32
$\begingroup$

Well I am just asking myself if there's a more elegant way of proving $$2<\exp(1)=\mathrm e<3$$ than doing it by induction and using the fact of $\lim\limits_{n\rightarrow\infty}\left(1+\frac1n\right)^n=\mathrm e$, is there one (or some) alternative way(s)?

$\endgroup$
1
  • 6
    $\begingroup$ Is this last thing your definition of $e$? $\endgroup$ Jul 3 '12 at 21:54
70
$\begingroup$

What answer you find most elegant may depend on what definition of $e$ you're starting with, as Dylan suggests, but I find this argument quite short and sweet: $$\begin{align} &\quad 1 + 1 &= 2\\ &< 1 + 1 + \frac12 + \frac1{2\cdot3} + \frac1{2\cdot3\cdot4} + \cdots &= e \\ &< 1 + 1 + \frac12 + \frac1{2\cdot2} + \frac1{2\cdot2\cdot2} + \cdots &= 3 \end{align}$$

$\endgroup$
5
  • 8
    $\begingroup$ And from this you see that it's also at least 2.5. $\endgroup$ Jul 3 '12 at 22:27
  • $\begingroup$ This looks like the Taylor expansion of $\exp$ at 0 evaluated at 1. Or which definition of $e$ did you use here? $\endgroup$
    – krlmlr
    Jul 3 '12 at 22:51
  • 6
    $\begingroup$ @user946850, I consider $e = \sum_{n=0}^\infty 1/n!$ itself to be one of the many equivalent definitions of $e$. $\endgroup$
    – user856
    Jul 3 '12 at 22:53
  • $\begingroup$ I don't understand the "$1 + 1 + \frac12 + \frac1{2\cdot2} + \frac1{2\cdot2\cdot2} + \cdots = 3$" part. Is there a formula to get it? $\endgroup$
    – yaobin
    May 18 '19 at 1:36
  • 3
    $\begingroup$ @yaobin Yes, there is. This sum is $1 + \sum_{k = 0}^{\infty} 2^{-k}$ which is a geometric series with $r = \frac{1}{2}$. Therefore, $$\sum_{k = 0}^{\infty} 2^{-k} = \frac{1}{1 - \frac{1}{2}} = 2.$$ $\endgroup$
    – Ramanujan
    Jun 3 '19 at 18:00
18
$\begingroup$

You can use integration by parts to show:

$$\int_1^e (\ln x)^2 dx = e-2$$

$$\int_1^e (\ln x )^3 dx = 6- 2e$$

Since $\ln(x)$ is strictly positive above $1$, we get

$$e-2>0$$ $$6-2e>0$$

so that $2<e<3$.

$\endgroup$
2
  • 4
    $\begingroup$ By the way, this argument left alone might look a bit ad-hoc, but it generalizes: integrating $\int_1^e (\ln x)^n dx$ by parts shows that $\frac{n!}{!n}$ alternates around $e$. This sequence actually converges extremely fast, so you can use it pretty easily to compute arbitrary precision rational approximations to $e$. $\endgroup$ Oct 20 '18 at 21:42
  • 1
    $\begingroup$ Actually, $\int_1^e (\ln x)^n dx = a_n e + b_n$, where $(-1)^n a_n$ is oeis.org/A000166 and $(-1)^{n+1} b_n=n!$. $\endgroup$
    – lhf
    Jan 10 '20 at 10:07
11
$\begingroup$

You can use

$$e =\sum_{n=0}^\infty \frac{1}{n!}= 2+\sum_{n=2}^\infty \frac{1}{n!}< 2+\sum_{n=2}^\infty \frac{1}{n(n-1)}=3 \,,$$

with the last equality following immediately from the fact that $\sum_{n=2}^\infty \frac{1}{n(n-1)}$ is telescopic.

Of course it depends on the way you define $e$, anyhow the equality

$$\sum_{n=0}^\infty \frac{1}{n!}=\lim\limits_{n\rightarrow\infty}(1+\frac1n)^n$$ can be established easily using the binomial theorem.


Second solution

You can use the fact that $a_n=(1+\frac{1}{n})^{n+1}$ is decreasing. The inequality $a_{n+1} < a_n$ is an immediate consequence of Bernoulli Inequality.

