3

Does anyone have proof that $\lim_{n\to \infty }\left(1+\frac{z}{n}\right)^n=e^x(\cos y+i\sin y)$ because the proof I have is unclear and/or incorrect.

$$z=x+iy \in \mathbb C$$

For a large enough $n$: $\left|\arg\left(1+\frac{1}{n}\right)\right|<\frac{\pi}{2}$ (Why/how?) Lets look at the real sequences $z_n=\left(1+\frac{z}{n}\right)^n,|z_n| \text{ and } \arg z_n.$

It is proven that $\lim_{n\to \infty }|1+\frac{z}{n}|^n=e^x$

then it says, which makes no sense to me from here on(I'll highlight what lines are unclear):

$$\arg (z_n)=\arg\left(\left(1+\frac{z}{n}\right)^n\right)=$$

$$n\arctan\left(\frac{y}{x+n}\right) \text{ then it's just some blank space and } \mod2\pi \to y \text{ and again some blank space} \mod 2\pi \text{ when } n\to \infty, \text{therefore }\lim_{n\to \infty }\left(1+\frac{z}{n}\right)^n=e^x(\cos y+i\sin y) $$

Kamil Jarosz
  • 4,984
  • 1
    Maybe helpful: http://math.stackexchange.com/questions/374747/if-z-n-to-z-then-1z-n-nn-to-ez –  Oct 12 '15 at 17:51
  • I'm not quite sure what you're asking here, but the reason that $|\arg(1 + z/n)| < \frac{\pi}{2}$ is that $1 + z/n \to 1$ as $n\to\infty$ (for fixed $z$), and $\arg(1) = 0 < \pi/2$. (There's a bit of ambiguity here in the definition of $\arg$ and what it means for the given inequality to hold, but the point is that $1 + w$ stays close to $1$ and away from the imaginary axis for $w$ small.) – anomaly Oct 12 '15 at 18:33
  • See this beautiful answer http://math.stackexchange.com/a/1451245/72031 – Paramanand Singh Oct 13 '15 at 03:32
  • also see http://math.stackexchange.com/a/1668179/72031 – Paramanand Singh Feb 25 '16 at 03:34

3 Answers3

3

I've always thought the binomial theorem and dominated convergence theorem are the way to go to prove this. For $n=1,2,\dots $ define

$$c_n(0) = 1, c_n(k) = \frac{n(n-1)\cdots (n-k+1)}{n^k},\, k = 1,2,\dots n, \,c_n(k) = 0\, \text {for}\,\, k> n.$$

Then the binomial theorem gives $$\tag 1(1+z/n)^n = \sum_{k=0}^{\infty}c_n(k)\frac{z^k}{k!}.$$ For each fixed $k, |c_n(k)|\le 1$ for all $n,$ and $\lim_{n\to \infty}c_n(k) =1.$ Because $\sum |z|^k/k!<\infty$ for each $z,$ the limit of $(1)$ as $n\to \infty$ equals $\sum_{k=0}^{\infty}1\cdot\frac{z^k}{k!} = e^z$ by the DCT.

zhw.
  • 105,693
  • This is a very efficient way forward, so well done! But does it address the OP's quite specific question(s)? – Mark Viola Oct 15 '15 at 14:37
  • It answers the question "does anyone know a proof". But you're right, I didn't address the specific problems the OP was having with that particular approach. – zhw. Oct 15 '15 at 15:56
  • 1
    Yes, it does answer the headline. Don't you love when a headline mismatches the issue in the ensuing text? ;-)) – Mark Viola Oct 15 '15 at 16:18
2

We know that principal value of the argument function of a complex number $z$ is defined as

$$\text{Arg} (z)=\begin{cases} \arctan\left(\frac{\text{Im}(z)}{\text{Re}(z)}\right)&,\text{Re}(z)>0\\\\ \arctan\left(\frac{\text{Im}(z)}{\text{Re}(z)}\right)+\pi&,\text{Re}(z)<0,\,\text{Im}(z)\ge 0\\\\ \arctan\left(\frac{\text{Im}(z)}{\text{Re}(z)}\right)-\pi&,\text{Re}(z)<0,\,\text{Im}(z)< 0\\\\ \pi/2&,\text{Re}(z)=0,\,\text{Im}(z)> 0\\\\ -\pi/2&,\text{Re}(z)=0,\,\text{Im}(z)< 0\\\\ \end{cases}$$

The argument of $z$ is

$$\arg (z)=\text{Arg}(z)+2\ell \pi$$

for integer values of $\ell$.

