4

One definition of $e$ that I am fond of is that it is the number $a$ such that $$ \lim_{h\to 0}\frac{a^{h} -1}{h}=1 $$ The reason for this is that it cuts to the heart of the special property of all exponential functions. If we have $f(x)=a^x$, then \begin{align} f'(x)&=\lim_{\Delta x \to 0}\frac{a^{x+\Delta x} -a^x}{\Delta x} \\ &=\lim_{\Delta x \to 0}a^x\frac{a^{\Delta x} - 1}{\Delta x} \\ &=a^x\lim_{\Delta x \to 0}\frac{a^{\Delta x} - 1}{\Delta x} \end{align} However, presumably there is a caveat to this approach. We need to show that a number like $e$ exists in the first place! In other words, we need to show that $g(a)=\lim_{\Delta x \to 0}\frac{a^{\Delta x} -1}{\Delta x}$ takes the value of $1$ for some value of $a$. How might we do this?

Joe
  • 14,185
  • 2
  • 28
  • 65
  • 3
    Maybe try to show that $g(a) = \lim_{h\to 0}\frac{a^{h} -1}{h}=1$ is continuous, and there exist $a,b$ with $g(a)<1, g(b)>1$. – GEdgar Aug 31 '20 at 21:45
  • 4
    You could define the exponential function by its taylor series $\exp(x) := \sum_{n=0}^\infty x^n/n!$, show convergence, that its derivative is the exponential function itself, and its property $\exp(a+b) = \exp(a) \exp(b)$. Therefore for $e := \exp(1)$, we have $\exp(x) = e^x$. How would you even define $a^x$ without the exponential function in the first place? – Xileflix Aug 31 '20 at 22:05
  • 1
    @Xileflix I know it is common to define exponentiation with $e$, but I have also heard of the following approach: for positive integers, $a^x = a \cdot a \cdot a \dots a \text{($x$ times)}$, then define $a^{p/q}=\sqrt[q]{a^p}$ for rational numbers, and finally define $a^x$ using limits for irrational numbers. In this [post](https://math.stackexchange.com/questions/55068/can-you-raise-a-number-to-an-irrational-exponent/55078#55078), Andre Nicolas briefly outlines this approach. – Joe Aug 31 '20 at 22:14
  • @Joe: That is a correct way to define the exponential function for $a > 1$. Then it may be extended for $0 < a \leqslant 1$, shown to be continuous, etc. – RRL Aug 31 '20 at 22:58
  • Well, this begs the question as to the actual definition of an exponent function. But you can show that for $0 < a < b$ then $\lim \frac {a^h -1}h < \frac {b^h-1}h$ and that a function $f(k) = \lim \frac {k^h -1}h$ is continuous and that $f(2) < 1$ and $f(3) > 1$. ***BUT*** this is all begging the question as $a^x$ hasn't actually been defined yet. – fleablood Sep 01 '20 at 03:53
  • 1
    You may want to visit [this question](https://math.stackexchange.com/q/925454/72031). – Paramanand Singh Sep 01 '20 at 16:55
  • You can proceed by showing that $g(a) $ is defined for all $a>0$, it is continuous and strictly increasing function of $a$ and maps $(0,\infty) $ onto $\mathbb{R} $. And more importantly it satisfies $g(ab) =g(a) +g(b) $ and forms one of the most important elementary function. – Paramanand Singh Sep 01 '20 at 16:59
  • I have discussed existence of $g(a) $ in [this answer](https://math.stackexchange.com/a/1491888/72031). – Paramanand Singh Sep 01 '20 at 17:05

3 Answers3

4

Independently, we can prove that there exists a number $e$ such that $e = \lim_{n \to \infty} (1+1/n)^n$ along with the property that for all $n \in \mathbb{N}$,

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

It follows that

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

where the right-hand inequality is obtained using Bernoulli's inequality $(1 + 1/n)^{1/n} \leqslant 1 + 1/n^2$.

By the squeeze theorem we get

$$\tag{*}\lim_{n \to \infty}\frac{e^{1/n}-1}{\frac{1}{n}} = 1$$

From here it is not difficult to show that

$$\tag{**} \lim_{h \to 0+}\frac{e^{h}-1}{h} = 1$$

Taking $n = \lfloor1/h\rfloor$ when $h > 0$, 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 $n \to \infty$ if and only if $h \to 0$ we obtain (**) by the squeeze theorem using the previous result (*).

