27
$\begingroup$

I know that $\cos(\pi/n)$ is a root of the Chebyshev polynomial $(T_n + 1)$, in fact it is the largest root of that polynomial, but often that polynomial factors. For example, if $n = 2 k$ then $\cos(\pi/n)$ is the largest root of $T_k$, which is a polynomial of lower degree, and if $n = 3$ then $\cos(\pi/n)$ is a root of $2 x - 1$, again lower degree than $T_3 + 1$.

How can I compute, for a given $n$, a polynomial in $\mathbb{Q}[x]$ of minimal degree that $\cos(\pi/n)$ is a root of?

$\endgroup$
0

5 Answers 5

56
$\begingroup$

The minimal polynomial of $\cos(2\pi/n)$ (by William Watkins and Joel Zeitlin, The American Mathematical Monthly Vol. 100, No. 5 (May, 1993), pp. 471-474) has full clarity on this matter (just take their result for even $n$ to resolve your case).

$\endgroup$
3
  • $\begingroup$ It's really surprising this was published in 1993. If you had asked me before, I would have wagered it should have been done by 1950. Do you know if there is any substantial difficulty involved, or is it just that no one bothered before? $\endgroup$
    – tomasz
    Commented Nov 28, 2017 at 10:54
  • 4
    $\begingroup$ @tomasz: A slightly less explicit recipe (more or less the same as the one Fedor Petrov gives in another answer) was given already by D.H.Lehmer (A Note on Trigonometric Algebraic Numbers, The American Mathematical Monthly Vol. 40, No. 3 (Mar., 1933), pp. 165-166, jstor.org/stable/2301023) in 1933, 60 years before Watkins and Zeitlin. So in a way it has been known for a very long time, but Watkins and Zeitlin found a nice formula relating it to Chebyshev's polynomials. $\endgroup$ Commented Nov 28, 2017 at 13:27
  • $\begingroup$ I see. That seems more reasonable, thanks for clearing that up. $\endgroup$
    – tomasz
    Commented Nov 28, 2017 at 18:19
17
$\begingroup$

I guess you mean a polynomial $p(x)$ with rational coefficients. Then, once $\cos {\pi/n}$ is a root of $p(x)$, $\deg p=d$, $e^{i\pi/n}$ is a root of a polynomial $t^dp((t+1/t)/2)$. But $e^{i\pi/n}$ is a root of a cyclotomic polynomial $g(t)=\Phi_{2n}(t)$, which is irreducible, thus $\Phi_{2n}(t)$ should divide $t^dp((t+1/t)/2)$, that is, for any $k\in \{0,1,\dots,2n-1\}$ coprime to $2n$ the number $\cos \pi k/n$ is also a root of $p(x)$. For $k$ and $2n-k$ we get the same value of a cosine, so we get $\varphi(2n)/2$ different roots. Actually the polynomial with all these $\varphi(2n)/2$ roots has rational coefficients. To see this observe that $\Phi_{2n}(t)=t^{\varphi(2n)/2}H(t+1/t)$ for some polynomial $H$, which of course has rational (even integer) coefficients, and this $H$ has roots $2\cos \pi k/n$ for $k$ coprime to $2n$.

$\endgroup$
5
$\begingroup$

There is a Wikipedia entry dedicated to this, which contains an alternative method to compute the minimal polynomial of $2\cos(\pi/n)$, which is essentially the same as for $\cos(π/n)$. In fact, denoting the minimal polynomial of $2\cos(π/n)$ by $\Psi_n(x)$, we have by the quoted AMM article in Vladimir Dotsenko's answer for odd $n=2k+1$ $$\prod_{d \mid n}\Psi_d(x) = 2\Big(T_{k + 1}(\frac x2) - T_k(\frac x2)\Big)$$ and for even $n=2k$ $$\prod_{d \mid n}\Psi_d(x) = 2\Big(T_{k + 1}(\frac x2) - T_{k-1}(\frac x2)\Big).$$ But there is in fact no need to calculate the Chebyshev polynomials. Defining for odd $n=2k+1$ $$\chi_{n}(x):= \sum_{k=0}^m (-1)^{\lfloor k/2\rfloor}\binom {m-\lfloor (k+1)/2\rfloor}{\lfloor k/2\rfloor} x^{m-k},$$ we have directly

$$\prod _{d\mid n}\Psi_{d}( x)=(x-2)\chi_{n}(x),$$ noting that $\Psi_{1}( x)=x-2$.

For even $n=2k$, we just need to sum up two of those polynomials: $$\prod _{d\mid n}\Psi_{d}( x)=(x-2)\Big(\chi_{n+1}(x)+\chi_{n-1}(x)\Big).$$

$\endgroup$
2
$\begingroup$

Maple 2017.3 helps you. For example,

convert(cos((1/7)*Pi), RootOf);
RootOf(8*_Z^3-4*_Z^2-4*_Z+1, .9009688679)
convert(cos((1/27)*Pi), RootOf)
RootOf(512*_Z^9-1152*_Z^7+864*_Z^5-240*_Z^3+18*_Z-1, .9932383577)

See convert and RootOf for info.

Addition. Also

convert(sin(7*Pi*(1/22)), RootOf);
RootOf(32*_Z^5+16*_Z^4-32*_Z^3-12*_Z^2+6*_Z+1, .8412535328)
convert(sin(7*Pi*(1/22))^2, RootOf);
-(1/4)*(RootOf(_Z^20-_Z^18+_Z^16-_Z^14+_Z^12-_Z^10+_Z^8-_Z^6+_Z^4-_Z^2+1, index = 1)^7+
RootOf(_Z^20-_Z^18+_Z^16-_Z^14+_Z^12-_Z^10+_Z^8-_Z^6+_Z^4-_Z^2+1, index = 1)^15)^2
$\endgroup$
1
$\begingroup$

In PARI/ GP :

r2n = Mod(x, polcyclo(2*n))

minpoly((r2n + 1/r2n) / 2)

$\endgroup$

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .