12
$\begingroup$

How do I prove the inequality $e^{-2x}\leq1-x$ for $0\leq x\leq1/2$?

$\endgroup$
3
  • 1
    $\begingroup$ @TheChaz: Out of curiosity: where did you see any research effort? $\endgroup$
    – user6701
    Commented Jan 11, 2012 at 5:34
  • 6
    $\begingroup$ @Tim: Nowhere. I downvoted this. $\endgroup$ Commented Jan 11, 2012 at 14:33
  • 2
    $\begingroup$ @TheChaz, Sarcasm does not translate well on the internet. It would have been much more helpful and less confusing had you given your actually reason why you downvoted this question. $\endgroup$
    – user4143
    Commented Jan 11, 2012 at 23:43

6 Answers 6

34
$\begingroup$

For example, $(1-x)(1+2x)=1+x-2x^2\geq 1$ so $$ e^{2x}\geq 1+2x \geq \frac{1}{1-x} $$

$\endgroup$
16
$\begingroup$

A slightly ugly but standard beginning calculus approach is to let $$f(x)=1-x-e^{-2x}.$$ We want to show that $f(x)\ge 0$ in the interval $[0,1/2]$.

A first experimental step might be to use software to graph $y=f(x)$ as $x$ ranges over our interval. If we have a high degree of trust in the graphing software, the picture tells us that $f(x)$ is quite likely to be $\ge 0$ in our interval.

Certainty is better. Use the derivative to study the behaviour of $f(x)$. Note that $f(0)=0$, and $$f'(x)=2e^{-2x} -1.$$ The derivative is positive at $x=0$, and is clearly decreasing. It reaches $0$ at $x=(\ln 2)/2\approx 0.3465$. So $f(x)$ is increasing in the interval $[0,(\ln 2)/2]$, and decreasing from $(\ln 2)/2$ on. By $x=1/2$, $f(x)$ is about $0.13212$, and in particular still positive.

Thus $f(x)$ is $\ge 0$ for $0 \le x\le (\ln 2)/2$, and $f(x) >0$ for $(\ln 2)/2 \le x\le 1/2$. It follows that $f(x)\ge 0$ on the whole interval $[0,1/2]$ (and somewhat beyond $1/2$).

Comment: There are far better ways to prove the inequality. But let's stick to calculusy approaches. To cut down on the negatives, note that equivalently we want to show that $e^{2x}(1-x) \ge 1$ on our interval (we multiplied both sides by the positive number $e^{2x}$). Let $g(x)=e^{2x}(1-x)-1$. Then $g(0)=0$. Also, $g'(x)=(1-2x)e^{2x}$, so $g$ is increasing on the interval $[0,1/2]$, and we are finished.

$\endgroup$
3
  • $\begingroup$ Perhaps it's good to give an example where a software plot is misleading: this function has a value $<0$ at $x=2$, yet the plot seems to be positive for all positive $x$. – Actually, I'm a bit surprised Wolfram|Alpha does this, a good CAS certainly should be able to proove that the plotted function is well-behaved between two evaluation points, and only then simply connect them. $\endgroup$ Commented Jan 10, 2012 at 23:22
  • 1
    $\begingroup$ @leftaroundabout: The SE software seems to have mangled your link. I think you mean something like this instead (with ^ encoded as %5E). $\endgroup$ Commented Jan 11, 2012 at 12:45
  • $\begingroup$ @IlmariKaronen yeah that's what I meant. $\endgroup$ Commented Jan 11, 2012 at 15:33
15
$\begingroup$

Let $f(x)=e^{-2x}-(1-x)$. We have:

  • $f''(x)=4e^{-2x}$, which is always positive. Hence, $f$ is convex.
  • $f(0)=0$
  • $f\left(\dfrac{1}{2}\right)=\dfrac{1}{e}-\dfrac{1}{2}\leq0$.

Since $f$ is convex, its curve is always below $0$ on $\left[0,\dfrac{1}{2}\right]$. This proves that $e^{-2x}\leq 1-x$.

$\endgroup$
3
$\begingroup$

By Taylor expansion, we have $$\ln \frac{1}{1-x} = \sum_{n=1}^{\infty} \frac{x^n}{n},$$ whose convergence radius is $R=1$.

The equation above can also be achieved by integrating both sides of $$\frac{1}{1-x} = \sum_{n=0}^{\infty} {x^n} .$$

For $0\leq x\leq 1/2$ we have the following upper bound $$ \sum_{n=1}^{\infty }\frac{x^{n}}{n}\leq \sum_{n=1}^{\infty }\frac{1}{ n\left( 2^{n}\right) }=\ln 2 \leq 2. $$ Therefore $$ -\ln \left( 1-x\right) =\sum_{n=1}^{\infty }\frac{x^{n}}{n}\leq 2x. $$ The given inequality follows.

$\endgroup$
2
$\begingroup$

We may start out in an elementary manner with a very known result that comes from W. Sierpiński inequality or $(e^x)$ Taylor expansion, namely:

$$e^x\geq 1+x $$

We get that: $$e^{-x}\leq\frac1{x+1}; \space e^{-2x}\leq\frac1{2x+1}\leq1-x,\space \space \space 0\leq{x}\leq\frac{1}{2}$$

The proof is complete.

$\endgroup$
2
  • 1
    $\begingroup$ In fact your proofs works for $0\le x\le 2$. We have (for $x\ge 0$) the following: $\frac1{1+2x}\le 1-x$ $\Leftrightarrow$ $1\le (1+2x)(1-x)=1+x-2x^2$ $\Leftrightarrow$ $0\le x-x^2=x(x-2)$ $\Leftrightarrow$ $x\in[0,2]$. $\endgroup$ Commented Jun 19, 2012 at 12:58
  • $\begingroup$ @Martin Sleziak: thanks for your comment! $\endgroup$ Commented Jun 19, 2012 at 13:03
1
$\begingroup$

For $x > 0$, the Taylor series for $e^{-2x}$ is the alternating series $\sum_n {(-2x)^n \over n!} = 1 - 2x + 2x^2 -...$ If $x < {1 \over 2}$, the terms decrease in magnitude. It's a fact about alternating series that in an alternating series with terms of decreasing magnitude, the overall sum is going to be less than what you get if you add finitely many terms, stopping with a positive term. So we have $$e^{-2x} < 1 - 2x + 2x^2$$ But after a little simple algebra, for $x > 0$ the statement $1 - 2x + 2x^2 < 1 - x$ is equivalent to the statement that $x < {1 \over 2}$. So the inequality follows. (The endpoint case $x = {1 \over 2}$ can be gotten by just plugging in the value).

$\endgroup$

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .