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.

Question. Let $(X, M, m)$ be an arbitrary measure space. Let $f \in L^1_m(X)$. Then $$\lim_{n \rightarrow \infty} \int_X n \log \left( 1 + \frac{|f(x)|^2}{n} \right) \,dm(x) = 0.$$


Solution.

Observe that $$n \log \left( 1 + \frac{|f(x)|^2}{n} \right) \leq n \log \left( 1 + \frac{|f(x)|}{n} \right)^2 = 2n \log \left( 1 + \frac{|f(x)|}{n} \right) \leq 2n \frac{|f(x)|}{n} = 2|f(x)|.$$

Then $n \log \left( 1 + \frac{|f(x)|^2}{n} \right)$ is dominated by $2|f(x)|$.

Moreover, we have

$$\lim_{n \rightarrow \infty} \frac{\log \left( 1 + \frac{|f(x)|^2}{n} \right)}{1/n} = \frac{0}{0}.$$

So we apply L'Hopital's Rule and we obtain:

$$\lim_{n \rightarrow \infty} \frac{\log \left( 1 + \frac{|f(x)|^2}{n} \right)}{1/n} = \lim_{n \rightarrow \infty} \frac{\frac{-|f(x)|^2}{n^2} (-n^2)}{ 1 + \frac{|f(x)|^2}{n} } = |f(x)|^2.$$

Then by Lebesgue Dominated Convergence Theorem, we conclude that

$$\lim_{n \rightarrow \infty} \int_X n \log \left( 1 + \frac{|f(x)|^2}{n} \right) \,dm(x) = \int_X |f(x)|^2 \,dm(x).$$


Where is the proof wrong? Thanks.


Added Later. I found the mistake: $n \log \left( 1 + \frac{|f(x)|^2}{n} \right) \leq n \log \left( 1 + \frac{|f(x)|}{n} \right)^2$ is not true for all $n$. As @did suggested, dominating function should be $|f|^2$.

share|improve this question
1  
Your title does not seem to be correct, since the limit is not $0$ but the integral of $|f|^2$. –  1015 Jan 6 '13 at 20:36
1  
And please replace every $dm$ by $dm(x)$. –  Did Jan 6 '13 at 23:02

1 Answer 1

up vote 2 down vote accepted

Let $g_n=n\log(1+|f|^2/n)$, then $\log(1+t)\leqslant t$ for every $t$, hence $g_n\leqslant|f|^2$. If $f$ is in $L^2$, this shows that $(g_n)$ is dominated. Since $g_n\to|f|^2$ pointwise, one gets $\int\limits_Xg_n\to\int\limits_X|f|^2$.

If $f$ is not in $L^2$, let $A(C)=\{|f|\leqslant C\}$ for some $C\gt0$, then $g_n\geqslant n\log(1+|f|^2/n)\mathbf 1_{A(C)}$. Furthermore, $\log(1+t)\geqslant t\log2$ for every $t\leqslant1$ hence, for every $n\geqslant C^2$, $g_n\geqslant(\log2)|f|^2\mathbf 1_{A(C)}$ and $\liminf\limits_{n\to\infty}\int\limits_Xg_n\geqslant\int\limits_{A(C)}(\log2)|f|^2$. This holds for every $C$ and $\int\limits_{A(C)}|f|^2$ can be made as large as one wants hence $\int\limits_Xg_n\to+\infty$.

share|improve this answer
    
Your $n\log(1+|f|^2/n)\leqslant2|f|$ is wrong. –  Did Jan 6 '13 at 19:38
    
Then the question is not correct? I mean, $\lim_{n \rightarrow \infty} \int_X n \log \left( 1 + \frac{|f(x)|^2}{n} \right) \,dm \neq 0.$ –  James Jan 6 '13 at 19:43
    
Absolutely. (Except if $f=0$ almost everywhere.) –  Did Jan 6 '13 at 19:49
    
If $f$ is not in $L^2$, the the limit is still $\int_X|f|^2=+\infty$. –  1015 Jan 6 '13 at 20:38
1  
Try $|f(x)|=n$. –  Did Jan 6 '13 at 20:48

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.