5
$\begingroup$

I was wondering what is the value of $\lim\limits_{x\to 0+} x^{x}$.

Assuming that the limit exists, I could show using the usual logarithm techniques that the limit is $1$. However, I am not able to show that the limit exists. Could some one help on that?

EDIT: It would be so nice if we could do it without l'Hopital...

$\endgroup$
  • $\begingroup$ Maple helps you with it by the $ Student[Calculus1][LimitTutor](); $ command. See the output in the screen. $\endgroup$ – user64494 Aug 22 '13 at 13:13
  • $\begingroup$ What for without l'Hopital? $\endgroup$ – user64494 Aug 22 '13 at 13:17
  • 2
    $\begingroup$ Are ruler and compass allowed? $\endgroup$ – user64494 Aug 22 '13 at 13:49
  • $\begingroup$ BTW, the l'Hopital rule also bases the existence of the limit $\lim_{x\to a+} \frac {f(x)} {g(x)}$ if the limit $\lim_{x\to a+} \frac {f'(x)} {g('x)}$ exists. $\endgroup$ – user64494 Aug 22 '13 at 15:23
  • 8
    $\begingroup$ Considering all the comments together, "ruler and compass" is a way of saying that the restriction to do it without L'Hopital is (in one user's opinion) unnatural, unnecessary, undesirable, or un-somethingelse. That is one answer to the question about this on the meta site. $\endgroup$ – zyx Aug 28 '13 at 14:54
6
$\begingroup$

Hint: Using the change of variable $x=e^{-t}$ show that $$\lim_{x \to 0+} x \ln x = 0.$$

$$\lim_{x \to 0+} x \ln x = \lim_{t \to +\infty} -\frac{t}{e^t} = 0.$$

To prove the last equality you can use the Taylor expansion for $e^t$, for instance.

| cite | improve this answer | |
$\endgroup$
  • $\begingroup$ Using Taylor expansion is basically L'Hopital. Which OP wants to avoid if possible. $\endgroup$ – Arthur Aug 22 '13 at 13:17
  • 1
    $\begingroup$ He added that condition after my answer. Then he can prove $\frac{t}{e^t} \to 0$ first for natural numbers (e.g., using binomial formula) then extend it to reals. $\endgroup$ – njguliyev Aug 22 '13 at 13:22
  • 2
    $\begingroup$ This makes perfect sense as for me, $e^{x}$ is defined as the series expansion. $\endgroup$ – Vishal Gupta Aug 22 '13 at 13:22
6
$\begingroup$

By the l'Hôpital theorem we have

$$\lim_{x\to0}x\log x=\lim_{x\to0}\frac{\log x}{\frac 1 x}=-\lim_{x\to0}\frac{\frac 1 x}{\frac{ 1 }{x^2}}=-\lim_{x\to0}x=0$$

and conclude with $$x^x=e^{x\log x}$$

Edited To find the limit without the l'Hôpital theorem: for $0<x<1$: $$ 0\leq|x\log x|=x\int_x^1\frac{dt}{t}\leq x\int_x^1\frac{dt}{t^{3/2}}=-2x\frac{1}{\sqrt{t}}\Big|_x^1\to0$$

| cite | improve this answer | |
$\endgroup$
  • 1
    $\begingroup$ That is a nice answer. But can we do it without recourse to l'Hopital? $\endgroup$ – Vishal Gupta Aug 22 '13 at 13:09
  • $\begingroup$ @ Vishal: What for? $\endgroup$ – user64494 Aug 22 '13 at 13:16
  • 1
    $\begingroup$ @Vishal I edited my answer. $\endgroup$ – user63181 Aug 22 '13 at 13:25
4
$\begingroup$

Assume you know $$\lim_{n \rightarrow \infty}\sqrt[n]{n}=1$$ Proof:

$$1 \leq \sqrt[n]{n}<\frac{n-2+2\sqrt n}{n}, \,\,\,\,\,\,\,\,\,\, \mathrm{GM \leq AM}$$

Now for each $x$ near zero define $n(x)=[\frac{1}{x}]$ where $[\cdot]$ is the integral part function. Let's call $n(x)$ out of simplicity $n$.

Therefore \begin{align} &\frac{1}{n} \geq x \geq \frac{1}{n+1} \frac{}{} \Rightarrow\\ \left (\frac{1}{n} \right )^\frac{1}{2n}\geq &\left (\frac{1}{n} \right )^{\frac{1}{n+1}} \geq x ^{x}\geq \left (\frac{1}{n+1} \right )^\frac{1}{n} \geq \left (\frac{1}{2n} \right )^\frac{1}{n}& \end{align}

And because as $x$ tends to zero $n(x)$ tends to infinity taking limits at both sides you are done.

| cite | improve this answer | |
$\endgroup$

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.