# If $f$ is continuous in $[0,1]$, find $\lim\limits_{x \to 0^{+}}x\int\limits_x^1 \frac{f(t)}t dt$

I'm solving this problem and I guess it shouldn't be too hard. Since $f$ is continuous it is bounded, so one has

$$\left| {x\int\limits_x^1 {\frac{{f\left( t \right)}}{t}dt} } \right| \leq x\int\limits_x^1 {\left| {\frac{{f\left( t \right)}}{t}} \right|dt} \leqslant Mx\int\limits_x^1 {\frac{{dt}}{t}} = - Mx\log x \to 0$$

Where $M=\operatorname{sup}\{|f(x)|:x\in[0,1]\}$

I'm not 100% certain on this, so I want a better, clearer approach.

Then, there is a second problem, similar, which is:

If $f$ is integrable on $[0,1]$ and $\exists\lim\limits_{x\to0}f(x)=L$, find

$$\ell = \mathop {\lim }\limits_{x \to {0^ + }} x\int\limits_x^1 {\frac{{f\left( t \right)}}{{{t^2}}}dt}$$

$$\mathop {\lim }\limits_{x \to {0^ + }} x\int\limits_x^1 {\frac{{f\left( t \right)}}{{{t^2}}}dt} =\mathop {\lim }\limits_{x \to {0^ + }} f\left( x \right) - xf\left( 1 \right) + x\int\limits_x^1 {\frac{{f'\left( t \right)}} {t}dt}$$

$$= L + \mathop {\lim }\limits_{x \to {0^ + }} x\int\limits_x^1 {\frac{{f'\left( t \right)}}{t}dt}$$

So, what can I sat about $f'(t)$ given $f(t)$ is integrable on $[0,1]$ that will allow me to apply the first case to the last limit?

• The first argument is fine, except that the first $<$ should be $\le$, and you should say explicitly that $M=\sup\{|f(x)|:x\in[0,1]\}$ and that at the end you’re taking the limit as $x\to 0^+$. Apr 8 '12 at 19:23
• @Brian Ok. I thought the $M$ was implicitly defined, but I guess it is appropriate to clarify. Any hints on the second one? Apr 8 '12 at 19:24
• Not offhand; there might be after I think about it a bit, but someone else is likely to get there first. Apr 8 '12 at 19:26
• You can't speak about $f'$ in second problem, since $f$ is only integrable Apr 8 '12 at 19:44
• @Norbert Is there any way to prove that $x\int\limits_x^1 {\frac{{f'\left( t \right)}}{t}dt} \to 0$? Apr 8 '12 at 19:48

For the first problem, your approach is fine (but the first inequality maybe be an equality when $f$ is non-negative). For the second, denote $L:=\lim_{x\to 0}f(x)$. Fix $\varepsilon>0$. We can find $\delta>0$ such that if $0\leq x\leq \delta$ then $|f(x)-L|\leq \varepsilon$, so for $0\leq x\leq \delta$: $$x\int_x^1\frac{f(t)}{t^2}dt=x\int_x^1\frac{f(t)-L}{t^2}dt+Lx\int_x^1\frac{dt}{t^2}=x\int_x^1\frac{f(t)-L}{t^2}dt+L\left(\frac 1x-1\right)x$$ hence \begin{align*}\left|x\int_x^1\frac{f(t)}{t^2}dt-L\right|&\leq x\int_x^1\frac{|f(t)-L|}{t^2}dt+|Lx|\\ &=x\int_x^\delta\frac{|f(t)-L|}{t^2}dt+ x\int_\delta^1\frac{|f(t)-L|}{t^2}dt+|Lx|\\ &\leq x\int_x^\delta\frac{\varepsilon}{t^2}dt+ x\int_\delta^1\frac{|f(t)-L|}{t^2}dt+|Lx|\\ &=\varepsilon x\left(\frac 1x-\frac 1{\delta}\right)+x\int_\delta^1\frac{|f(t)-L|}{t^2}dt+|Lx|\\ &=\varepsilon-\frac{\varepsilon}{\delta}x+x\int_\delta^1\frac{|f(t)-L|}{t^2}dt+|Lx| \end{align*} so $$\limsup_{x\to 0^+}\left|x\int_x^1\frac{f(t)}{t^2}dt-L\right|\leq \varepsilon$$ and since $\varepsilon$ was arbitrary, $L=\ell$.

• I'm OK with this, but can't you devise an approach that avoids the prediction that the limit is indeed $L$? (See my last edit, where it suffices to show that the integral goes to zero). Apr 8 '12 at 19:41
• @PeterT.off My approach gives prediction for integrabale $f$ :) Apr 8 '12 at 19:43
• First we work in the case on which $\ell=0$, , using $g=f-\ell$, then we try to generalize. Apr 8 '12 at 19:43

If $f$ is continuous let's use L'Hopital rule $$\lim\limits_{x\to+0}x\int\limits_{x}^{1}\frac{f(t)}{t}dt= \lim\limits_{x\to+0}\frac{\int\limits_{x}^{1}\frac{f(t)}{t}dt}{x^{-1}}= \lim\limits_{x\to+0}\frac{-\int\limits_{1}^{x}\frac{f(t)}{t}dt}{x^{-1}}= \lim\limits_{x\to+0}\frac{-\frac{f(x)}{x}}{-x^{-2}}= \lim\limits_{x\to+0}xf(x)=0$$ $$\lim\limits_{x\to+0}x\int\limits_{x}^{1}\frac{f(t)}{t^2}dt= \lim\limits_{x\to+0}\frac{\int\limits_{x}^{1}\frac{f(t)}{t^2}dt}{x^{-1}}= \lim\limits_{x\to+0}\frac{-\int\limits_{1}^{x}\frac{f(t)}{t^2}dt}{x^{-1}}= \lim\limits_{x\to+0}\frac{-\frac{f(x)}{x^2}}{-x^{-2}}= \lim\limits_{x\to+0}f(x)$$ P.S. Big thanks to David Mitra, he pointed out that requirement for integrals to be divergent is unnecessary!

• Accordingly upvoted. However, I'm looking for an approach following what I proposed. Apr 8 '12 at 19:45
• Remember that you can only use l'Hopital's rule if the limit is in an indeterminate form (in this case, $\infty/\infty$). So your proof works only in the case that the integral in the numerator goes to infinity. Apr 8 '12 at 20:11
• Ok, I will add this restrictions Apr 8 '12 at 20:13
• @GregMartin For the infinite limit case, you do not need the numerator to tend to infinity to use L'Hopital's rule, only the denominator. Apr 8 '12 at 20:17
• @GregMartin For your example, the limit of the quotient of the derivatives does not exist. L'Hopital doesn't apply (Lopital states if the limit of the quotient of the derivatives exists, then... Apr 8 '12 at 20:27