Of course for $x\neq 0 $ and $\alpha$ in radians. Can you define them?

  • 1
    $\begingroup$ Sine is transcendental. Any rational input would get your an irrational number. $\endgroup$ – Shuhao Cao May 8 '13 at 23:45
  • 1
    $\begingroup$ @ShuhaoCao, wrong. $\sin 0 = 0$. $\endgroup$ – vonbrand May 9 '13 at 1:05
  • 1
    $\begingroup$ @vonbrand Ok, you are right, except for the input $0$, :p $\endgroup$ – Shuhao Cao May 9 '13 at 1:10

I think the Lindenmann-Weierstrass theorem is an overkill for this problem. Ivan Niven got an elementary proof of

For any rational $r\neq 0$, $\cos r$ is irrational.

in his book Irrational Numbers by exploiting the fact that rational number being the roots of a polynomial, the proof is not lengthy, and is very inspiring, I learned the proof back in college days. Here I just copy the proof and adapt it for sine function:

Claim: For any rational $r\neq 0$, $\sin r$ is irrational.

For a positive $r = a/b$, where $a,b\in \mathbb{Z}^+$. For $p$ being an odd prime to be specified, define $$ f(x) = \frac{x^{p-1}(a-bx)^{2p}(2a-bx)^{p-1}}{(p-1)!} = \frac{(r-x)^{2p}\big(r^2 - (r-x)^2\big)^{p-1}b^{3p-1}}{(p-1)!}\tag{1} $$ For $0<x<r$: $$ 0<f(x)< \frac{r^{2p}(r^2)^{p-1}b^{3p-1}}{(p-1)!} = \frac{r^{4p-2}b^{3p-1}}{(p-1)!} \tag{2} $$ Next, define $F(x)$ using all the even derivatives of $f(x)$ $$ F(x) = f(x) - f''(x) + f^{(4)}(x) - f^{(6)}(x) + \cdots - f^{(4p-2)}(x). $$ Thus $$ \big(F'(x)\cos x + F(x)\sin x\big)' = F''(x)\cos x + F(x)\cos x = f(x)\cos x $$ and $$ \int^r_0 f(x)\cos x \,dx = F'(r)\cos r + F(r)\sin r + F(0). \tag{3} $$ Next, what Ivan Niven does is proving $F'(r)=0$ by construction to get rid of $\cos$, computing $F(r)$ and $F(0)$, setting $\sin r$ to be rational, then to get a contradiction.

  • $F'(r)=0$ follows from $f(x)$ has a factor of $(r-x)^2$.

  • For $f(x)$ has form $x^{p-1}g(x)/(p-1)!$, $g(x)$ has all integer coefficients, $f^{(j)}(0)$ is an integer, also notice $f^{(j)}(0)$ is a multiple of $p$ unless $j=p-1$. Direct computation from (1) gives $$ f^{(p-1)}(0) = a^{2p}(2a)^{p-1}. $$ If $p>a$, and $p$ is an odd prime, then $f^{(p-1)}(0)$ is not divisible by $p$. Thus $F(0) = q$, where $q$ and $p$ are relatively prime.

  • The definition (1) implies that $$ f(r-x) = \frac{x^{2p}(a^2 - b^2x^2)^{p-1}b^{p+1}}{(p-1)!}. $$ Using similar argument as above, $f^{(j)}(r)$ is a multiple of $p$. Hence let $F(r) = pm$ for some integer $m$.

Now assume

$\sin r = d/k$ is rational for $d,k\in \mathbb{Z}^+$.

It suffices to show for positive sine value for $\sin(\cdot)$ is an odd function. Plugging all above results into (3): $$ k\int^r_0 f(x)\cos x \,dx = pmd + kq\tag{4} $$ The contradiction will be obtained by choosing $p$:

Choose $p$ large enough such that the right side of (4) is a non-zero integer, and left side is strictly in $(-1,1)$.

For $p>a$ is already required, now let $p>k$, thus $p$ is relatively prime with $kq$, thus right side of (4) is a non-zero integer. Now by (2): $$ \left|k\int^r_0 f(x)\cos x \,dx\right| < kr^3b^2\frac{(r^4b^3)^{p-1}}{(p-1)!} $$ Since $$ \lim_{p\to \infty} \frac{(r^4b^3)^{p-1}}{(p-1)!} = 0 $$ So we can choose $p$ large enough so that the left side of (4) is strictly in $(-1,1)$. Contradiction, hence the claim holds.

Notice Ivan Niven's proof tells us nothing about whether $\sin r$ can be algebraic. For that part I believe we have to borrow the Lindenmann-Weierstrass theorem, if I am wrong here someone please point out.

  • $\begingroup$ @PeterTamaroff Yes, I posted the adapted version. $\endgroup$ – Shuhao Cao May 9 '13 at 2:16

When $\alpha$ is rational and nonzero, $\sin \alpha$ is transcendental. This follows from the Lindenmann-Weierstrass theorem.


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.