Now, for $z=x+iy$ with fixed values of $x$ and $y$ and $n>-x$, we have

$$\begin{align} \arg\left(\left(1+\frac zn\right)^n\right)&=\arg\left(\left(1+\frac xn+i\frac yn\right)^n\right) \tag 1\\\\ &=n\arctan\left(\frac{y/n}{1+x/n}\right)+2\ell \pi \tag 2\\\\ &\to y+2\ell \pi \,\,\text{as}\,\,n\to \infty \tag 3 \end{align}$$

Therefore, since $\cos (y+2\ell \pi)=\cos y$ and $\sin (y+2\ell \pi)=\sin y$, we have

$$\lim_{n\to \infty}\left(1+\frac zn\right)^n=e^x\left(\cos y+i\sin y\right)$$

as was to be shown!


NOTE $1$:

In going from $(1)$ to $(2)$, we made use inductively of the relationship

$$\arg(z^2)=2\arg(z)+2\ell \pi \tag 4$$

We can verify $(4)$ as follow. First, let $z=x+iy$. Then, for $x^2-y^2>0$, and $x>0$, we have

$$\begin{align} \arg(z^2)&=\text{Arg}(z^2)+2\ell \pi\\\\ &=\arctan\left(\frac{2xy}{x^2-y^2}\right)+2\pi \ell\\\\ &=\arctan\left(\frac{2(y/x)}{1-(y/x)^2}\right)+2\pi \ell\\\\ &=2\arctan(y/x)+2\pi \ell\,\,\dots \,\text{where we used the double angle formula for the tangent}\\\\ &=2\arg(z)+2\ell \pi \tag 5 \end{align}$$

Now, let $z=\left(1+\frac xn\right)+i\left(\frac yn\right)$ in $(5)$. Then, for $n$ sufficiently large, $\left(1+\frac xn\right)^2-\left(\frac yn\right)^2>0$ and $1+\frac xn >0$, we find

$$\arg\left(z^2\right)=2\arg(z)+2\pi \ell$$

Proceeding inductively reveals that

$$\arg(z^n)=n\arg (z)+2\ell \pi$$

as was to be shown.


NOTE $2$:

In going from $(2)$ to $(3)$, we made use of THIS ANSWER, in which I established the inequalities for the arctangent function

$$\frac{|x|}{\sqrt{1+x^2}} \le|\arctan(x)|\le |x|$$

Then, we have

$$\left|\frac{\frac{y}{1+x/n}}{\sqrt{1+\left(\frac{y/n}{1+x/n}\right)^2}} \right|\le \left| n\arctan\left(\frac{y/n}{1+x/n}\right)\right| \le \left|\frac{y}{1+x/n}\right|$$

whereupon applying the squeeze theorem reveals the limit

$$\lim_{n\to \infty}n\arctan\left(\frac{y/n}{1+x/n}\right)=y$$

as was to be shown!

