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.

Please try to help me with a question that I'm trying to solve.

$f(x)$ is continuous in the range of $[0, \infty)$ and $\lim \limits_{x \to \infty}x^2 f(x) = 1$.

Calculate $$\lim_{n \to \infty} \int_0^1 f(nx) \,dx .$$

share|improve this question
    
Thanks for the edit –  Guy Dec 16 '11 at 16:33
4  
Try a few examples to get a guess for the answer. –  GEdgar Dec 16 '11 at 16:42
3  
Note that being continuous at zero matters. $f(x) = x^{-2}$ doesn't give the same result. –  Michael Lugo Dec 16 '11 at 16:59

4 Answers 4

up vote 9 down vote accepted

Here's an argument based on l'Hôpital's rule:

$$ \begin{array}{rcl} \lim_{u \to \infty} \int_0^1 f(ux) \, dx &\stackrel{\color{Brown}{ux = y}}{=}& \lim_{u \to \infty}\frac{\int_0^u f(y) \, dy}{u} \\&\stackrel{\color{Brown}{\text{l'Hôpital}}}{=}& \lim_{u \to \infty} \frac{\frac{d}{du}\int_0^u f(y) \, dy}{\frac{d}{du}u} \\& \stackrel{\phantom{ux \to y}}{=}& \lim_{u \to \infty} f(u) \\&\stackrel{\phantom{ux \to y}}{=}& 0, \end{array} $$ since $u^2 f(u) \to 1$ as $u \to \infty$.

In this proof, we use a general version of l'Hôpital's rule that is apparently not as well-known as it deserves to be. This requires only the denominator to approach $\infty$; it is not necessary that the numerator also go to infinity. This has been explained by Bill Dubuque in several posts in this site; e.g., see here and here. The linked posts contain a formal statement of the theorem.

share|improve this answer
    
Nice. It might be a good idea to point out which version of L'Hopital you're using (the "anything" over $\infty$ version; which, I don't think is very well known to students) :) –  David Mitra Dec 16 '11 at 17:18
    
how can you use by l'Hôpital's rule? don't you need to know that the numerator also -> to infinity? –  Guy Dec 16 '11 at 17:19
    
@David, Thanks for your comment. I am actually searching for the precise version right now. =) Bill Dubuque has explained this in quite a few posts; I will update my answer with a link. –  Srivatsan Dec 16 '11 at 17:20
    
@Guy You only need the denominator going to infinity. This is why I made my first comment. –  David Mitra Dec 16 '11 at 17:20
    
Thank you very much –  Guy Dec 16 '11 at 18:10

First note that, since $f$ is continuous on $[0,\infty)$, it is integrable over any interval of finite length contained in $[0,\infty)$.

I assume you want to compute $$ \lim_{n\rightarrow \infty}\int_0^1 f(nx)\,dx.$$

The above limit is equal to (making the substitution $u=nx$): $$ \lim_{n\rightarrow \infty}{1\over n}\int_0^n f(u)\,du. $$

Since $\lim\limits_{x\rightarrow\infty}\bigl[ x^2 f(x)\bigr]=1$, for sufficiently large $x$, say $x\ge A$: $$ \tag{1}{1\over 2x^2}\le f(x)\le {2\over x^2} $$

Now, for any number $C $ $$ \tag{2}\lim_{n\rightarrow \infty}{1\over n}\int_A^n {C\over x^2}\,dx= \lim_{n\rightarrow \infty}{1\over n} \Bigl[{C\over A}-{C\over n}\,\Bigr]=0. $$ From (1) and (2), it follows that $$\lim\limits_{n\rightarrow \infty}{1\over n}\int_A^n f(u)\,du=0.$$

Thus $$\eqalign{ \lim_{n\rightarrow \infty}{1\over n}\int_0^n f(u)\,du &=\lim_{n\rightarrow \infty}\Bigl[\,{1\over n}\int_0^A f(u)+{1\over n}\int_A^n f(u)\,du\Bigr]\cr &=\lim_{n\rightarrow \infty} \,{1\over n}\int_0^A f(u)\ +\ \lim_{n\rightarrow \infty}{1\over n}\int_A^n f(u)\,du \cr &=0. } $$

share|improve this answer
    
Thank you very much –  Guy Dec 16 '11 at 18:10

$\int_0^1 f(nx)dx = \frac{1}{n}\int_0^n f(x)dx$. But we can easily show by the property of $f(x)$ that $\int_0^\infty f(x)dx = C$ exists, and that there is an $x_0$ such that $f(x)>0$ if $x>x_0$. From that we determine that, when $n>x_0$.

$$\int_0^1 f(nx)dx = \frac{1}{n} \int_0^n f(x)dx < \frac{1}{n}\int_0^\infty f(x)dx = \frac C n$$. So for $\epsilon>0$ if $n>C/\epsilon$, $\int_0^1 f(nx)dx < \epsilon$.

Now you just need t show that the limit must be non-negative, and then you've shown the limit must be zero.

share|improve this answer

We have $\int_0^1f(nx)dx=\frac 1n\int_0^nf(t)dt$ after the substitution $t=nx$. We fix $\varepsilon>0$ and $A$ such that $|x^2f(x)-1|\leq \varepsilon$ if $x\geq A$. We have \begin{align*} \left|\int_0^1f(nx)dx\right|&=\frac 1n\left|\int_0^1f(t)dt\right|+\frac 1n\left|\int_1^n\frac{t^2f(t)-1}{t^2}dt+\int_1^n\frac{dt}{t^2}\right|\\ &\leq \frac 1n\left|\int_0^1f(t)dt\right|+\frac 1n\left|\int_1^A\frac{t^2f(t)-1}{t^2}dt\right|+\frac 1n\int_A^n\frac{|t^2f(t)-1|}{t^2}dt+\frac 1n\int_1^{+\infty}\frac{dt}{t^2}\\ &\leq \frac 1n\left|\int_0^1f(t)dt\right|+\frac 1n\left|\int_1^A\frac{t^2f(t)-1}{t^2}dt\right|+\frac{\varepsilon}n\int_A^n\frac{dt}{t^2}+\frac 1n\\ &\leq \frac 1n\left(\left|\int_0^1f(t)dt\right|+\left|\int_1^A\frac{t^2f(t)-1}{t^2}dt\right|+\varepsilon+1\right). \end{align*} Now, one can find a $n_0$ such that if $n\geq n_0$ then $\left|\int_0^1f(nx)dx\right|\leq \varepsilon$, and the limit is $0$.

share|improve this answer

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.