Note that this implies (induction hidden here) that $a_n \leq a_6 <3$ for all $n \geq 3$, and that

$$e =\lim a_n \leq a_6 <3 \,.$$


Here is one more:

$$e^{-1}=1-\frac{1}{1!}+\frac{1}{2!}-\frac{1}{3!}+.. \,.$$

Since the series is alternating and $\frac{1}{n!}$ is decreasing, it is obvious (very easy to show) that the series oscilates around the limit and

$$s_{2n+1}=1-\frac{1}{1!}+\frac{1}{2!}-\frac{1}{3!}+....+\frac{1}{(2n)!}-\frac{1}{(2n+1)!} \leq \frac{1}{e} \leq 1-\frac{1}{1!}+\frac{1}{2!}-\frac{1}{3!}+....+\frac{1}{(2n)!}=s_{2n}$$

[ Actually in the proof of the Alternating series test, one proves the stronger statement that for such a series we have $s_{2n}$ decreasing, $s_{2n+1}$ increasing and $s_{2n+1} \leq s_{2n}$. ]

The inequality

$$1-\frac{1}{1!}+\frac{1}{2!}-\frac{1}{3!} < \frac{1}{e} < 1-\frac{1}{1!}+\frac{1}{2!}$$ is

$$\frac{1}{3} < \frac{1}{e} < \frac{1}{2} \,.$$

$\endgroup$
2
  • 1
    $\begingroup$ It was not clear to me how to easily show using binomial theorem that both definitions of $e$ are equal. Fortunately, Wikipedia had the answer at en.wikipedia.org/wiki/Binomial_theorem#Series_for_e $\endgroup$ May 16 '13 at 6:15
  • 2
    $\begingroup$ "Note that this implies (induction hidden here) that $a_n\le a_6<3$ for all $n\ge 3$" — I guess you mean $n\ge 6$ here. $\endgroup$
    – celtschk
    Jul 23 '16 at 6:52
9
$\begingroup$

It's equivalent to show that the natural logarithm of 3 is bigger than 1, but this is $$ \int_1^3 \frac{dx}{x}. $$ A right hand sum is guaranteed to underestimate this integral, so you just need to take a right hand sum with enough rectangles to get a value larger than 1.

$\endgroup$
4
  • $\begingroup$ (to show that $e>2$, integrate from 1 to 2 and show you get less than 1 by estimating with a left hand sum). $\endgroup$
    – user29743
    Jul 3 '12 at 22:09
  • $\begingroup$ You can also use the fact that $\int_1^3 \frac{1}{x^a} >1$ when $a$ is close enough to $1$. I think it holds for $a=1.1$, but that is not easy to show.... $\endgroup$
    – N. S.
    Jul 3 '12 at 22:35
  • 3
    $\begingroup$ Replacing the integrand $1/t$ by its tangent at $t = (x+1)/2$ shows that $\log(x) = \int_1^x \tfrac{dt}{t} \geq 2 \tfrac{x-1}{x+1}$ for all $x \geq 1$. In particular $\log(3) \geq 1$. $\endgroup$
    – WimC
    Jul 8 '12 at 17:46
  • 1
    $\begingroup$ @WimC, wow, graphing software shows that is an outstanding estimate between .5 and 1.5, had not seen that one before. That's really cool. $\endgroup$
    – user29743
    Jul 14 '12 at 15:00
1
$\begingroup$

First let's consider a simple heuristic argument to show that $2<e<4$. It is easy to prove using the definition of the derivative that if $f(x)=2^x$ then $f'(x) = (\text{constant}\cdot 2^x$). The curve gets steeper as $x$ increases, and the average slope between $x=0$ an $x=1$ is $(2^1-2^0)/(1-0)= 1$. Therefore, the slope at $x=0$ is less than $1$; hence the "constant" is less than $1$. Now do the same with $g(x)=4^x$ on the interval from $x=-1/2$ and $x=0$, and conclude that the slope at $x=0$ is more than $1$; hence the "constant" you get there is more than $1$.

So $2$ is too small, and $4$ is too big, to serve as the base of the natural exponential function.

It's messier to do the same with $3$, but the interval from $x=-1/6$ to $x=0$ will do it, and you conclude $3$ is too big to be the base of the natural exponential function.

$\endgroup$
1
$\begingroup$

The sequence $x_n=(1+1/n)^{n+1}$ is strictly decreasing and converges to $e.$ So $e<x_5=2.985984 <3.$

Remark: $\ln x_n=-(n+1)\ln (1-1/(n+1))=1+\sum_{j=1}^{\infty}(j+1)^{-1}(n+1)^{-j}.$ Comparing the terms of this series to the corresponding terms in the series for $\ln x_{n+1} , $ we see that $\ln x_n>\ln x_{n+1} , $ so $ x_n>x_{n+1}.$

$\endgroup$
0
$\begingroup$

Observe that, $(1 + \frac 1 n)^n = 1 + 1 + (1 - \frac 1 n)/2! + (1 - \frac 1 n)(1 - \frac 2 n)/3! + ... + (1 - \frac 1 n)(1 - \frac 2 n)...(1 - (n-1)/n)/n! < 1 + 1 + 1/2! + ... + 1/n! < 1 + 1 + 1/2 + 1/2^2 + ... + 1/2^{n-1}$, [since $n! > 2^{n-1}, \forall_n \geq 3$]. So, $(1 + \frac 1 n)^n < 3 - (\frac 1 2)^n$.Now as $n$ is tending to infinity $(\frac 1 2)^n$ diminishes indefinitely and ultimately we obtain $e < 3$.

$\endgroup$

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.