Mark Viola
  • 179,405
  • can you please explain in greater detail the step where you went from $arg \left( \left( 1 + \frac{x}{n} + i\frac{y}{n}\right)^{n}\right)$ to $n \arctan \left(\frac{y/n}{1+x/n}\right) + 2\mathcal{l}\pi$? I'm working on this same problem, and I understand that $arg\left(1+ \frac{x}{n} + i\frac{y}{n}\right) = \arctan\left(\frac{y/n}{1+x/n}\right) = 2\mathcal{l}\pi$, but I don't understand where the $n$ out front came from. Nor do I understand why the $n$ out front doesn't make the limit blow up to $\infty$. Can you please explain this in greater detail? Thanks :) –  Feb 24 '16 at 08:25
  • @JessyCat Jessy. Do you understand why $\arg(z_1,z_2)=\arg(z_1)+\arg(z_2)$? If so, then inductively $\arg (nz)=n\arg(z)$. And can you show from the inequalities $x\cos(x)\le \sin(x)\le x$ for $\pi/2\ge x\ge 0$ that the arctangent function satisfies the inequalities $x/\sqrt{1+x^2} \le \arctan(x)\le x$ for $0\le x$? And from those bounds on the arctangent, can you see why the limit in question is indeed correct? - Mark – Mark Viola Feb 24 '16 at 16:44
  • Bozo. Please let me know how I can improve my answer. I really just want to give you the best answer I can. - Mark – Mark Viola Feb 24 '16 at 16:45
  • actually, I cannot show the second set of inequalities from the first set. I was given the hint for this problem to show that $\displaystyle \lim_{n \to \infty}\left(\frac{\arctan\left(\frac{y/n}{1+(x/n)} \right)}{\frac{y/n}{1+(x/n)}} \right) = 1$ in order to show that we can replace $\arctan$ with its argument as $n \to \infty$. If I could do that, then, I could simply bring the $n$ inside the argument, but I would get $y/x$ as my limit. I'm still very confused... –  Feb 24 '16 at 17:54
  • er...rather that the $\lim_{n \to \infty} \left( \text{that whole thing} \right)^{n/2} = 1$. –  Feb 24 '16 at 18:13
  • if you could help me figure this out, I'd be eternally grateful! –  Feb 24 '16 at 18:45
  • 1
    @JessyCat Jessy. I've added two detailed notes. Let me know if this helps. - Mark – Mark Viola Feb 24 '16 at 19:02
  • it did!! Now I understand everything! Thank you so much for being so patient with me and answering my follow-up questions! –  Feb 24 '16 at 22:21
  • @jessycat Jessy. You're welcome. My pleasure. - Mark – Mark Viola Feb 24 '16 at 22:36
  • I have a problem understanding step 2). Is ℓ independent of the chosen n? I think it should be $ℓ_n$. – user1868607 Apr 15 '16 at 10:06
  • @Rodrigo $\ell$ is any integer value and does not depend, therefore, on $x/n$ or $y/n$. – Mark Viola Apr 15 '16 at 13:28
0

In my opinion this could be answered as follows:

Lemma: Let $z_n$ be a sequence of non-zero complex numbers and suppose that for each $n \in \mathbb{N}$ we have $\phi_n \in Arg(z_n)$ where $Arg(z_n)$ is the set of arguments (I usually denote by arg(z) the principal value). Assume that $\phi_n \to \phi$ and that $|z_n| \to \rho$ then we have $z_n \to \rho(cos \phi+isen\phi)$.

So once you've proved this lemma you can work as follows:

$lim |1+\frac{z}{n}|^n= lim \sqrt{\left( (1+\frac{x}{n})^2 + (\frac{y}{n})^2\right)^n}= lim \sqrt{\left(1+\frac{x^2/n+y^2/n+2x}{n}\right)^n}=(e^{2x})^{1/2}=e^x$

where I've used the real case of the limit and I've written $z=x+iy$.

Now by De Moivre's formula $n \ arg(1+\frac{z}{n}) \in Arg((1+\frac{z}{n})^n)$ and because of the previous lemma the only thing I need to proof is that $n \ arg(1+\frac{z}{n}) \to y$.

You should remember at this point the value of the argument expressed in a general way which you can find here.

Using that form we have:

$n \ arg(1+\frac{z}{n}) = 2n \ arctg\left(\frac{y/n}{\sqrt{(1+x/n)^2+(y/n)^2}+(1+x/n)}\right)$ (1)

Note that this form could be wrong if $1+\frac{z}{n}$ was a negative real number but clearly there exists $n_0$ such that for all $n \ge n_0$ $1+\frac{z}{n}$ is not a negative real number.

Finally, if you apply that $arctg x$ and $x$ are equivalent infinitesimals when $x \to 0$ then in (1) you can put:

$2n \left(\frac{y/n}{\sqrt{(1+x/n)^2+(y/n)^2}+(1+x/n)}\right)$

which clearly converges to y.

If you apply the first lemma you are done.

user1868607
  • 5,791