Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

I've never worked out the general proof of l'Hôpital before, and am taking the opportunity to do so before I return to the soul-crushing grind of college life.

I looked in Spivak and found a fine proof of the $(x \rightarrow a$ and $f(x), g(x) \rightarrow 0)$ case; but when he went to prove the $(x \rightarrow \infty, f(x), g(x) \rightarrow 0)$ case, he made a suggestion which actually does not work. (He proposed applying l'Hôpital for the first case above to $f(1/x) / g(1/x)$ as $x \rightarrow 0$, but since he used the fact that $f(0)$ and $g(0)$ could be defined to prove the first case, his version doesn't seem to apply. Correct me if I'm wrong!)

I adapted this proof of the second case.

Theorem: Suppose that $\lim_{x \to \infty} f(x), \lim_{x \to \infty} g(x) = 0$, and $\lim_{x \to \infty} \frac{f'(x)}{g'(x)} = l$. Then $ \lim_{x \to \infty} \frac{f(x)}{g(x)} = l.$

Proof: Note as a lemma that there exists an interval $I = (c, \infty)$ on which

  1. $g'(x) \neq 0$
  2. $g(x) \neq 0$

(The hypothesis implies (1); to prove (2), note that every interval $(\eta, \infty)$ would otherwise have $g(\xi) = 0$, so taking any $\eta ' > \eta$, there would be $g'(x) = 0$ on $(\xi, \xi ')$, and thus on $(\eta, \infty)$, against (1).)

Fix $\epsilon > 0$. By hypothesis, $\exists M: x > M \implies |\frac{f'(x)}{g'(x)} - l| < \frac{\epsilon}{2}$. If $M < c$, choose $M = c$ so $M \in I$.

Now pick any $y > M$. By Cauchy, $\forall x > y$ we get $($for $\mu \in (y, x))$

$$\frac{f(x) - f(y)}{g(x) - g(y)} = \frac{f'(\mu)}{g'(\mu)}.$$

But if we take $\lim_{x \to \infty}$ of both sides, we find

$$\lim_{x \to \infty}\frac{f(x) - f(y)}{g(x) - g(y)} = \frac{f(y)}{g(y)} = \lim_{x \to \infty} \frac{f'(\mu)}{g'(\mu)}. $$

We know that the RHS limit is determinate; we also know that

$$ l - \frac{\epsilon}{2} < \frac{f'(\mu)}{g'(\mu)} < l + \frac{\epsilon}{2} \implies l - \frac{\epsilon}{2} \leq \lim_{x \to \infty} \frac{f'(\mu)}{g'(\mu)} \leq l + \frac{\epsilon}{2}$$

$$\therefore l - \epsilon < \frac{f(y)}{g(y)} < l + \epsilon .$$

We have therefore furnished $M : |\frac{f(x)}{g(x)} - l| < \epsilon$ if $x > M$. This proves the theorem.

I now feel a sudden urge to begin overusing l'Hôpital! Someone, please tell me if this is correct, so I can go at it.

share|cite|improve this question
+1, if only for "soul-crushing grind of college life". I sincerely hope you were being facetious (else I'd change my upvote to a downvote). – Rick Decker Aug 8 '12 at 0:54
@Rick I thought that would be a winner :D – Chris Aug 8 '12 at 2:04
I upvoted since you bothered to check l'Hôpital; not too many students do that. – J. M. Aug 8 '12 at 3:17
up vote 1 down vote accepted

I don't see any major problems with your proof. Though, expanding on Spivak's hint, $f(1/x)g(1/x)$ doesn't work immediately. But, note that $$ \lim_{x \to 0^+} f(1/x) = \lim_{u \to \infty} f(u) = 0 $$ and similarly for $g$. Hence the functions $$ f^*(x) = \lim_{z\to x} f(1/z), $$ $$ g^*(x) = \lim_{z\to x} g(1/z) $$ do satisfy the hypotheses for L'Hopital on some interval $[0,\epsilon]$ for $\epsilon > 0$, (think of taking the reciprocal of the interval $[M,\infty]$, then $\epsilon = 1/M$).

share|cite|improve this answer
I am not precisely clear on your meaning. Could you carry this out? (Also: no major problems, but any minor ones?) – Chris Aug 8 '12 at 2:20
I don't really want to type out all the details. Is here a certain part in particular that is unclear? – nullUser Aug 8 '12 at 2:23
Well, it seems that you're suggesting we patch the continuity hypothesis by taking advantage of the differentiability of $f$ for real arguments which are large (considering the reciprocals of small numbers), and then defining $f^*(0) = \lim_{z \to x}f(\frac{1}{z})$. Am I correct on that? – Chris Aug 8 '12 at 2:28
Yes, (assuming you mean $f^*(0) = \lim_{z\to 0} f(1/z)$ of course). – nullUser Aug 8 '12 at 2:29
Alright, I checked the details. Thank you! You're free to go :D – Chris Aug 8 '12 at 2:44

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.