1

I want to look at the summation for both cos,sin..

$cos(x) = 1-\frac{x^2}{2!}+\frac{x^4}{4!}-...= \sum\nolimits_{i=0}^\infty \frac{(-1)^nx^{2n}}{(2n)!}$ so, $cos(x+\frac{\pi}{2})=\sum\nolimits_{i=0}^\infty \frac{(-1)^n(x+\frac{\pi}{2})^{2n}}{(2n)!}$. So how I need to expand $(x+\frac{\pi}{2})^{2n}$ using the binomial formula maybe?

I want to look at manipulating the POWER series specifically for proving this identity.

Mr.Fry
  • 5,003
  • 3
  • 19
  • 28
  • 4
    Taylor series is certainly the hard way to prove the result in the title. – Gerry Myerson Sep 14 '13 at 23:43
  • 1
    I depends on your definitions. – Pedro Sep 14 '13 at 23:44
  • This will help me build some intuition on how to show, $\cos(x+y)=cos(x)cos(y)-sin(x)sin(y)$. Specifically by manipulating the power series. – Mr.Fry Sep 14 '13 at 23:46
  • 1
    Do you know (in the sense that you allow yourself to use the fact) that $\cos \frac{\pi}{2} = 0$ and $\sin \frac{\pi}{2} = 1$? – Daniel Fischer Sep 14 '13 at 23:46
  • Actually, using the power series, I think the general addition theorem is easier to show than this specific instance. – Daniel Fischer Sep 14 '13 at 23:47
  • As an exercise and tour de force, showing this with power series can be nice (for some...I guess...). If you want to go highschoolish use the good' ol trigonometric circle and the basic definitions there. – DonAntonio Sep 15 '13 at 11:15

4 Answers4

3

$$ \begin{align} \cos\left(x+\frac\pi2\right) &=\sum_{k=0}^\infty\frac{(-1)^k}{(2k)!}\left(x+\frac\pi2\right)^{2k}\\ &=\sum_{k=0}^\infty\sum_{j=0}^{2k}\frac{(-1)^k}{(2k)!}\binom{2k}{j}x^j(\pi/2)^{2k-j}\\ &=\sum_{j=0}^\infty\sum_{k=\lceil j/2\rceil}^\infty\frac{(-1)^k}{(2k)!}\binom{2k}{j}x^j(\pi/2)^{2k-j}\\ &=\sum_{j=0}^\infty\sum_{k=\lceil j/2\rceil}^\infty\frac{(-1)^k}{(2k-j)!j!}x^j(\pi/2)^{2k-j}\\ &=\sum_{j=0}^\infty\frac{x^j}{j!}\sum_{k=\lceil j/2\rceil}^\infty\frac{(-1)^k}{(2k-j)!}(\pi/2)^{2k-j}\tag{1} \end{align} $$ Notice that when $j$ is even, $$ \begin{align} \sum_{k=\lceil j/2\rceil}^\infty\frac{(-1)^k}{(2k-j)!}(\pi/2)^{2k-j} &=\sum_{k=0}^\infty\frac{(-1)^{k+j/2}}{(2k)!}(\pi/2)^{2k}\\ &=(-1)^{j/2}\cos(\pi/2)\\[12pt] &=0\tag{2} \end{align} $$ Notice that when $j$ is odd, $$ \begin{align} \sum_{k=\lceil j/2\rceil}^\infty\frac{(-1)^k}{(2k-j)!}(\pi/2)^{2k-j} &=\sum_{k=0}^\infty\frac{(-1)^{k+(j+1)/2}}{(2k+1)!}(\pi/2)^{2k+1}\\ &=(-1)^{(j+1)/2}\sin(\pi/2)\\[12pt] &=(-1)^{(j+1)/2}\tag{3} \end{align} $$ Plugging $(2)$ and $(3)$ back into $(1)$, removing the even terms and reindexing the odd terms, yields $$ \begin{align} \cos\left(x+\frac\pi2\right) &=-\sum_{j=0}^\infty\frac{(-1)^j}{(2j+1)!}x^{2j+1}\\ &=-\sin(x)\tag{4} \end{align} $$

robjohn
  • 345,667
2

What you sketch there is definitely the hard way to do this. When you expand the binomials, each of them will contribute terms to every term of lower degree in the resulting series, and then you have to sum all of these to find the coefficients.

It's much easier (not to mention more instructive) to follow the programs outlined in answers to How to prove periodicity of sin(x) or cos(x) starting from the Taylor series expansion?.

