$h\in \mathbb{R}$, because we have defined the Trigonometric Functions only on $\mathbb{R}$ so far.

I have a look at $e^{ih}=\sum_{k=0}^{\infty}\frac{(ih)^k}{k!}=1+ih-\frac{h^2}{2}+....$

How can one describe the nth term of the sum?

Then I look at $\frac{e^{ih}-1}{h}=\frac{(1-1)}{h}+i-\frac{h}{2}+...=i-\frac{h}{2}+....$

Again how can I describe that the nth term of the sum?

Because $\frac{e^{ih}-1}{h}=\sum_{k=1}^{\infty}\frac{\frac{(ih)^k}{h}}{k!}<\sum_{k=0}^{\infty}\frac{\frac{(ih)^k}{h}}{k!}=\sum_{k=1}^{\infty}\frac{(ih^{-1})^k}{k!}=e^{ih^{-1}}$

and $ih^{-1}$ is a complex number and the exponential-series converges absolutely for all Elements in $\mathbb{C}$, I have found a convergent majorant. And I can apply the properties of Limits on $\frac{e^{ih}-1}{h}\forall, h\in \mathbb{R}$.

How can I now prove formally (i.e by chosing an explicit $\delta$) that

$$\forall_{\epsilon>0}\exists_{\delta>0}\forall_{h\in\mathbb{R}}|h-0|=|h|<\delta\Longrightarrow |(\frac{e^{ih}-1}{h}=i-\frac{h}{2}+...)-i|<\epsilon$$

I am also seeking for advice how to argue in such cases more intuitively (i.e by not always ginving an explicit $\delta$ ).

  • 2
    $\begingroup$ (1) for the very first one, you already have an expression for the $k$th term, otherwise how can you use sigma notation for a sum? (2) for the bit before you ask for a formal proof, you can't use inequalities when you are trying to compare complex numbers in $\mathbb C$ $\endgroup$ – Calvin Khor Jan 23 at 13:42
  • $\begingroup$ @CalvinKhor On (2), If I would put $|\cdot |$ around the sums would that work? $\endgroup$ – RM777 Jan 23 at 13:47
  • 1
    $\begingroup$ You should check absolute convergence, because this is something that extends to complex numbers nicely (and has $|\cdot|$s everywhere) $\endgroup$ – Calvin Khor Jan 23 at 13:49

Sketch that follows the spirit of your approach, rather than using trigonometry etc. A useful technical tool:

Theorem. Suppose the continuous functions $f_n=f_n(h)$ taking values in $\mathbb C$ are such that $\sum_{n=0}^\infty f_n(h) := \lim_{N\to\infty} \sum_{n=0}^N f_n(h)$ converges absolutely uniformly on some interval $h\in [-a,a]$ (i.e. $\sum_{n=0}^\infty \|f_n\|_{\infty} < \infty $). Then $$ \lim_{h\to 0} \sum_{n=0}^\infty f_n(h) = \sum_{n=0}^\infty f_n(0)$$

Sketch proof of theorem: Recall that the uniform limit of continuous functions is continuous. (see Proof of uniform limit of Continuous Functions or Wikipedia ). The $N$th partial sums $F_N(h) := \sum_{n=0}^N f_n(h)$ are continuous, and they converge pointwise on $[-a,a]$ to $F(h):=\sum_{n=0}^\infty f_n(h)$, and the absolutely uniformly convergent assumption implies that this convergence is uniform on $[-a,a]$. Hence, $F$ is continuous at all points in $[-a,a]$, and in particular at $0$. The result follows.

Application: $$\frac{e^{ih}-1}{h}\\= \frac{(\sum_{n=0}^\infty (ih)^n/n! )- 1} {h} \\= \frac{\sum_{n=1}^\infty (ih)^n/n! } {h} \\= \sum_{n=1}^\infty \frac{i(ih)^{n-1}}{n!} \\= \sum_{m=0}^\infty \frac{i(ih)^{m}}{(m+1)!} $$ After you verify we can apply the theorem, this yields $\lim_{h\to 0 } \frac{e^{ih}-1}{h} = i+0+0+\dots = i$.

PS make sure you realise that all infinite sums are defined in terms of limits in $\mathbb C$ (e.g. $a_n \to a \in\mathbb C$ iff $|a_n - a| \to 0$)

PPS this works to find the derivative at an arbitrary point of any function expressed as (for example) a Taylor series.

  • $\begingroup$ Nice answer, Calvin. I was wondering if you had a source or proof of the theorem? $\endgroup$ – Jam Jan 23 at 14:15
  • $\begingroup$ @Jam it essentially follows from the fact that the uniform limit of continuous functions is continuous (and the sequence of partial sums is a sequence of continuous functions that converge uniformly under the above assumptions). Does this help? $\endgroup$ – Calvin Khor Jan 23 at 14:16
  • $\begingroup$ It does. Thank you $\endgroup$ – Jam Jan 23 at 14:17
  • $\begingroup$ @Jam you're welcome $\endgroup$ – Calvin Khor Jan 23 at 14:21
  • 1
    $\begingroup$ @RM777 I have added the proof to the answer, as in the comment to Jam $\endgroup$ – Calvin Khor Jan 23 at 22:11

Hint: Use Euler's formula and split the limit into well known trigonometric limits.


Here is a simple proof which I first read in Hardy's A Course of Pure Mathematics.

To reiterate the question for clarity we have $h\in\mathbb{R} $ and the symbol $e^{ih} $ is defined by the series $\sum_{n=0}^{\infty} (ih) ^n/n! $. Let's assume $|h|< 1$ and observe that \begin{align*} \left|\frac {e^{ih} - 1}{h}-i\right|&=\left|\frac{i^2h}{2!}+\frac{i^3h^2}{3!}+\dots\right|\\ &\leq\frac{|h|}{2!}+\frac{|h|^2}{3!}+\dots\\ &\leq\frac{|h|} {2}+\frac{|h|^2}{2^2}+\frac{|h|^3}{2^3}+\dots\\ &=\frac{|h|}{2-|h|}\text{ (sum of infinite GP)} \\ &<|h| \end{align*} Thus given any $\epsilon>0$ if we choose $\delta=\min(1,\epsilon) $ then we have $$\left|\frac{e^{ih} - 1}{h}-i\right|<\epsilon $$ whenever $0<|h|<\delta$. Therefore by definition of limit we have $$\lim_{h\to 0}\dfrac{e^{ih}-1}{h}=i$$

Hardy uses the above limit to prove the formula $$e^{ix} =\cos x+i\sin x \,\forall x\in\mathbb {R} $$ The idea is to show that the function $f(x) =e^{ix} $ is differentiable with derivative $f'(x) =if(x) $ and then consider the function $$g(x) =(\cos x-i\sin x) f(x) $$ It is easily proved that $g'(x) =0$ and hence $g$ is constant. Thus $g(x) =g(0)=1$ and $f(x) =\cos x+i\sin x$.


Show that $\lim_{h\rightarrow 0}\frac{e^{ih}-1}{h}=i$:

As we know: $ \cos \theta + i \sin \theta = e^{i \theta}$

So, $$\lim \limits_{h\rightarrow 0} [\frac { \cos h + i \sin h - 1 } {h} ] = \lim \limits_{h \rightarrow 0} [ \frac {1 - \frac{h^2}{2!} + \phi_1 (h^4) + i( h - \frac{h^3}{3!}... \phi_2(h^5) )-1} {h} ] = \lim \limits_{h \rightarrow0} \frac { i(1 - higher \space powers\space of\space h) }{1} = i$$


$$e^{ih} = 1 + ih + \frac{-h^2}{2!}...$$ will give the same result, just take terms of $i$ to one side and none $i$ ones to other, then take $i$ common and cancel 1 from numerator, then take $h$ common and cancel it from denominator, you'll get the desired result!


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.