With some more work we can show that the limit $1$ is also attained as $h \to 0-$.

RRL
  • 84,288
  • 7
  • 55
  • 114
1

Let $f(x)$ be given by the

$$f(x)=\int_1^x \frac1t\,dt$$

It is easy to show that $f(x)$ is continuous and increasing for $x>0$ with $f(1)=0$ and $\lim_{x\to\infty}f(x)=\infty$. Then, by the intermediate value theorem, there exists a number $a>1$ such that

$$f(a)=\int_1^a \frac1t\,dt=1\tag1$$


We enforce the substitution $t\mapsto (1+ht)^{1/h}$ in $(1)$ to obtain

$$1=\int_0^{(a^h-1)/h}\frac1{1+ht}\,dt\tag2$$


For $h>0$, $\frac{a^h-1}{h}>0$ and $1\le 1+ht\le a^h$ when $t\in [0,(a^h-1)/h]$. Therefore, from $(2)$ we find that

$$\left(\frac{a^h-1}{h}\right)a^{-h}\le 1\le \left(\frac{a^h-1}{h}\right)\tag3$$

Rearranging $(3)$, we have the bounds

$$1\le \left(\frac{a^h-1}{h}\right)\le a^h$$

whence application of the squeeze theorem yields

$$\lim_{h\to 0^+}\left(\frac{a^h-1}{h}\right)=1$$

We leave it as an exercise for the reader to show that the left-side limit is also $1$ from which we conclude

$$\lim_{h\to 0}\left(\frac{a^h-1}{h}\right)=1$$

for some $a>1$ such that $1=\int_1^a \frac1t\,dt$.

Mark Viola
  • 166,174
  • 12
  • 128
  • 227
  • 1
    @joe Please let me know how I can improve my answer. I really want to give you the best answer I can. – Mark Viola Sep 01 '20 at 15:38
  • Thanks for contacting me. I'm having trouble understanding how you got from equation $1$ to equation $2$. In particular, I'm not sure how you converted the bounds of the integral, and also how your substitution works. – Joe Sep 01 '20 at 15:44
  • 1
    Hi Joe. You're welcome. My pleasure. In going from $(1)$ to $(2)$, we made the substitution $t= (1+hu)^{1/h}$. Then, $dt= \frac1h \times (1+hu)^{1/h -1}\times h\,du=(1+hu)^{1/h-1}\,du$. Then, $\frac1t \,dt =\frac1{1+hu}\,du$. The limits of integration transform also. When $t=1$ , $u=0$. And when $t=a$, $u=\frac{a^h-1}{h}$. Finally, $u$ is a dummy variable so after we make the substitution, we can relabel it as $t$. – Mark Viola Sep 01 '20 at 16:07
  • 1
    @joe Hi Joe. Did my previous comment answer your questions? – Mark Viola Sep 03 '20 at 17:27
  • Yes it did. Sorry for the late reply—I was busy preparing for my return to school. Since this answer seems quite elegant, I am going to accept it. – Joe Sep 03 '20 at 18:12
  • 1
    Hi Joe. Much appreciated. And best of luck in your returning to classes. Stay safe and healthy! – Mark Viola Sep 03 '20 at 18:13
0

Let me suggest one more, if it can be so called, functional, approach. Let's consider function $f$ with following 3 property:

  1. $f(x_1+x_2)=f(x_1)f(x_2)$ for $\forall x_1,x_2 \in \mathbb{R}$
  2. $f(0)=1,f(1)=a$ where $a>0$
  3. $f$ is continuous in $x=0$

It can be proved, that exists unique continuous on all $\mathbb{R}$ function, which satisfies brought properties and it we define as $a^x$.

Can be proved, that exists limit $$\lim\limits_{x \to \infty}\left( 1+\frac{1}{x} \right)^x= \lim\limits_{x \to 0}\left( 1+ x\right)^{\frac{1}{x}}$$ and we denote it by $e$. For $e$, at last, we obtain $$\lim\limits_{x \to 0}\frac{e^x-1}{x}=1$$ If this approach is satisfactory and you are interested in the details, then write which part you would like to see more fully written.

zkutch
  • 11,079
  • 2
  • 11
  • 26