24
$\begingroup$

\begin{eqnarray*} \sin(x) & = & x - \frac{x^3}{3!} + \frac{x^5}{5!} - \ldots\\ \sin(x+2π) & = & x + 2\pi - \frac{(x+2π)^3}{3!} + \frac{(x+2π)^5}{5!} - \ldots \\ \end{eqnarray*}

Those two series must be equal, but how can you show that by only manipulating the series?

$\endgroup$
  • 2
    $\begingroup$ You are missing a $2\pi$ on the right hand side of the second equation. $\endgroup$ – copper.hat Apr 14 '15 at 6:07
  • $\begingroup$ For the latter, at which point do you expand the function ? $\endgroup$ – Vim Apr 14 '15 at 6:07
  • $\begingroup$ not easy is it... $\endgroup$ – JonMark Perry Apr 14 '15 at 6:08
  • 1
    $\begingroup$ first try to show that $2\pi-(2\pi)^3/3!+(2\pi)^5/5! - ...=0$ $\endgroup$ – JonMark Perry Apr 14 '15 at 6:19
  • 3
    $\begingroup$ What are you using for a definition of $\pi$? $\endgroup$ – Robert Israel Apr 14 '15 at 6:44
27
$\begingroup$

A fairly easy way to introduce $\pi$ in trigonometric functions defined by series is:

  1. Define

$$e^z=\sum_{n=0}^{\infty} \frac{z^n}{n!}$$

Then use Euler's formula to define $\sin$ and $\cos$:

$$\sin z=\sum_{n=0}^\infty(-1)^n\frac{z^{2n+1}}{(2n+1)!}$$ $$\cos z=\sum_{n=0}^\infty(-1)^n\frac{z^{2n}}{(2n)!}$$

  1. Then prove that $\cos$ has a least positive root, which you call $\pi/2$. For this, you can show easily that $\cos 0>0$ and $\cos 2<0$ (the latter using majoration of the rest in the series, which is alternating).

  2. Prove and use $e^{a+b}=e^ae^b$ (it's a Cauchy product) to derive similar identities for $\sin$ and $\cos$.

  3. Use (2) and (3) to prove that $\sin$ and $\cos$ are $2\pi$-periodic.


Here is the detailed derivation

First, define

$$e^z=\sum_{n=0}^{\infty} \frac{z^n}{n!}$$

The series converges for all $z\in\Bbb C$ by the ratio test, thus it defines an entire function on the complex plane. It is $C^{\infty}$ on $\Bbb C$, and the restriction to real $z$ is real-valued and also $C^\infty$.

Putting $z=0$, you have $e^0=1$, and by differentiating the series, you get $\dfrac{\mathrm{d}e^z}{\mathrm{d}z}=e^z$.

Then you define

$$\cos z=\frac{e^{iz}+e^{-iz}}{2}=\sum_{n=0}^\infty(-1)^n\frac{z^{2n}}{(2n)!}$$ $$\sin z=\frac{e^{iz}-e^{-iz}}{2i}=\sum_{n=0}^\infty(-1)^n\frac{z^{2n+1}}{(2n+1)!}$$

And while we are at it,

$$\cosh z=\frac{e^{z}+e^{-z}}{2}=\sum_{n=0}^\infty\frac{z^{2n}}{(2n)!}$$ $$\sinh z=\frac{e^{z}-e^{-z}}{2}=\sum_{n=0}^\infty\frac{z^{2n+1}}{(2n+1)!}$$

Thus $\cos$ and $\cosh$ are even, while $\sin$ and $\sinh$ are odd.

Of course, the restriction of these functions to real $z$ are real-valued.

You have

$$e^{iz}=\cos z+i\sin z$$

And the derivatives $\sin'=\cos$ and $\cos'=-\sin$.

