6

Prove that to each $\epsilon > 0$ there exists a $\delta > 0$ such that $\displaystyle \int_E |f| d \mu < \epsilon$ whenever $\mu (E) < \delta$, where $f \in L^1(\mu)$.

I found this question asked on MSE here, here, and here, but some of the suggestions seem relatively complicated (at least compared to what I did). When I was solving the problem, no such ideas/theorems found in those links came to my mind; indeed, I just followed my nose by playing with inequalities. But, after looking at some other solutions, I am beginning to doubt that mine is correct. Here is what I did:

First note that by definition

\begin{align} \int_{E} |f|d \mu &= \sup \{\sum_{i=1}^m a_i \mu (A_i \cap E) \mid m \in \Bbb{N}, 0 \le a_i \le |f|, E = \bigsqcup_{i=1}^m A_i, A_i \mbox{measurable} \} \\ &\le \sup\{\max \{a_i\} \sum_{i=1}^m \mu(A_i \cap E) \mid ... \} \\ &= \sup\{\max \{a_i\} \mu(E) \mid ... \} \\ &= \mu(E) \cdot \sup\{\max \{a_i\} \mid ... \} \\ &\le \mu(E) \inf_{x \in E} |f(x)| \\ \end{align}

The first inequality follows simply because the elements in the one set are larger than the other. The second because $0 \le a_i \le |f|$ on $E$ implies $a_i \le \inf_{x \in E} |f(x)|$ and therefore $\max \{a_i \} \le \inf_{x \in E} |f(x)|$. Now, if $\inf_{x \in E} |f(x)| = 0$, then any $\delta > 0$ will do, since the above together with this implies $\int_{E} |f| d \mu =0$. If not, then $\displaystyle \delta = \frac{\epsilon}{\inf_{x \in X} |f(x)|}$ will work.

I can't spot the error.

user193319
  • 7,940
  • This should be “$0 ≤ a_i ≤ \lvert f|_{A_i} \rvert$”, right? – k.stm Jan 07 '18 at 19:25
  • @k.stm Hmm...I think so, but I'm not sure. – user193319 Jan 07 '18 at 19:27
  • It should be $\sup$ for the smallest upper bound, and you can have $\sup |f|=\infty$, that is, your argument is only correct if $f$ is essentially bounded.. – Lutz Lehmann Jan 07 '18 at 19:28
  • @LutzL Smallest upper bound? Where exactly in my proof are you pointing to? From my understanding, $\inf_{x \in E} |f(x)|$ is larger than every $a_i$. Why would I want to substitute in $\sup |f|$? – user193319 Jan 07 '18 at 19:31
  • Consider $f(x)=1/x^2$ and $X=(0,1)$ – Matematleta Jan 07 '18 at 19:36
  • The infimum is the largest lower bound, the supremum the smallest upper bound, thus $\inf |f|\le |f(x)|\le\sup|f|$ over the selected domain. $\inf |f|$ could very well be zero for a non-zero function. This might even be true for the essential bounds. – Lutz Lehmann Jan 07 '18 at 19:36
  • @chilangoincomprendido Isn't such a function ruled out by the stipulation that $f \in L^{1}(\mu)$? – user193319 Jan 07 '18 at 19:38
  • @LutzL I'm not sure what you are getting at. If $\inf |f|$ is zero, then we are done as I noted in my proof; in that case we can choose any $\delta > 0$. – user193319 Jan 07 '18 at 19:46
  • I think the conclusion at the end is wrong. You can only conclude that, for a fixed decomposition $A_1, \ldots, A_m$, $\sum_{i=1}^m a_i \mu(A_i \cap E) \leq \mu(E) \inf_{x \in A_k} |f(x)|$ for some $k$.

    Take, for instance, $f(x) = x$ on $(0,1)$ and $E = (0,1)$ with the Lebesgue measure as a counterexample.

    – Pedro M. Jan 07 '18 at 19:50
  • No, you have just mixed the meanings of infimum and supremum, check your sources. if the least upper bound of $f$ is zero, then the whole function is zero and the result, indeed, is trivial. – Lutz Lehmann Jan 07 '18 at 19:54
  • How the heck did you come up with that last inequality?? – mathworker21 Jan 07 '18 at 21:31
  • @user193319 it's a typo: I meant $f(x)=1/\sqrt x$ – Matematleta Jan 07 '18 at 22:04

0 Answers0