9

How can we prove the inequality $e^x \le x + e^{x^2}$ for $x\in\mathbb{R}$?

Element118
  • 3,793

4 Answers4

4

Note the inequality $e^t \ge t + 1$ for all $t \in \mathbb{R}$. In particular $e^{x^2} \ge x^2 + 1$.

If $x \le -1$, then $$ e^{x^2} - e^x + x \ge x^2 + 1 - e^0 + x = x(x+1) \ge 0. $$ If $-1 < x < 1$, then \begin{align*} e^{x^2} - e^x + x &\ge x^2 + x + 1 - \sum_{k \ge 0} \frac{x^k}{k!} \\ &= \frac{x^2}{2!} - \sum_{k \ge 3} \frac{x^k}{k!} \\ &\ge \frac{x^2}{2!} - \sum_{k \ge 3} \frac{x^2}{k!} \\ &= x^2 \left( \frac12 - \left[e - 1 - \frac{1}{1!} - \frac{1}{2!} \right]\right) \\ &= x^2 \left( 3 - e \right) \ge 0. \end{align*}

Finally, if $x \ge 1$, then $$ e^{x^2} - e^x + x > e^{x} - e^x + x > 0. $$

1

You have that $e^1< 1+e^{1^2}$. Now see what happens with the derivative on each side, e.g., we have for $x>1$ that $e^x < 1+2xe^{x^2}$ , since $2x>1$ and $e^{x^2}>e^{1}=e$ , so that , when $x>1$, the left-hand side is smaller than the R.H side, and then the R.H side grows faster nafterwards, for $x>1$. Try seeing what happens left of $1$.

We also have that for $x=1 , e^{-1}<-1+e^1$ , and $h(x)=1+2xe^{x^2}$ decreases faster than $e^{x}$ in $(-\infty,-1]$ (Meaning $h(|x|) $ grows faster than $e^{|x|}$). I don't think this is too hard to prove; for $x<-1$ , we have $1-2e^{x^2}< e^{x}$.

The whole issue comes down to the fact that $e^{x^2}$ grows way faster than $e^x$ in $[1, \infty)$ and decreases way faster in $(-\infty, -1]$. In $[-1,1]$ the result still holds; it is a little late now, but I will have a proof by Tuesday night, or I will delete this.

user99680
  • 6,836
  • 16
  • 25
  • Is "seeing what happens left of 1" supposed to be simple? – MT_ May 05 '14 at 03:57
  • @MichaelT : Rather, I suggest choosing $e^0 = 0 + e^{0^2}$ as the start. – Graham Kemp May 05 '14 at 03:59
  • @GrahamKemp I would too, but the trouble of comparing first derivatives for this problem IMO is exploring $x \in (0, 1)$ since $x^2 < x$ and, if using derivatives, $2x < 1$. – MT_ May 05 '14 at 04:01
  • @MichaelT:The situation left of 1 is very similar to that right of 1. Things do get a little more complicated in $(-1,1)$ – user99680 May 05 '14 at 06:26
  • Any comments on the down vote? – user99680 May 05 '14 at 06:26
  • @MichaelT:Sorry, I meant left of $-1$. – user99680 May 05 '14 at 06:39
  • @user99680 Yes, that was me. When you said "left of $1$" it seemed that you were leaving the only non-trivial part of the problem (i.e. $x \in [-1, 1]$) as an "exercise for the OP" kind of thing. Now that you plan to update your answer, I will remove it. – MT_ May 05 '14 at 11:20
1

$$e^{x} = 1 + x + \frac{x^2}{2!} + \dots \le 1 + x + 2 \left(\left( \frac{x}{2} \right )^2+\left( \frac{x}{2} \right )^3 +\dots \right)\le 1 + x+\left( \frac{x}{2} \right )^2 \left( \frac {4}{2 - x}\right)$$

$$x + e^{x^2} \ge x +1 + x^2 + \frac{x^4}{2!} = 1+ x+ \left( \frac{x}{2} \right )^2 2(2+x^2) \ge 1+ x+ 4 \left( \frac{x}{2} \right )^2 $$

The inequality is true for all $\Bbb R \setminus (1,3)$, let $b \in (1,2)$, Also, $$e^{(1+b)^2}+1+b \ge e^{(1+b)^2} \ge e^{1+b}$$ hence the inequality holds for all $x\in \Bbb R$

S L
  • 11,961
-2

Hmmm.... I will give it a try. Consider the function $f(x) = e^{x²} + x - e^x$, $x \in \mathbb{R} $. Find the global minimum of this function. It turns out to be zero. Does that prove it ?

Note: The second derivative is $e^{x^2} (4 x^2+2)-e^x$ which is strictly positive. $e^{x^2}$ becomes less than $x$ between $0$ and $1$ but that difficulty is overcome by multiplying by 2 (*)

(*)That difficulty is not overcome just by multiplying by 2, I learned it the hard way. You still have to prove it !

I did not answer the question, but I will keep the post here. Reminding me of my once failed efforts!

  • How did you find the global minimum? – MT_ May 05 '14 at 03:47
  • Yes, you are right. I am stuck at proving that x=0 is the global minimum. – Jonas Lehmann May 05 '14 at 03:48
  • @MichaelT $f'(x) = 2x e^{x^2} + 1 - e^x, f'(x)=0 \implies x=0$ – Graham Kemp May 05 '14 at 03:50
  • @MichaelT I don't know, wouldn't it work out if we take the derivative and equate it to zero. – The very fluffy Panda May 05 '14 at 03:50
  • @GrahamKemp How do we know that there aren't other solutions? Even if we do show that to be the only critical point and it to be a minimum, that wouldn't necessary mean it is a global minimum. Now, it isn't necessarily very hard to do these things, but it's important that they are done. – MT_ May 05 '14 at 03:52
  • @MichaelT You are right, not turning out to be that easy. – The very fluffy Panda May 05 '14 at 03:54
  • 2
    The second derivative at 0 is strictly positive. So your function is concave up around 0. If the graph turned around at some point, the derivative would vanish elsewhere (which it doesn't). –  May 05 '14 at 03:55
  • 1
    Since the second derivative is positive everywhere, which can be shown by noting that $e^{x^2} > e^x$ for all x, then the first derivative is always increasing so x=0 is the absolute min. – Steven F May 05 '14 at 04:00
  • @MichaelT Examine the second derivative as Bryan and StevenF note. – Graham Kemp May 05 '14 at 04:02
  • 3
    @StevenF $e^{x²}$ is not greater than $e^x$ for all $x$ – The very fluffy Panda May 05 '14 at 04:06
  • @PandaBear You are right. I was hasty there. The entire inequality is $(4x^2+2) e^{x^2} - e^x > 0$ The issue is in (0,1). – Steven F May 05 '14 at 04:16
  • @PandaBear I am not sure that just saying that multiplying by 2 is enough to make it always bigger. It is true but in full rigor we might have to recognize to show that $2 e^{x^2}- e^x > 0 $ is to show that $4 e^{x^2} - e^x > 0 $. I arrive here by looking of the second derivative of the former. This process can be continued out ad infinitum to $2^n e^{x^2} - e^x > 0 $. But I still wonder if even that serves rigorously enough. – Steven F May 05 '14 at 04:41
  • @StevenF Oh yes.. you are right. It's not that easy. – The very fluffy Panda May 05 '14 at 04:49