I am trying to prove:

If $\cos(\pi\alpha) = \frac{1}{3}$ then $\alpha \in \mathbb{R} \setminus \mathbb{Q}$

So far, I've tried making it into an exponential, since exponentials are easier to manipulate (at least for me), when compared to $\cos$ or $\sin$. So:

$$\cos(\pi\alpha)^2 + \sin(\pi\alpha)^2 = 1$$ $$\Big(\frac{1}{3}\Big)^2 + \sin(\pi\alpha)^2 = 1$$ $$\sin(\pi\alpha)^2 = \frac{8}{9}$$ $$\sin(\pi\alpha) = \frac{\sqrt{8}}{3}$$ $$\sin(\pi\alpha) = \frac{2\sqrt{2}}{3}$$

Then we can use Euler's formula:

$$e^{i\pi\alpha} = \cos(\pi\alpha) + i\sin(\pi\alpha) = \frac{1}{3} + i\frac{2\sqrt{2}}{3} = \frac{1+i2\sqrt{2}}{3}$$

Now take the log of both sides:

$$ \ln(e^{i\pi\alpha}) = \ln(\frac{1+i2\sqrt{2}}{3}) = \ln(1+i2\sqrt{2}) - \ln(3) = i\pi\alpha$$

$$\therefore \alpha = \frac{\ln(1+i2\sqrt{2}) - \ln(3)}{i\pi}$$

But, here I get stuck, and don't know how to show it is irrational, any ideas? (Even though it looks pretty darn irrational to me...) Should I be trying something else? I'm so lost...

  • $\begingroup$ Write $\alpha$ as $a+bi$ and use the angle sum formula for $\cos$. That should quickly let you establish that $\alpha\in\mathbb{R}$. $\endgroup$ – alex.jordan Sep 1 '14 at 22:02
  • $\begingroup$ If it was rational, $e^{i\pi a}=\dfrac13+i\dfrac{2\sqrt2}3$ would have to be a root of unity (why?). However, I'm not sure where to go from there. $\endgroup$ – Akiva Weinberger Sep 1 '14 at 22:03
  • 1
    $\begingroup$ As for not being in $\mathbb{Q}$, there is this theorem. $\endgroup$ – alex.jordan Sep 1 '14 at 22:05
  • $\begingroup$ @alex.jordan We want to show that it's irrational. We already know it's real. $\endgroup$ – Akiva Weinberger Sep 1 '14 at 22:05
  • 1
    $\begingroup$ I think $\mathbb R \setminus \mathbb Q$ is correct for "$\alpha$ is an irrational number", which is after all a much stronger statement than merely "$\alpha$ is not a rational number" (since neither is, say, a brick). $\endgroup$ – Ben Millwood Sep 2 '14 at 10:08

Just in case it is not assumed that $\alpha\in\mathbb{R}$, let $\alpha=a+bi$ with $a,b\in\mathbb{R}$. Then $$\begin{align} \cos(a\pi+b\pi i)&=\cos(a\pi)\sin(b\pi i)+\sin(a\pi)\cos(b\pi i)\\ \frac13&=i\cos(a\pi)\sinh(b\pi)+\sin(a\pi)\cosh(b\pi)\\ \end{align}$$ Since the left side is real, either $\cos(a\pi)$ or $\sinh(b\pi)$ is $0$. The former implies $\frac13=\pm\cosh(b\pi)$, which is impossible, So $\sinh(b\pi)=0$, which implies $b=0$, and so $\alpha\in\mathbb{R}$.

Now assume that $\alpha$ is rational: $\alpha = \frac{a}{b}$ with $a,b\in \mathbb{Z}$ and consider

$$\left(e^{i\alpha\pi}\right)^b = e^{i\pi a} = (-1)^a$$

Now use that $e^{i\alpha\pi} = \frac{1}{3} \pm i\frac{2\sqrt{2}}{3}$ to derive a contradiction.

  • 1
    $\begingroup$ Classical approach (looking what a fraction as solution would imply). Nice. $\endgroup$ – mvw Sep 1 '14 at 22:07
  • $\begingroup$ It's not clear that "the opposite" is that $\alpha$ is rational. OP seems to not be assuming that $\alpha$ is real, and needs that established too (which is easy enough). $\endgroup$ – alex.jordan Sep 1 '14 at 22:08
  • $\begingroup$ @alex.jordan: Honestly, I thought it had to be real... $\endgroup$ – Dair Sep 1 '14 at 22:09
  • $\begingroup$ Hope you don't mind me editing your answer. Sorry if I overstepped---I won't be upset if you roll back. $\endgroup$ – alex.jordan Sep 1 '14 at 22:17
  • $\begingroup$ @alex.jordan Fair enough. I took it for granted that $\alpha$ was real here. Yeah, no problem, saves me the work:) $\endgroup$ – Winther Sep 1 '14 at 22:17

Another possible approach. Assume, by contradiction, that $\alpha=\frac{p}{q}\in\mathbb{Q}$. Consider the polynomial: $$ g(x)=T_{2q}(x)-1 $$ where $T_{2q}(x)$ is a Chebyshev polynomial of the first kind. We have that $\frac{1}{3}$ is a rational root of $g(x)$, but this contradicts the rational root theorem, since $g(0)=-2$ and the leading coefficient of $g(x)$ is a power of $2$.

  • 2
    $\begingroup$ Interesting! I have never heard of a Chebyshev polynomial! I will definitely look into this when I get further along in my mathematical career! :D $\endgroup$ – Dair Sep 1 '14 at 23:16
  • $\begingroup$ Heard about it but nice to see the application for this! $\endgroup$ – Elimination Feb 7 '16 at 15:18

A different approach is to use Dirichlet's well-known discontinuous function

$$ f(x)=\lim_{k\to\infty}\left(\lim_{j\to\infty}(\cos(k!\pi x))^{2j}\right), $$

which is $1$ if $x$ is rational and $0$ if $x$ is irrational.

Since $\cos(\pi\alpha)=\frac{1}{3}$ implies $\alpha=\pm\frac{1}{\pi}\arccos\frac{1}{3}+2n$, with $n$ an arbitrary integer, we get

$$ f(\alpha)=\lim_{k\to\infty}\left(\lim_{j\to\infty}[\cos(\pm k!\arccos(1/3))]^{2j}\right). $$

Next, we use the fact that $\cos(nx)=T_{n}(\cos x)$, where $T_n$ denotes the $n$-th Chebyshev polynomial of the first kind. We see that $$ \cos(\pm k!\arccos(1/3))]=T_{k!}(1/3) $$

Using the recurrence relation $T_{n+1}(1/3)=\frac{2}{3}T_n(1/3)-T_{n-1}(1/3)$ (with $T_0(1/3)=1$ and $T_1(1/3)=1/3$), it is not difficult to prove that $|T_n(1/3)|<1$ for all $n\geq 1$. (I did it by solving the recurrence explicitly.) It is then trivial to show that both the outer and the inner limit equal $0$.


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.