Notice also that the terms in the series of $e^x$ are increasing for increasing $x\geq0$, thus $x\rightarrow e^x$ is increasing for $x\geq0$ and you have $e^x\geq1+x$ for $x\geq0$, and $e^x\underset {x\rightarrow+\infty}\longrightarrow+\infty$.

Let $(a,b)\in\Bbb C^2$. Since the series of $e^z$ is absolutely convergent for all $z$, the following equality holds

$$e^ae^b=\sum_{i=0}^\infty \frac{a^i}{i!}\sum_{j=0}^\infty \frac{a^j}{j!}=\sum_{n=0}^{\infty} u_n$$

With

$$u_n=\sum_{k=0}^n\frac{a^kb^{n-k}}{k!(n-k)!}=\frac{1}{n!}\sum_{k=0}^n {n\choose k}a^kb^{n-k}=\frac{(a+b)^n}{n!}$$

Thus $e^ae^b=e^{a+b}$ for all complex $a,b$.

Thus you have $e^ze^{-z}=1$, and $e^z$ is never zero.


Digression on the real exponential

Hence for real $x$, $e^x\neq0$, and since the function is $C^0$ (even $C^\infty$), its sign does not change, and $\forall x\in\Bbb R, e^x>0$.

Also, since $e^xe^{-x}=1$ and $e^x\underset {x\rightarrow+\infty}\longrightarrow+\infty$, you have $e^x\underset {x\rightarrow-\infty}\longrightarrow0$.

And since the derivative of $e^x$ is itself, the derivative is also always positive, and the exponential is increasing on $\Bbb R$.

You can conclude it's a bijection, and since $e^ae^b=e^{a+b}$ and $e^0=1$, this proves that the exponential is a group isomorphism between $(\Bbb R,+)$ and $(\Bbb R^\star_+,\cdot)$.

Call $\log$ the inverse isomorphism, defined on $\Bbb R^\star_+$, with $\log (ab)=\log(a)+\log(b)$ for all $a>0, b>0$. Also, using the formula of derivation of an inverse function, you have $\log'(x)=1/x$.


Trigonometric identities

From $e^ae^b=e^{a+b}$ and using Euler's identity, you can derive the usual trigonometric (and hyperbolic trigonometry) identities. I'll show how on an example:

$$\cos a\cos b-\sin a\sin b=\frac{e^{ia}+e^{-ia}}{2}\frac{e^{ib}+e^{-ib}}{2}-\frac{e^{ia}-e^{-ia}}{2i}\frac{e^{ib}-e^{-ib}}{2i}$$ $$=\frac14\left[(e^{ia}+e^{-ia})(e^{ib}+e^{-ib})+(e^{ia}-e^{-ia})(e^{ib}-e^{-ib})\right]$$ $$=\frac{1}{4}\left[\left(e^{i(a+b)}+e^{i(a-b)}+e^{i(b-a)}+e^{-i(a+b)}\right)+\left(e^{i(a+b)}-e^{i(a-b)}-e^{i(b-a)}+e^{-i(a+b)}\right)\right]$$ $$=\frac{e^{i(a+b)}+e^{-i(a+b)}}{2}=\cos(a+b)$$

Likewise, you have $\sin(a+b)=\sin a\cos b+\sin b\cos a$, and a bunch of other formulas.

In particular, you have for all $z\in\Bbb C$:

$$\cos^2 z + \sin^2 z=\cos(z-z)=1$$ $$\cos 2z=\cos^2 z-\sin^2 z=2\cos^2 z-1$$

These are true for real $z$, and since the functions are then real-valued, you have $|\cos x|\leq 1$ and $|\sin x| \leq 1$ for all $x\in\Bbb R$.


Definition of $\pi$

You have $\cos 0=1$ from the series definition, and

$$\cos 2=\sum_{n=0}^\infty (-1)^n\frac{2^{2n}}{(2n)!}$$

The series is alternating with decreasing term after $n=1$, thus

$$\cos 2<1-\frac{2^2}{2!}+\frac{2^4}{4!} = -\frac13 <0$$

Since $\cos$ is continuous, it has at least one root in $]0,2[$.

