7

Without using differentiation, logarithmic function, rigorously, prove that $$e^x\ge x+1$$ for all real values of $x$.

8 Answers8

28

Bernoulli's Inequality: for any $\,n\in\Bbb N\,$

$$1+x\leq\left(1+\frac{x}{n}\right)^n\xrightarrow [n\to\infty]{} e^x$$

The inequality above is true for $\,x\geq -1\,$ , and since the wanted inequality is trivial for $\,x<-1\,$ we're done.

Joe
  • 22,603
DonAntonio
  • 214,715
  • 1
    Showing Bernoulli's inequality for $-1 < x < 0$ does require some effort though... – Zarrax Dec 06 '12 at 21:06
  • @DonAntonio: Nice. Might be clearer if said inequality is trivial for $x\lt -1$. – André Nicolas Dec 06 '12 at 21:11
  • 2
    @AndréNicolas, that was the intention yet some of my fingers believe they have free will...Thanks. – DonAntonio Dec 06 '12 at 21:53
  • 1
    @Zarrax, a very minimal effort: the inequality follows by induction in the general case. – DonAntonio Dec 06 '12 at 21:53
  • It is quite easy to demonstrate that Bernoull’s inequality is strict for $x\neq 0$ and $n>1$, but the limit introduces it again. Can we extend this to show that the (question) inequality is strict iff $x\neq 0$? – Steve Powell Sep 25 '13 at 14:26
  • it's wrong. This is not Bernoulli inequality – frhack Feb 10 '18 at 21:00
  • @frhack Of ourse it is. If you were more careful you would check the above thoroughly before downvoting a correct answer. – DonAntonio Feb 10 '18 at 22:15
  • Bernoulli's Inequality is different (1+nx)<=(1+x)^n and not directly useful to demostrate the target inequality. So if you have the arguments you have to show them given that the question ask to prove it. – frhack Feb 11 '18 at 06:54
  • 1
    @frhack That's exactly what is written there with $;\cfrac xn;$ instead of $;x;$ alone. Do you Really have problems to see this, or you just wanted to downvote something that is correct? – DonAntonio Feb 11 '18 at 09:11
  • 1
    ok damn me, you are right ! sorry ! – frhack Feb 11 '18 at 13:22
  • 1
    @frhack Don't worry, it happens to us all some time. It is advisable, though, to wait a little more for downvoting even if there's an outstanding mistake in the answer. Sometimes typos go under the radar or someone gets confused. Better, to comment and wait for the answerer to address the doubt. – DonAntonio Feb 11 '18 at 13:35
  • @DonAntonio i waited a few hours before downvoting. :( Peraphs i was a little bit confused by this demostration. https://jeremykun.com/2015/11/23/the-inequality/ – frhack Feb 11 '18 at 13:41
9

I don't know if it's cheating, but you didn't say that integration is forbidden. Since $\exp$ is increasing we know that $e^{-x} \leq 1$ for all $x \geq 0$. Integrating, we get $$ \forall x\geq 0,\qquad 0 = \int_0^x 0\,dt \leq \int_0^x (1-e^{-t})\,dt = x + e^{-x} - 1. $$

The inequality $\forall x \geq 0, \;e^x \geq 1 + x$ follows in the same lines as the former.

Siméon
  • 10,844
6

Write out the expansion of $(1 + {x \over n})^n$ as $$(1 + {x \over n})^n = 1 + n {x \over n} + {n(n-1) \over 2}{x^2 \over n^2} + ...$$ If $x \geq 0$, then all terms are nonnegative, so the sum is at least the sum of the first two terms, namely $1 + x$. If $x \leq - 1$, then for even $n$, the expression $(1 + {x \over n})^n $ is nonnegative, so the limit must be at least zero, which is greater than or equal to $1 + x$.

So it remains to look at the case where $-1 < x < 0$. In this case write $x = -y$ and you have $$(1 - {y \over n})^n = 1 - n {y \over n} + {n(n-1) \over 2}{y^2 \over n^2} - ...$$ This is an (finite) alternating series, and the absolute value of the ratio of two consecutive terms of this series is of the form ${n - k \over k+ 1} {y \over n}$, which is less than $1$ since $0 < y < 1$. So the terms decrease in absolute value. Hence the overall sum is at least what you get if you truncate after a negative term. So truncating after two terms you get $$(1 - {y \over n})^n \geq 1 - y$$ Equivalently, $$(1 + {x \over n})^n \geq 1 + x$$ Taking limits as $n$ goes to infinity gives what you're looking for.

Zarrax
  • 46,591
4

If $ x\ge 0 $ as $ \exp $ is increase we have $ e^{x} \ge 1 $. Then by mean value theorem there exist $ c \in (0,x) $ such that \begin{equation} e^x - 1 = e^c x. \end{equation} Thus \begin{equation} e^x \ge x +1. \end{equation} If $ x \le -1 $, we have \begin{equation} e^x \ge 0 > 1+x \end{equation} Now if $ -1 < x < 0 $ we have \begin{equation} e^c < 1 \Rightarrow -e^c > -1 \Rightarrow -e^c x < -x \end{equation} Thus \begin{equation} 1 - e^x = e^c (-x) < -x \end{equation} and \begin{equation} e^x > 1+x \end{equation}

user29999
  • 4,721
2

Hint:

Prove for rational $x=\frac{a}{b}$ ($b > 0$) that $e^{\frac{a}{b}} \ge \frac{a}{b}+1$. You can do this by showing that $$e^a \ge \left(\frac{a}{b}+1\right)^b.$$ Finally, argue by continuity.

Zach
  • 246
2

Assuming $x > 0$ and regardless of your definition of $e$, you can show that $e^x = \displaystyle\lim_{n \to \infty} \left( 1 + \frac{x}{n} \right)^n$. Then:

\begin{align} \left( 1 + \frac{x}{n} \right)^n &= \sum_{k=0}^n {n \choose k}\frac{x^k}{n^k} \\ &= 1 + x + \sum_{k=2}^n {n \choose k} \frac{x^k}{n^k} \end{align}

Since each term ${n \choose k} (x/n)^k$ is positive, it follows that

$$ \left( 1 + \frac{x}{n} \right)^n - x - 1 > 0. $$ Hence,

$$ \lim_{n \to \infty} \left( 1 + \frac{x}{n} \right)^n - x - 1 \geq 0. $$ from which the desired conclusion follows for $x > 0$.

JavaMan
  • 13,375
1

Suppose $x \le 0$. From $t^n - 1 = (t-1)(1 + t + \ldots + t^{n-1})$ (with $t = 1 + x/n$) we get $$(1+x/n)^n - 1 = \frac{x}{n} \sum_{j=0}^{n-1} (1+x/n)^j$$ If $n$ is large enough that $1+x/n \ge 0$ we have $(1+x/n)^j \le 1$, so $$\sum_{j=0}^{n-1} (1+x/n)^j \le \sum_{j=0}^{n-1} 1 = n $$ and since $x/n \le 0$ $$ \frac{x}{n} \sum_{j=0}^{n-1} (1+x/n)^j \ge \frac{x}{n} n = x $$ Thus $$(1+x/n)^n \ge 1 + x$$ Now take the limit as $n \to \infty$.

Robert Israel
  • 470,583
0

Hint: Break it into two cases, $x \leq 0$ and $x > 0$. One of them is trivial. For the other case, consider the Taylor Series expansion.

Jonathan Christensen
  • 3,900
  • 16
  • 21