Take the 2-minute tour ×
Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. It's 100% free, no registration required.

In a paper, it was claimed that $\lim_{x \to \infty} (1-\frac{f(x)}{x})^x \sim e^{-f(x)}$ when $\frac{(f(x))^2}{x}$ is $o(1)$.

I proved the claim in the following way; however, I'm seeking a simpler proof.

Define $g(x) = \frac{(1-f(x)/x)^x}{x}$ and $h(x) = \frac{e^{-x}}{x}$. To prove the theorem, we must show that $\lim_{x \to \infty} \frac{g(x)}{h(f(x))} = 1$ when $\frac{(f(x))^2}{x}$ is $o(1)$. To this end, we use the binary expansion for $g(x)$, and Taylor series for $h(x)$:

$\lim_{x \to \infty} g(x) = \lim_{x \to \infty} \frac{(x-f(x))^x}{x^{x+1}} = \lim_{x \to \infty} \frac{x^x - \binom{x}{1}x^{x-1}f(x) + \binom{x}{2}x^{x-2}(f(x))^2 - \cdots}{x^{x+1}} = \lim_{x \to \infty} \frac{1-f(x)}{x}$

(The last identity follows from the fact that $\frac{(f(x))^2}{x}$ is $o(1)$; that is $\lim_{x \to \infty} \frac{(f(x))^2}{x} = 0$.)

Now, since $h(x) \sim \frac{1 - x + x^2/2 - \cdots}{x}$, we have $h(f(x)) \sim \frac{1 - f(x) + (f(x))^2/2 - \cdots}{x}$, and $\lim_{x \to \infty} h(f(x)) = \lim_{x \to \infty} \frac{1-f(x)}{x}$. (Again, the last identity follows from the fact that $\frac{(f(x))^2}{x}$ is $o(1)$.)

Combining the two limits, we see that $\lim_{x \to \infty} \frac{g(x)}{h(f(x))} = 1$.

As I said, this proof is long, and to me, it is not appealing.

Does anyone know a better proof?

share|improve this question
2  
I think using the expansion of the $\ln$ is more natural than the binomial expansion. Not that that shortens the proof a lot. –  Raskolnikov Mar 29 '11 at 21:45
add comment

2 Answers 2

up vote 12 down vote accepted

Let'sconsider $$\lim_{x\to\infty} \left(1 - \frac{{f(x)}}{x}\right)^{x} = \lim_{x\to\infty} \exp\left[x \ln \left(1 - \frac{f(x)}{x}\right)\right].$$ Expanding the logarithm, we obtain $$x \ln \left(1 - \frac{f(x)}{x}\right) = x \left[ - \frac{f(x)}{x} + O\left(\frac{f(x)}{x}\right)^2\right] = -f(x) + O\left(\frac{f(x)^2}{x}\right).$$ So we obtain (using the fact that $f(x)^2/x$ is $o(1)$) $$\left(1 - \frac{{f(x)}}{x}\right)^{x} = e^{-f(x) + o(1)} \sim e^{-f(x)}.$$

share|improve this answer
    
+1, but the end should be revised. –  Shai Covo Mar 29 '11 at 22:33
    
@Shai Covo: I will try to fix that; I forgot that $f(x)$ depends on $x$... ;-) –  Fabian Mar 29 '11 at 23:03
add comment

It might be useful to consider $(1 - \frac{{f(x)}}{x})^{(x/f(x))f(x)}$.

EDIT: This is a comment (in response to the OP's comment below).

First of all, note that without loss of generality we can suppose that $f(x) \neq 0$. Now, at first glance, it seemed that writing $(1 - \frac{{f(x)}}{x})^x $ as $(1 - \frac{{f(x)}}{x})^{(x/f(x))f(x)}$ might be useful, since $(1 - \frac{{f(x)}}{x})^{x/f(x)} \to e^{-1}$. Indeed, note that since $\frac{f^2(x)}{x}$ is $o(1)$, $f(x)/\sqrt{x} \to 0$, hence also $f(x)/x \to 0$. However, this approach only gives intuition for the desired asymptotic equality, and beyond that is apparently not useful in our context. Fabian's approach seems the best to me.

share|improve this answer
    
Too many upvotes for this hasty suggestion... –  Shai Covo Mar 29 '11 at 22:01
    
Should I balance them out?? ;) –  The Chaz 2.0 Mar 30 '11 at 1:06
    
@The Chaz: I will try to improve the answer... –  Shai Covo Mar 30 '11 at 1:35
    
@Shai Covo: I was joking! You are very helpful. Have a good day! –  The Chaz 2.0 Mar 30 '11 at 2:22
    
@The Chaz: Thanks! –  Shai Covo Mar 30 '11 at 3:01
show 2 more comments

Your Answer

 
discard

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.