The series for $\sin x$ is also alternating for $0< x\leq 2$, and its general term is decreasing after $n=0$, thus for $x\in[0,2]$,

$$\sin x \geq x-\frac{x^3}{6}=x\left(1-\frac{x^2}{6}\right)$$

The RHS of the inequality has roots $0$ and $\pm\sqrt{6}$, and $\sqrt{6}>2$, thus for $x\in]0,2]$, $\sin x>0$.

Since $\cos'=-\sin$, you have that the function $\cos$ is decreasing on $]0,2[$.

Therefore, $\cos x=0$ has one and only one root in $[0,2]$. Let's call this root $\frac{\pi}2$.

We have then $\cos \frac{\pi}2=0$, thus $\cos^2 \frac{\pi}2+ \sin^2 \frac{\pi}2=1$ implies $\sin \frac{\pi}2=\pm1$, and since it's positive, $\sin \frac{\pi}2=1$.

Also, from $\cos 2x=2\cos^2x-1$, you get that $\cos \pi=-1$, and then $\sin\pi=0$.

Notice that you have also

$$e^{i\pi}=\cos \pi+i\sin\pi=-1$$


Trigonometric functions are periodic

From the identities

$$\cos (a+b)=\cos a\cos b - \sin a \sin b$$ $$\sin (a+b)=\sin a\cos b + \cos a \sin b$$

You get

$$\cos (a+\pi)=\cos a\cos \pi - \sin a\sin \pi=-\cos a$$ $$\sin (a+\pi)=\sin a\sin \pi + \cos a\sin \pi=-\sin a$$

And finally

$$\cos (a+2\pi)=\cos a$$ $$\sin (a+2\pi)=\sin a$$

Thus $\cos$ and $\sin$ are $2\pi$-periodic. We have still to prove it's the smallest possible period, but before, let's have a look at variations of $\cos$ and $\sin$ on one period $[0,2\pi]$.

We already know that for $x\in[0,\pi/2]$, $\cos x\geq 0$ and $\sin x\geq 0$, where the former is decreasing from $1$ to $0$, and the latter is increasing from $0$ to $1$.

First, we complete an half-period. Using the previous identities:

$$\cos (\pi-x)=-\cos x$$ $$\sin (\pi-x)=\sin x$$

Thus for $x \in [\pi/2,\pi]$, $\cos$ is decreasing from $0$ to $-1$, and $\sin x$ is decreasing from $1$ to $0$.

Then we complete the full period with

$$\cos (a+\pi)=-\cos a$$ $$\sin (a+\pi)=-\sin a$$

This means that for $x\in[2\pi]$, the only roots of $\cos x$ are $\pi/2$ and $3\pi/2$, and the only roots of $\sin x$ are $0$, $\pi$ and $2\pi$.

Now, is $2\pi$ the smallest period? Suppose there is a $\lambda \in ]0,2\pi[$ such that for all $a$, $\cos (a+\lambda)=\cos a$, then

$$\cos(a+\lambda)=\cos a\cos\lambda-\sin a\sin \lambda=\cos a$$

And for $a=\pi/2$,

$$-\sin \lambda=0$$

Thus $\lambda=\pi$, but then $\cos a=\cos(a+\lambda)=-\cos a$, which is not true for example for $a=0$. Thus $2\pi$ is the minimal period.


What next?

You could define $\tan x=\frac{\sin x}{\cos x}$ and derive identities, then define inverse trigonometric functions on some wise restriction (since a periodic function has no inverse), and also define $a^b=e^{b\log a}$. And you have a construction of all so-called elementary functions.

