This is almost surely an Arzela-Ascoli question, since it comes from an old exam of which such problems are quite common. Unfortunately, I can't seem to get it though.

$\{ f_n \}$ is a sequence of functions $[0,1] \to \mathbb{R}$ satisfying $|f_n'(x)| \leq \frac{1 + |\ln (x)|}{\sqrt{x}}$ and $-10 \leq \int_0^1 f_n(x) dx \leq 10$ for every $n$. The question is to show the existence of a uniformly convergent subsequence.

In these kinds of situations, the bounds given usually turn into something nice to force uniform boundedness. Maybe I could hope to establish equicontinuity using the integral somehow, but I'm not told that the $f_n$ are always positive, so the absolute values would probably screw things up anyway.

Is it an easy problem I'm missing or is this totally the wrong track? If so, what's the right track?

  • $\begingroup$ Sorry for the confusion. I want to show there is a uniformly convergent subsequence. That's why Arzela-Ascoli seems to be the way to go. $\endgroup$ – Alfred Yerger May 14 '17 at 22:10
  • $\begingroup$ Are the $f_n$ continuously differentiable? $\endgroup$ – zhw. May 15 '17 at 0:20
  • $\begingroup$ Everywhere except at 0 I imagine. The other half of this comment is false and so I removed it. $\endgroup$ – Alfred Yerger May 15 '17 at 0:28
  • $\begingroup$ I edited my previous hint to a more complete solution. $\endgroup$ – zhw. May 17 '17 at 6:37
  • $\begingroup$ @zhw. For some reason I couldn't even see your answer. I offered the bounty to attract an answer! I didn't know there already was one. $\endgroup$ – Alfred Yerger May 17 '17 at 15:55

I assume the $f_n$ are differentiable on $[0,1]$ and are continuously differentiable on $(0,1].$

Here's a sketch that should help you along (ask if you have questions). First, the function

$$x\to \int_0^x \frac{1+|\ln t|}{\sqrt t}\,dt$$

is continuous, hence uniformly continuous, on $[0,1].$ Thus given $\epsilon>0,$ we will have

$$ \int_x^y \frac{1+|\ln t|}{\sqrt t}\, dt < \epsilon$$

if $x,y$ are close to each other.

It follows that for such $x,y$ we have

$$|f_n(y) - f_n(x)| = |\int_x^y f_n'(t)\,dt| \le \int_x^y |f_n'(t)|\,dt \le \int_x^y \frac{1+|\ln t|}{\sqrt t}\, dt < \epsilon$$

for every $n.$ This shows the $f_n$ are equicontinuous on $[0,1].$ By Arzela-Ascoli, we will have the desired conclusion if we show the $f_n$ are uniformly bounded.

Now the given condition on the integrals implies that for each $n,$ there is $x_n\in [0,1]$ such that $|f(x_n)|\le 10.$ Thus for each $n$ we have

$$|f_n(x)| = |f_n(x_n) + \int_{x_n}^x f_n'(t)\,dt\,|\le |f_n(x_n)| + \int_{x_n}^x |f_n'(t)|\,dt$$ $$ \le 10 + \int_{x_n}^x \frac{1+|\ln t|}{\sqrt t}\, dt \le 10 + \int_{0}^1 \frac{1+|\ln t|}{\sqrt t}\, dt <\infty. $$

This gives uniform boundedness and we're done.

| cite | improve this answer | |

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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