\pi}{30},, 2 cos\frac{13 \pi}{30}, 2 cos\frac{17 \pi}{30}, 2 cos\frac{19\pi}{30}, 2 cos\frac{23 \pi}{30}, 2 cos\frac{29\pi}{30} $$

  • $\begingroup$ Sorry, i forgot to say $n$ is even. $\endgroup$ – A. Mpi Dec 6 '17 at 13:59
  • $\begingroup$ If $n$ is even, i checked by computer and it doesn't have root of the given form. $\endgroup$ – A. Mpi Dec 6 '17 at 14:04
  • 1
    $\begingroup$ Indeed, if you take $x = 2 \cos(\alpha)$, $z = e^{i\alpha}$ is a root of $z^{16}+z^{14} - z^{10} - z^8 - z^6 + z^2 + 1$, which is $C_{30}(z^2)$ where $C_{30}$ is the 30'th cyclotomic polynomial. $\endgroup$ – Robert Israel Dec 6 '17 at 23:47

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.