1

$Sin(nx)$ and $cos(nx)$ can be expressed as polynomials of sin(x) and cos(x). I am interested in the way of this expression and a proof (preferably at secondary-school level) as well.

gen
  • 1,518
  • Are you familiar with the following formulas? $x$ is any real number. $e^{ix} = \cos x + i \sin x$, $(e^{ix})^n = \sum_{j=0}^n {n\choose j}(\cos x)^j(i\sin x)^{n-j}$, and also $(e^{ix})^n = e^{inx} = \cos(nx) + i \sin(nx).$ – jflipp Nov 18 '14 at 21:52
  • well I am, but I would much more like to see a solution based on identities of trigonometric functions. But I would love to see your solution based on the above as well. – gen Nov 18 '14 at 22:15

2 Answers2

6

Here's a proof that $\sin(2x)$ can't be written as a polynomial in $\sin x$.

Recall that $\sin(2x) = 2\cos(x)\sin(x)$. Now assume we have a polynomial $p \in \mathbb R[x]$ such that $\sin(2x) = p(\sin(x)).$ Then we also have $$ 2\cos(x)\sin(x) = p(\sin(x)). $$ Since the l.h.s. is $0$ at $x = 0$, we must have $0 = p(\sin(0)) = p(0)$. This means that $p$ doesn't have a constant term, so we can write $p(x) = xq(x)$ with another polynomial $q \in \mathbb R[x].$ Thus we have $$ 2\cos(x)\sin(x) = \sin(x)q(\sin(x)) $$ for any $x \in \mathbb R$. So we have $$ 2\cos(x) = q(\sin(x)) \quad \forall x \in \mathbb R\setminus\pi\mathbb Z $$ since $\pi\mathbb Z$ is the zero set of $\sin(x)$. Since both the l.h.s. and the r.h.s of the preceding equation are continuous and $\pi\mathbb Z$ is discrete in $\mathbb R$, we get by continuity $$ 2\cos(x) = q(\sin(x)) \quad \forall x \in \mathbb R. $$ Now we square this equation and get $$ 4(1-\sin^2(x)) = q(\sin(x))^2 \quad \forall x \in \mathbb R. $$ Put $r(x) = 4(1-x^2) \in \mathbb R[x]$. With this we can write $$ r(\sin(x)) = q(\sin(x))^2 \quad \forall x \in \mathbb R. $$ Since $\sin(x)$ ranges over $[-1,1]$ as $x$ ranges over $\mathbb R$, this is equivalent to $$ r(x) = q(x)^2 \quad \forall x \in [-1,1]. $$ Now we use the fact that if two polynomials coincide on a non-empty open interval (and $[-1,1]$ contains $(-1,1)$), then these polynomials must be equal. So we get $$ r(x) = q(x)^2 \quad in\ \mathbb R[x]. $$ But $$ r(x) = 4(1-x^2) = 4(1-x)(1+x)\quad in\ \mathbb R[x] $$ is not a square.

So we have reached a contradiction, and the polynomials $q$ and $p$ don't exist.

jflipp
  • 4,786
  • Wow, found this years later as I was struggling with the same problem myself. The proof is really clear, but is there any intuitive reason why $\cos(nx)$ has polynomial representations in terms of $\cos(x)$ but there is no similar relationship for $\sin$? Does this have something to do with the fact that $\cos$ is even and $\sin$ is odd? I'm really curious about this. – K.defaoite Aug 11 '20 at 02:40
3

Here you are.

Let me state the formulas again. For any $x \in \mathbb R$, we have the following. $$ e^{ix} = \cos x + i\sin x $$ $$ (e^{ix})^n = (\cos x + i\sin x)^n = \sum_{j=0}^n{n\choose j}(\cos x)^{n-j}(i\sin x)^j $$ $$ (e^{ix})^n = e^{inx} = \cos(nx) + i\sin(nx) $$ Note that the second formula is a little bit different than in my comment.

Now we equate real and imaginary parts in the second and third formula for $(e^{ix})^n$.

From the real parts we get $$ \cos(nx) = \sum_{k = 0}^{\lfloor \frac{n}{2} \rfloor} {n\choose 2k} (\cos x)^{n-2k}(i\sin x)^{2k} = \sum_{k = 0}^{\lfloor \frac{n}{2} \rfloor} {n\choose 2k} (\cos x)^{n-2k}(-1)^k(1-\cos^2x)^k. $$ Here, the r.h.s. is a polynomial in $\cos x$.

From the imaginary parts we get $$ \begin{align*} i \sin(nx) & = \sum_{k = 0}^{\lfloor \frac{n-1}{2} \rfloor} {n\choose 2k+1} (\cos x)^{n-2k-1}(i\sin x)^{2k+1} \\ & = i\sin x\sum_{k = 0}^{\lfloor \frac{n-1}{2} \rfloor} {n\choose 2k+1} (\cos x)^{n-2k-1}(-1)^k(1-\cos^2x)^k \end{align*} $$ Here, the sum on the r.h.s. is a polynomial in $\cos x$. And of course we can cancel the common factor $i$.

jflipp
  • 4,786
  • Please see my edit to the last equation of your anwser. If it is, right then your formula isn't a polynomial of cos x. Also, isn't it possible to derive a sin(nx) in form of a polynomial of sin(x) ? – gen Nov 19 '14 at 09:45
  • I'm not very familiar with the editing and reviewing system on this site. It seems that your edit of my answer above was rejected. Can you describe what you think is wrong with my last equation? Regarding an expression for $\sin(nx)$ as a polynomial in $\sin(x)$, see my answer below. – jflipp Nov 19 '14 at 14:13
  • So equation number 5 begins with isin(nx) and ends with isin(x). Is that right? Or the i*sin(x) part is a factor of the last line (e.g.) the sum? – gen Nov 19 '14 at 18:02
  • Yes. The sum in the very last line and the $i\sin(x)$ at the right end of the preceding line belong together. I'll improve the formatting there. – jflipp Nov 19 '14 at 21:34