$\endgroup$
  • $\begingroup$ An alternative way to get (3) that reveals the underlying reason for the series of the exponential function is that it is its own derivative and hence the Taylor series of $x \mapsto e^{a+x}$ immediately gives $e^{a+x} = e^a e^x$. Also, the geometric properties of $\exp,\sin,\cos$ is an alternative way to obtain periodicity, which is the approach I took at math.stackexchange.com/a/802678/21820. $\endgroup$ – user21820 Apr 14 '15 at 9:27
  • $\begingroup$ @JessePFrancis Fantastic Job! This is an excellent and concise tutorial - one of the best synopses on this material that I've seen. What was the inspiration? Was there a useful publication/text that was useful in compiling this? $\endgroup$ – Mark Viola Apr 14 '15 at 16:36
  • $\begingroup$ @Jean-ClaudeArbaut Well, great job! I have seen most of this a long, long time ago, but it was a treat to have this synopsis presented herein! $\endgroup$ – Mark Viola Apr 14 '15 at 16:58
9
$\begingroup$

With series manipulation, we can get $$ \begin{align} \sin(x+y) &=\sum_{k=0}^\infty(-1)^k\frac{(x+y)^{2k+1}}{(2k+1)!}\tag{1a}\\ &=\sum_{k=0}^\infty\sum_{j=0}^{2k+1}\frac{(-1)^k}{(2k+1)!}\binom{2k+1}{j}x^jy^{2k+1-j}\tag{1b}\\ &=\sum_{k=0}^\infty\sum_{j=0}^{2k+1}(-1)^k\frac{x^j}{j!}\frac{y^{2k+1-j}}{(2k+1-j)!}\tag{1c}\\ &=\sum_{k=0}^\infty\sum_{j=0}^k(-1)^j\frac{x^{2j+1}}{(2j+1)!}(-1)^{k-j}\frac{y^{2k-2j}}{(2k-2j)!}\tag{1d}\\ &+\sum_{k=0}^\infty\sum_{j=0}^k(-1)^j\frac{x^{2j}}{(2j)!}(-1)^{k-j}\frac{y^{2k-2j+1}}{(2k-2j+1)!}\tag{1e}\\ &=\sum_{j=0}^\infty(-1)^j\frac{x^{2j+1}}{(2j+1)!}\sum_{k=0}^\infty(-1)^{k}\frac{y^{2k}}{(2k)!}\tag{1f}\\ &+\sum_{j=0}^\infty(-1)^j\frac{x^{2j}}{(2j)!}\sum_{k=0}^\infty(-1)^k\frac{y^{2k+1}}{(2k+1)!}\tag{1g}\\[6pt] &=\sin(x)\cos(y)+\cos(x)\sin(y)\tag{1h} \end{align} $$ Explanation:
$\text{(1a)}$: series definition
$\text{(1b)}$: binomial theorem
$\text{(1c)}$: break binomial coefficient into factorials
$\text{(1d)}$: add the odd powers of $x$ from $\text{(1c)}$
$\text{(1e)}$: to the even powers of $x$ from $\text{(1c)}$
$\text{(1f)}$: change the order of summation and substitute $k\mapsto k+j$
$\text{(1g)}$: change the order of summation and substitute $k\mapsto k+j$
$\text{(1g)}$: series definitions


