My friend and I were working to see if we could use the squeeze theorem to prove that the derivative of $e^x$ is $e^x$.

We said that by definition, the derivative is

$$\frac{d}{dx} e^x =\lim \limits_{h\to0} \frac{e^{x+h}-e^x}{h} = \lim \limits_{h\to0} \frac{e^{x}(e^h-1)}{h} = \lim \limits_{h\to0} e^x *\lim \limits_{h\to0} \frac{(e^h-1)}{h}$$

$$\frac{1}{2}h+1 \leq \frac{(e^h-1)}{h} \leq |h| + e^h$$

$$\lim \limits_{h\to0} \frac{1}{2}h+1 = \lim \limits_{h\to0} |h| + e^h = 1$$

$$\lim \limits_{h\to0} e^x *\lim \limits_{h\to0} \frac{(e^h-1)}{h} = \lim \limits_{h\to0} e^x = e^x$$

However, this seems more simplified than any other proof we've seen, so we're wondering if there's a fault here. Neither of us have taken analysis yet, so we might be assuming something incorrect.

  • $\begingroup$ Where did $\frac{1}{2}h+1 \leq \frac{(e^h-1)}{h} \leq |h| + e^h$ come from? $\endgroup$ – Henry May 7 '18 at 0:22
  • $\begingroup$ @Henry ah sorry I should've specified, we wanted to bound above and below and we came up with those functions to do so. $\endgroup$ – rb612 May 7 '18 at 0:22
  • $\begingroup$ What is your definition of $e$? $\endgroup$ – Jeffery Opoku-Mensah May 7 '18 at 0:23
  • $\begingroup$ @JefferyOpoku-Mensah I suppose it's the traditional definition of $\lim \limits_{x\to \infty} (1+1/x)^x$. $\endgroup$ – rb612 May 7 '18 at 0:24
  • $\begingroup$ A complete proof would need to show why these are bounds $\endgroup$ – Henry May 7 '18 at 0:25

A pretty nice way to do this is as follows:

Note that $$|e^h - 1 - h| = \left|\frac{h^2}{2} + \frac{h^3}{6} + \frac{h^4}{24} +...\right| $$ $$ = h^2 \left|\frac{1}{2} + \frac{h}{6} + \frac{h^2}{24} +...\right|\leq h^2 \left(\frac{1}{2} + \frac{1}{4} + \frac{1}{8} + ... \right) = h^2$$

So if $h\neq 0$, we have $\left|\frac{e^h - 1}{h} - 1 \right| \leq |h|$, i.e. $$\lim_{h \rightarrow 0} \left(\frac{e^h - 1}{h} - 1\right) = 0$$

Note that this tells us that $e^x$ is differentiable at the origin with derivative 1.

Now: $$\lim_{h \rightarrow 0} \left(\frac{e^{x+h} - e^x}{h}\right) = e^x \lim_{h \rightarrow 0} \frac{e^h - 1}{h} = e^x$$

as required.

  • $\begingroup$ Thank you, except doesn’t this require knowing the derivative of $e^x$ in order to create this Maclaurin expansion? Unless I suppose you define $e$ to be that series. $\endgroup$ – rb612 May 7 '18 at 0:34
  • $\begingroup$ Not if you define $e^x$ by the power series. Even if you use the sequence definition as you wrote in the comments, you can easily show this is equivalent to the series representation of $e^x$ $\endgroup$ – PhysicsMathsLove May 7 '18 at 0:35
  • $\begingroup$ I could show that it's equal to $\lim \limits_{x\to\infty} (1+\frac{1}{x})^x$ easily? $\endgroup$ – rb612 May 7 '18 at 0:39
  • $\begingroup$ Yes. It is a standard argument in elementary real analysis. For example, see: math.stackexchange.com/questions/637255/… $\endgroup$ – PhysicsMathsLove May 7 '18 at 0:48

Here is a pure squeeze theorem proof without using a Taylor expansion and using the definition $e = \lim_{n \to \infty}(1 +1/n)^n$.

For $h > 0$, let $n = \lfloor1/h\rfloor$. We have $n \leqslant 1/h < n+1$ and

$$\frac{n}{n+1}(n+1)(e^{1/n+1} - 1)= n(e^{1/n+1} - 1 ) \leqslant \frac{e^h -1 }{h} \leqslant (n+1)(e^{1/n} -1) = \frac{n+1}{n}n(e^{1/n}-1).$$

Since $n \to \infty$ iff $h \to 0$ we obtain the limit $1$ by the squeeze theorem if we can show that $n(e^{1/n} - 1) \to 1$.

This follows from the inequality

$$\left(1 + \frac{1}{n} \right)^n < e < \left(1 + \frac{1}{n} \right)^{n+1},$$

which implies

$$1 < n(e^{1/n}-1) < n\left[\left(1 + \frac{1}{n} \right)^{1/n}\left( 1+ \frac{1}{n}\right)-1 \right] \leqslant 1 + \frac{1}{n} + \frac{1}{n^2}.$$

Here we used Bernoullis's inequality $(1 + 1/n)^{1/n} < 1 + 1/n^2$ to obtain the far right inequality. A second application of the squeeze theorem gives us $n(e^{1/n}-1) \to 1$.

If $h <0$ and the limit is approached from the left we can use the above result and

$$\frac{e^h -1}{h} = \frac{e^{-|h|}-1}{-|h|} = e^{-|h|}\frac{e^{|h|}-1}{|h|},$$

since $e^{-|h|} \to 1$ as $h \to 0$.


Claim 1: $\lim\limits_{x\to 0} \dfrac{\ln(x+1)}{x}=1$

Proof: by properties of log, $\lim\limits_{x\to 0} \dfrac{\ln(x+1)}{x}=\lim\limits_{x\to 0} \ln[(x+1)^{1/x}]=\ln(e)=1$.

Claim 2: $\lim\limits_{y\to0}\dfrac{e^y-1}{y}=1$

Proof: putting $x=e^y-1$ then $\lim\limits_{y\to0}\dfrac{e^y-1}{y}=\lim\limits_{x\to0}\dfrac{x}{\ln(x+1)}=\lim\limits_{x\to0}\dfrac{1}{\dfrac{\ln(x+1)}{x}}=\dfrac{1}{1}=1$

With this, $$\lim\limits_{h\to0}\dfrac{e^{x+h}-e^x}{h}=e^x$$


How did you get these inequalities ?

$$\frac{1}{2}h+1 \leq \frac{(e^h-1)}{h} \leq |h| + e^h$$

You need to prove these inequalities to have a valid proof.

You can not use Taylor Series because your statement is used in Taylor series.

  • $\begingroup$ Mohammad, thank you for your answer. I understand why we can't use the Taylor Series. I don't see exactly where in our proof we use the series expansion. $\endgroup$ – rb612 May 7 '18 at 0:41
  • $\begingroup$ My point is that in proving the inequalities you can not use the Taylor series. $\endgroup$ – Mohammad Riazi-Kermani May 7 '18 at 0:54

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.