4

I currently try to prove the following equation:

$$(\sin(x))^n=\sum_{k=0}^{n}{a_k\cos(kx)+b_k\sin(kx)}$$

with $a_0,...,a_k$ and $b_0,...b_k$ being real numbers for each $n$.

I tried to proof this by induction, then I could use the induction hypothesis for replacing $\sin(x)^{n}$ with $\sin(x)\cdot (\sum_{k=0}^{n}{a_k \cos(kx)+b_k\sin(kx)})$

Do you have any idea, how I can get this in the form $\sum_{k=0}^{n+1}{a_k \cos(kx)+b_k\sin(kx)}$?

Thanks

slkjck3
  • 51

3 Answers3

3

Hint

$$\sin^n(x)=\left(\frac{e^{ix}-e^{-ix}}{2i}\right)^n = ?$$

amWhy
  • 209,954
  • Yeah, binomical theorem but I don't think that this really helps me in this case, I think that it is much easier with addition theorems.. – slkjck3 Nov 25 '18 at 19:26
  • @slkjck3: this pretty much gives you the answer immediately. – Cheerful Parsnip Nov 25 '18 at 22:12
  • @CheerfulParsnip Thanks. {}{}{}{}{}{}{} – hamam_Abdallah Nov 25 '18 at 22:13
  • I don't see that.. – slkjck3 Nov 25 '18 at 23:02
  • @slkjck3: Once you expand it out using the binomial theorem, use that $(e^{ix})^k=e^{ikx}=\cos(kx)+i\sin (kx)$, so that each summand becomes a linear combination of terms of the form you are looking for. – Cheerful Parsnip Nov 25 '18 at 23:42
  • @CheerfulParsnip hey, this is my main account (I'm slkjck3). I had trouble getting this account working but now it's ok. I still don't see how I could use this as a proof. I think that mrtaurho's answer below is much better because in my oppinion I think that you still have to do crazy reshuffling of terms, etc. If not, then it would be really kind if you could explain it a bit more indepth. Because I think you get somethin like $sin((n-k)x)$ which is not equal to $sin(kx)$... – Gykonik Nov 26 '18 at 22:28
  • @Gykonik: It's true you have to do some reindexing but it's not bad. You get terms of the form $\sin((2k-n)x)$ where the coefficient $(2k-n)x$ can range from $-n$ to $n$, and using the oddness of sine, can be converted to all lie between $0$ and $n$ just as you desired. I'll do an example to convince you. Stay tuned. – Cheerful Parsnip Nov 27 '18 at 04:11
2

This is an example to illustrate Hamam Abdallah's method. Let's compute $\sin^6(x)$. \begin{align*} \sin^6(x)&=\frac{1}{(2i)^6}(e^{ix}-e^{-ix})^6\\ &=\frac{-1}{2^6}(e^{6ix}-6e^{(5-1)ix}+15e^{(4-2)ix}-20e^{(3-3)ix}+15e^{(2-4)ix}-6e^{(1-5)ix}+e^{-6ix})\\ &=-\frac{1}{2^6}((e^{6ix}+e^{-6ix})-6(e^{4ix}+e^{-4ix})+15(e^{2ix}+e^{-2ix})-20)\\ &=-\frac{1}{2^6}\left(2\cos(6x)-12\cos(4x)+30\cos(2x)-20\right) \end{align*} In general you can derive formulas for $\sin^n(x)$ based on the parity of $n$ modulo $4$. Symmetry among the terms in the binomial expansion will cause terms to group together to form sines or cosines.

Edit (11/29/18): Having nothing better to do on a rainy day, I worked out the general formulae. Enjoy:

\begin{align*} \sin^{2m}(x)&=\left(\frac{e^{ix}-e^{-ix}}{2i}\right)^{2m}\\ &=\frac{(-1)^m}{2^{2m}}\sum_{k=0}^{2m}(-1)^k\binom{2m}{k}e^{ikx-i(2m-k)x}\\ &=\frac{(-1)^m}{2^{2m}}\sum_{k=0}^{2m}(-1)^k\binom{2m}{k}e^{i(2k-2m)x}\\ &=\frac{(-1)^m}{2^{2m}}\left[\sum_{k=0}^{m-1} (-1)^k\binom{2m}{k}e^{i(2k-2m)x} + (-1)^m\binom{2m}{m}+\sum_{k=m+1}^{2m}(-1)^k \binom{2m}{k}e^{i(2k-2m)x}\right]\\ &=\frac{(-1)^m}{2^{2m}}\left[\sum_{u=1}^{m} (-1)^{m-u}\binom{2m}{m-u}e^{-i(2u)x} +(-1)^m \binom{2m}{m}+\sum_{u=1}^{m} (-1)^{m+u}\binom{2m}{m+u}e^{i(2u)x}\right]\\ &=\frac{(-1)^m}{2^{2m}}\left[(-1)^m \binom{2m}{m}+\sum_{u=1}^{m} (-1)^{m-u}\binom{2m}{m-u}(e^{-i(2u)x}+e^{i(2u)x})\right]\\ &=\frac{1}{2^{2m}}\left[ \binom{2m}{m}+\sum_{u=1}^{m}(-1)^{u} \binom{2m}{m-u}2\cos(2ux)\right]\\ \end{align*} \begin{align*} \sin^{2m+1}(x)&=\left(\frac{e^{ix}-e^{-ix}}{2i}\right)^{2m+1}\\ &=\frac{(-1)^m}{2^{2m+1}i}\sum_{k=0}^{2m+1}(-1)^{k+1}\binom{2m+1}{k}e^{ikx-i(2m+1-k)x}\\ &=\frac{(-1)^m}{2^{2m+1}i}\sum_{k=0}^{2m+1}(-1)^{k+1}\binom{2m+1}{k}e^{i(2k-2m-1)x}\\ &=\frac{(-1)^m}{2^{2m+1}i}\left[\sum_{k=0}^{m}(-1)^{k+1}\binom{2m+1}{k}e^{i(2k-2m-1)x}+\sum_{k=m+1}^{2m+1}(-1)^{k+1}\binom{2m+1}{k}e^{i(2k-2m-1)x}\right]\\ &=\frac{(-1)^m}{2^{2m+1}i}\left[\sum_{u=0}^{m}(-1)^{m-u+1}\binom{2m+1}{m-u}e^{-i(2u+1)x}+\sum_{u=0}^{m}(-1)^{u+m}\binom{2m+1}{u+m+1}e^{i(2u+1)x}\right]\\ &=\frac{(-1)^m}{2^{2m}}\sum_{u=0}^{m}(-1)^{m-u}\binom{2m+1}{m-u}\left(\frac{e^{i(2u+1)x}-e^{-i(2u+1)x}}{2i}\right)\\ &=\frac{1}{2^{2m}}\sum_{u=0}^{m}(-1)^{u}\binom{2m+1}{m-u}\sin((2u+1)x) \end{align*}

  • Oh yeah, thanks I understand this one. It would be easy if the exercise was "Find a formula for $\sin(x)^n$. Sadly the exercise was, that I should proof that $sin(x)^n$ corresponds to the above forumla and with your approach I think it is possible, but with much more effort then "simply" proof it by induction. Because I have to do a lot of reshuffling, arguing and splitting for n odd and n even, etc... (I think so) – Gykonik Nov 27 '18 at 11:03
  • @Gykonik: I'm really not getting your objection. This approach gives you explicit formulae for the coefficients $a_k$ and $b_k$ which is even stronger than what your problem asked for. – Cheerful Parsnip Nov 27 '18 at 17:20
1

One may consider the two formulae

$$\begin{align}\sin(n)\cos(m)&=\frac12(\sin(n+m)+\sin(n-m))\\ \sin(n)\sin(m)&=-\frac12(\cos(n+m)-\cos(n-m))\end{align}$$

which are both a direct consequence of the addition theorems for the sine function and the cosine function respectively. In your case we set $n=x$ and $m=kx$ to get

$$\begin{align} \sin(x)\cos(kx)&=\frac12(\sin(x+kx)+\sin(x-kx))=\frac12(\sin((k+1)x)-\sin((k-1)x))\\ \sin(x)\sin(kx)&=-\frac12(\cos(x+kx)-\cos(x-kx))=-\frac12(\cos((k+1)x)-\cos((k-1)x)) \end{align}$$

As you can see this produces a term of the type $a_{k+1}\cos((k+1)x)+b_{k+1}\sin((k+1)x)$ which implies a higher upper border and therefore you are done hence this procedure can be applied for any value of $k$ and $n$ respectively.

Concerning the special terms $\sin(-x)$ and $\cos(-x)$ occuring for $k=0$ recall the symmetric properties of the trigonometric functions.


EDIT

Since I realised for myself - by extending my answer within the comments - actually writing down a proof using these given formulae seems to be more difficult than I thought in the first place.