We can prove that $\cos(2\pi)=1$ and $\sin(2\pi)=0$ using a couple of derivatives and some geometry. First, $$ \begin{align} \frac{\mathrm{d}}{\mathrm{d}x}\sin(x) &=\frac{\mathrm{d}}{\mathrm{d}x}\sum_{k=0}^\infty(-1)^k\frac{x^{2k+1}}{(2k+1)!}\\ &=\sum_{k=0}^\infty(-1)^k\frac{x^{2k}}{(2k)!}\\[6pt] &=\cos(x)\tag{2} \end{align} $$ and $$ \begin{align} \frac{\mathrm{d}}{\mathrm{d}x}\cos(x) &=\frac{\mathrm{d}}{\mathrm{d}x}\left(1+\sum_{k=1}^\infty(-1)^k\frac{x^{2k}}{(2k)!}\right)\\ &=\sum_{k=1}^\infty(-1)^k\frac{x^{2k-1}}{(2k-1)!}\\ &=\sum_{k=0}^\infty(-1)^{k+1}\frac{x^{2k+1}}{(2k+1)!}\\[6pt] &=-\sin(x)\tag{3} \end{align} $$ Using $(2)$ and $(3)$, we get $$ \begin{align} \frac{\mathrm{d}}{\mathrm{d}x}\left(\sin^2(x)+\cos^2(x)\right) &=2\sin(x)\cos(x)-2\cos(x)\sin(x)\\ &=0\tag{4} \end{align} $$ Equation $(4)$ says that $\sin^2(x)+\cos^2(x)$ is a constant, and since $\sin(0)=0$ and $\cos(0)=1$, we have $$ \sin^2(x)+\cos^2(x)=1\tag{5} $$ Thus, $(\cos(x),\sin(x))$ is on the unit circle and moves with speed $$ \begin{align} \left|\frac{\mathrm{d}}{\mathrm{d}x}(\cos(x),\sin(x))\right| &=\left|\vphantom{\frac12}(-\sin(x),\cos(x))\right|\\[6pt] &=1\tag{6} \end{align} $$ To go around the unit circle once at speed $1$ takes $2\pi$. Thus, $$ \begin{align} (\cos(2\pi),\sin(2\pi)) &=(\cos(0),\sin(0))\\[6pt] &=(1,0)\tag{7} \end{align} $$


Combining $(1)$ and $(7)$ yields $$ \begin{align} \sin(x+2\pi) &=\sin(x)\cos(2\pi)+\cos(x)\sin(2\pi)\\[6pt] &=\sin(x)\tag{8} \end{align} $$

$\endgroup$
  • $\begingroup$ Well done. This is what I planned to write formally, but never got around to :) $\endgroup$ – nbubis Apr 14 '15 at 16:44
8
$\begingroup$

Note that $\sin$ is the unique solution of $y'' = -y$ subject to $y(0) = 0, y'(0) = 1$. Note that $x \mapsto \sin (x+2 \pi)$ also satisfies the same differential equation, so the question boils down to showing that $\sin (2 \pi) = 0, \cos(2 \pi) = 1$ (where $\cos = \sin'$).

Consider $\eta(x) = \sin^2 x + \cos^2 x$, we see $\eta(0) = 1$ and $\eta(x)'= 0$, hence $\sin^2 x + \cos^2 x =1$ for all $x$, and so we only need to establish that $\cos (2 \pi) = 1$.

It is straightforward to establish that there is a smallest $T>0$ such that $ \cos T = 0$. Rudin has a nice method in "Real & Complex Analysis". Alternatively, suppose $\cos x >0$ for all $x \ge 0$. We have $\cos 0 = 1$, and so we have some $x_0>0$ such that $\sin x_0 >0$. By presumption, $\sin $ is strictly increasing on $x \ge 0$, and since $\cos' x = - \sin x$, we have $\cos' x \le - \sin x_0$ for all $x \ge x_0$, which contradicts $\cos x >0$ for all $x \ge 0$. Since $\cos$ is continuous, there is a smallest positive number $T$ such that $\cos T = 0$, and we can see that $\sin T = 1$, since $\sin$ is increasing on $[0,T)$.

We define $\pi = 2 T$.

Let $\zeta(x) = \cos ( {\pi \over 2}-x)$. We note that $\zeta(0) = 0, \zeta'(0) = 1$ and $\zeta'' = - \zeta$, and so by uniqueness we have $\sin x= \cos ({\pi \over 2}-x) = \cos (x-{\pi \over 2})$ for all $x$ (since $\cos $ is even), and differentiating gives $\cos x = -\sin (x-{\pi \over 2})$ (since $\cos' = -\sin$). It is straightforward to establish that $\cos (2 \pi) = 1$ from this.