2

As an example, one can use the statement that

PROP If $f(0)=f'(0)=0$ and $f''+f=0$, then $f=0$.

P The ODE gives $f'f''+ff'=0$, so that $(f')^2+f^2=K$. But then $K=0$ from the initial conditions. This gives that $f=0$.

COR If $f(0)=a,f'(0)=b$ and $f''+f=0$ then $f=a\cos+b\sin$.

P Set $g=a\cos +b\sin $. Then $h=f-g$. It follows $h''+h=0$ and $h(0)=0,h'(0)=0$ so $h\equiv 0$ by the lemma.

(Giveaway: let $f(x)=\cos (x+y)-\cos x\cos y+\sin x\sin y$ in the first result, or $f(x)=\cos (x+y)$ in the corrollary)

Then you can prove that $\cos(x+y)=\cos x\cos y-\sin x\sin y$ using the above, letting $y$ be a constant.

Alternatively, let $f=\sin x+\cos \left(x+\dfrac \pi 2\right)$ above. You will need to use the special values $\sin \dfrac \pi 2=1$ and $\cos \dfrac\pi 2=0$, though. One can in fact define $\dfrac \pi 2$ to be the unique number determined by $$\inf\{x>0:\cos x=0\}$$

Pedro
  • 122,002
  • @robjohn One can prove it with the above, that's the point of it. To compute the derivatives is usual business, what do you mean? – Pedro Sep 15 '13 at 00:38
  • Okay. I agree. It looks as if you are assuming too much, but since the series gives you $\sin'(x)=\cos(x)$ and $\cos'(x)=-\sin(x)$, you only need to know $\sin(\pi/2)=1$ and $\cos(\pi/2)=0$ to get the rest. – robjohn Sep 15 '13 at 00:57
  • In effect, this uses the power series, but I'm not sure I would classify it as manipulating the power series. +1 for the concise (if a bit terse) presentation. – robjohn Sep 15 '13 at 01:01
  • @robjohn Heh, thanks. Terse? What do you mean? =O – Pedro Sep 15 '13 at 01:02
  • As a proof, it shows everything it needs to. As an explanation, it is a bit short. Just an opinion. – robjohn Sep 15 '13 at 01:17
2

Start from $$\cos(x+\pi/2) = \sum_{n\ge 0} (-1)^n \frac{(x+\pi/2)^{2n}}{(2n)!}.$$ Now ask yourself about the coefficient of $x^k$ in this series, it is given by $$\sum_{2n\ge k} \frac{(-1)^n}{(2n)!} \binom{2n}{k} (\pi/2)^{2n-k} = \frac{1}{k!} \sum_{2n\ge k} \frac{(-1)^n}{(2n-k)!} (\pi/2)^{2n-k}.$$ In this last series the $2n-k$ term starts at zero when $k$ is even and is given by $$\frac{1}{k!} \sum_{q\ge 0} \frac{(-1)^{q+k/2}}{(2q)!} (\pi/2)^{2q} = \frac{(-1)^{k/2}}{k!} \sum_{q\ge 0} \frac{(-1)^q}{(2q)!} (\pi/2)^{2q} = \frac{(-1)^{k/2}}{k!} \cos(\pi/2) = 0.$$ When $k$ is odd the $2n-k$ term starts at one and is given by $$\frac{1}{k!} \sum_{q\ge 0} \frac{(-1)^{q+(k+1)/2}}{(2q+1)!} (\pi/2)^{2q+1} \\= \frac{(-1)^{(k+1)/2}}{k!} \sum_{q\ge 0} \frac{(-1)^q}{(2q+1)!} (\pi/2)^{2q+1} = \frac{(-1)^{(k+1)/2}}{k!} \sin(\pi/2) = \frac{(-1)^{(k+1)/2}}{k!}.$$ It follows that $$\cos(x+\pi/2) = \sum_{m\ge 0} \frac{(-1)^{m+1}}{(2m+1)!} x^{2m+1} = -\sin(x).$$

Marko Riedel
  • 61,317
  • Congratulations to @robjohn who was faster than me by half a minute. – Marko Riedel Sep 15 '13 at 00:17
  • I get it, I did see it half a minute before I posted but the date stamp says five minutes. ;-) Tough luck for me I guess. – Marko Riedel Sep 15 '13 at 00:23
  • I wouldn't worry. I don't think any of these answers will get a windfall of votes :-) Not that they are not good, it's just that some topics attract a lot more voters. – robjohn Sep 15 '13 at 00:26