Therefore lets plug in the new terms constructed by using the given formulae. So we get

$$S=\sum_{k=0}^n \left[\frac{a_k}2(\sin((k+1)x)-\sin((k-1)x))-\frac{b_k}2(\cos((k+1)x)-\cos((k-1)x))\right]$$

Hence we are dealing with finite sums we are perfectly fine with rearraging the sums aswell as splitting them up. So lets do a little bit of reshaping

$$\begin{align} S&=\sum_{k=0}^n \left[\frac{a_k}2(\sin((k+1)x)-\sin((k-1)x))-\frac{b_k}2(\cos((k+1)x)-\cos((k-1)x))\right]\\ &=\sum_{k=0}^n \left[a_k'\sin((k+1)x)-b_k'\cos((k+1)x)+b_k'\cos((k-1)x))-a_k'\sin((k-1)x)\right]\\ &=\sum_{k=0}^n \left[a_k'\sin((k+1)x)-b_k'\cos((k+1)x)\right]+\sum_{k=0}^n \left[b_k'\cos((k-1)x))-a_k'\sin((k-1)x)\right]\\ &=\underbrace{\sum_{k=1}^{n+1}\left[a_{k-1}'\sin(kx)-b_{k-1}'\cos(kx)\right]}_{=S_1}+\sum_{k=0}^n \left[b_k'\cos((k-1)x))-a_k'\sin((k-1)x)\right]\\ &=S_1+b_0'\cos(-x)-a_0'\sin(-x)+\sum_{k=1}^n \left[b_k'\cos((k-1)x))-a_k'\sin((k-1)x)\right]\\ &=S_1+b_0'\cos(x)+a_0'\sin(x)+\underbrace{\sum_{k=0}^{n-1} \left[b_{k+1}'\cos(kx)-a_{k+1}'\sin(kx)\right]}_{=S_2}\\ &=a_0'\sin(x)+b_0'\cos(x)+S_1+S_2 \end{align}$$

Now we can argue there are terms for $k=0$ produced by the sum $S_2$, terms for $k=n+1$ produced by $S_1$, the extra term which matches with the case $k=1$ for either $S_1$ or $S_2$ and terms inbetween with coefficients composed out of these given by $S_1$ and $S_2$. Therefore we can proceed to say that

$$S=\sum_{k=0}^{n+1}a_k''\cos(kx)+b_k''\sin(kx)$$

for some $a_k''$ and $b_k''$ which can be computed by oberserving the matching constants infront of the sines and cosines for the respective values of $k$ in each sum $S_1$ and $S_2$ and for $k=1$ the remaining term respectively.

mrtaurho
  • 16,103
  • @saulspatz Well since the induction hypotesis states that there are terms of the type $a_k\sin(kx)+b_k\cos(kx)$ for any $k=0,1,...,n$ there are indeed such terms with an $k-1$ as argument beforehand. Furthermore the $a_k$s and $b_k$s are some constants not explicitly choosen so the new terms with the argument $(k-1)x$ can be absorbed within the given ones by considering the new constants $a_{k-1}'$ and $b_{k-1}'$. – mrtaurho Nov 25 '18 at 18:14
  • You're right; I don't know what I was thinking. I must still be asleep. – saulspatz Nov 25 '18 at 18:19
  • @mrtaurho yeah I understand this approach. I also tried it with the addition theorems but I don't get your second last pararaph. Why this produces a term of the this type? And the part with "implies a higher upper border" I totaly understand but is there a way of writing this down mathematicly correct? – slkjck3 Nov 25 '18 at 19:25
  • Split up the sum in an part with the $(k-1)$ terms and in one with the $(k+1)$ and apply some suitable index shifts, so you will get sums which will add up in one sum with the bounds $0$ and $n+1$. – mrtaurho Nov 25 '18 at 19:40
  • Mhh I can't figure out the index shifts. I tried to shift the two sums by +1 so I end up with a sum from k to n+1, which is my desired result but this wasn't good. Then I tried to shift the sum with $(k-1)$ terms, so I have $sin((k+1)x$ / $cos$ in both sums but also there I couldnt figure out any further... Could you give me another hint? – slkjck3 Nov 25 '18 at 20:19
  • @slkjck3 I have added a complete solution to my answer which hopefully clears all doubts :) – mrtaurho Nov 25 '18 at 22:08
  • @mrtaurho oh yeah, now it's much clearer! I only don't understand the "argumentation" (second last pagaraph) because $S_1$ is a sum from 0 to n+1 and $S_2$ is a sum from 0 to n-1. This would mean that you need the missing $S_2$-sums for n and n+1 and I don't know how you can argue that they exists :D – slkjck3 Nov 25 '18 at 22:53
  • @slkjck3 Actually the sum $S_1$ is from $1$ up to $n+1$ and $S_2$ from $0$ up to $n-1$ respectively. Concerning your doubts regarding to the $n-$th and $n+1-$th term: I would argue that their coefficients - $a_{n-1}',b_{n-1}',a_n'$ and $b_n'$ (which can be found by oberseving the sum $S_1$ for $k=n$ and $k=n+1$) - are all non-zero since they are just defined as $a_{n-1}'=\frac{a_{n-1}}2$ - and so on. As long as we claim that the originial sum has the characteristic that for all $k$ we can say $a_k$ and $b_k$ are non-zero this will be a valid argumentation. – mrtaurho Nov 25 '18 at 23:05