While this seemed (and probably does seem to some of you) like a simple question a first it stumbled me a bit.

We were asked to calculate the derivative of:

$$f(x) = \left\{ \begin{array}{lr} e^{\frac{-1}{x^2}} & x \ne0\\ 0 & x =0 \end{array} \right.$$

at $x=0$ by definition, meaning though I can solve it easily with L'Hopital or maybe a taylor expansion of $e$ I cannot use it, only definition of derivatives and limits.

So I started by the defintion of the derivative at $x=0$ to get: $\lim_{x \to 0} \frac{e^{\frac{-1}{x^2}}-0}{x-0} = \lim_{x \to 0} \frac{e^{\frac{-1}{x^2}}}{x}$

Thinking I'd simplify it a bit I got to $\frac{1}{x \cdot e^{\frac{1}{x^2}}}$

Knowing limits around $0$ could be tricky I thought to take the onesided limits of $0^+, 0^-$

Thinking this would be a simple job I set out the prove the limit was $0$ (purely by intuition).Looking for example at the limit as $x \to 0^+$ I started by letting there be an $\epsilon$ I want to prove that $\exists \delta : x<\delta -> |f(x)|<\epsilon $

This is where I got stuck since what I'm left with is (after some algebra) $x \cdot e^{\frac{1}{x^2}} > \epsilon$

Now I'm sure there some algerba involved here to "extract" our variable to allow me to control $x$ by $\delta$ but I can't see it. Also I'm sure there must be some other way to solve it (by definition of course) since I also tried to substitute $u=\frac{1}{x^2}$ to reach a similar limit only to get stuck on the limit of $\frac{1}{\frac{e^u}{\sqrt{u}}}$ as $u \to \infty $ which I also believe to $0$ but also I don't know how to prove.

Any guidance would be appreciated! Thank you!


Note that, for every $x\ne0$, $$\left|\frac{\mathrm e^{-1/x^2}}x\right|=\exp(-a(x^2)),\qquad a(u)=\frac1{u}+\frac12\log u,$$ hence, if ever the limit of $a(u)$ when $u\to0$, $u\gt0$, is $+\infty$, then the limit of the ratio defining the derivative, is $0$. Note that $$a(u)=\frac{2+u\log u}{2u},$$ whose numerator goes to $2$ and the denominator goes to $0$ while being positive, when $u\to0$, $u\gt0$. Can you finish?

  • $\begingroup$ This actually makes it a lot clearer, thought I wouldn't have thought of that myself. Hmm regarding the end, I am not sure what do you mean by finish since this seems finished to me (maybe that's what you meant). Only thing I'm missing is the limit of $u \cdots log(u)$ in your proof which seems to be zero, not sure how to prove that without L'Hopital but I guess that's a topic for another question :) Thank you! $\endgroup$ – byentropy Aug 25 '14 at 13:07

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.