Addendum: To show that the solution of $y''+y = 0$ is unique we could use the Picard Lindelöf theorem. Alternatively, suppose $y_1,y_2$ are two solutions with the same initial conditions and let $\delta = y_1-y_2$. Let $V = \delta^2 + (\delta')^2$ and note that $V' = 0$. Since $V(0) = 0$, we have $V(x) = 0$ for all $x$, that is, $y_1 = y_2$.

$\endgroup$
  • $\begingroup$ What is your justification that the differential equation has no other solution with those initial conditions? Using the complex exponential function takes just one line, but I don't see how to do it easily without. $\endgroup$ – user21820 Apr 14 '15 at 13:07
  • $\begingroup$ @user21820: I added an elaboration above. $\endgroup$ – copper.hat Apr 14 '15 at 17:25
  • $\begingroup$ Your alternative solution using $δ^2-δ'^2$ is very very nice! I was looking for such a method but couldn't find it. $\endgroup$ – user21820 Apr 16 '15 at 4:19
2
$\begingroup$

Define $f(x)$ as the residual: $$\sin(x+2\pi) = x + 2\pi - \frac{(x+2\pi)^3}{3!} + \frac{(x+2\pi)^5}{5!}+\ldots$$ $$\equiv x - \frac{x^3}{3!} + \frac{x^5}{5!}+\ldots + f(x)=\sin(x)+f(x)$$ Now prove it is zero: $$f(x) = 2\pi -\frac{1}{3!}\left((2\pi)^3 + 3(2\pi)^2 x + 3(2\pi)x^2 \right)+ \frac{1}{5!}\left( (2\pi)^5 +5x (2\pi)^4+10x^2(2\pi)^3+10x^3(2\pi)^2+5(2\pi)x^4\right)+\ldots$$ Extract all powers of $x$ lower than $2$, and define yet another residual, $g(x)$: $$f(x)=\sin(2\pi) + x\sum_{k=1}^\infty \frac{(-1)^k(2\pi)^{2k}}{(2k)!}+g(x)$$ $$=\sin(2\pi) +x(1-\cos(2\pi))+g(x)$$ The first terms are zero, and now we must prove that the remaining residual $g(x)$ is also zero, ad infinitum.

A more general continuation to follow...

$\endgroup$
0
$\begingroup$

Using term-wise derivation, we define the auxiliary sum $$e^{ix}=\sin'(x)+i\sin(x)=\sum_{k=0}^\infty\frac{(ix)^k}{k!}.$$ By a straightforward manipulation of this sum using the Binomial theorem,

$$e^{i(x+t)}=\sum_{k=0}^\infty\frac{(i(x+t))^k}{k!}=\sum_{k=0}^\infty\sum_{l=0}^k\frac{k!i^kx^lt^{k-l}}{k!l!(k-l)!}=\sum_{l=0}^\infty\sum_{j=0}^\infty\frac{(ix)^l(it)^{j}}{l!j!}=e^{ix}e^{it}.$$

Then we evaluate $$\sin(1)=1-\frac1{3!}+\text{ positive pairs of terms }\cdots>0$$ $$\sin(4)=4-\frac{4^3}{3!}+\frac{4^5}{5!}-\frac{4^7}{7!}+\frac{4^9}{9!}-\text{ positive pairs of terms }\cdots<0,$$ to show that $\sin$ has a root in $(1,4)$, let $\pi$.

As $|e^{i\pi}|^2=e^{i\pi}\left(e^{i\pi}\right)^*=e^{i\pi}e^{-i\pi}=e^0=1$, this root is such that $$e^{i\pi}=\pm1,$$ and to avoid the possible negative, we use $$e^{i2\pi}=\left(e^{i\pi}\right)^2=1.$$Thus there is a real $\pi\in(1,4)$ such that $$\sin(x+2\pi)=\Im\left(e^{ix+i2\pi}\right)=\Im\left(e^{ix}\right)=\sin(x).$$

$\endgroup$

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.