I'm doing an exercise in Lebesgue integral.

Definitions probably used:

Some of them can be acquired here: Questions of an exercise in Lebesgue integral.

Definition of Convergence in measure: Suppose $f(x), f_1(x), f_2(x),..., f_k(x), ...$ are Lebesgue measurable functions that are finite almost everywhere ($m(\{x \in E: |f_k(x)| = +\infty\}) = 0$). $\{f_k(x)\}$ converge in measure to $f(x)$ if for $\forall \epsilon > 0,$ $\lim_{k \to \infty} m(\{ x\in E: |f_k(x) - f(x)| \ge \epsilon\}) = 0$.


Plese note it is based on Lebesgue integral.

Suppose $0 \le f_1(x) \le f_2(x) \le ... \le f_k(x) \le ...(x \in E)$ and $f_k(x)$ converges in measurable to $f(x)$, then show $$\lim_{k \to \infty} \int_{E} f_k(x) dx = \int_{E} f(x) dx.$$

My trial and trouble:

There is no solution for this exercise. I tried to use Riesz's theorem of convergence in measure that is,

suppose $\{f_k(x)\}$ converge in measure to $f(x)$, then there exist a subsequence of $\{f_k(x)\}$, named $\{f_{k_i}(x)\}$ such that $\lim f_{k_i}(x) = f(x),$ almost everywhere, $ x \in E.$

I can prove $\lim_{i \to \infty} \int_{E} f_{k_i}(x) dx = \int_{E} f(x) dx.$ However, I don't know how to show $\lim_{k \to \infty} \int_{E} f_k(x) dx = \lim_{i \to \infty} \int_{E} f_{k_i}(x) dx.$ The trouble is I can take $\lim$ into $\int_{E} f_{k_i}(x)$ by monotone convergence theorem that is $$\lim_{i \to \infty} \int_{E} f_{k_i}(x) dx = \int_{E} \lim_{i \to \infty} f_{k_i}(x) dx.$$ But I can't do that for $\lim_{k \to \infty} \int_{E} f_k(x) dx$ coz $\lim_{k \to \infty} f_k(x)$ may not exist.

  • $\begingroup$ a monotone sequence of reals numbers has a accumulation point at most $\endgroup$ – user251257 Aug 8 '15 at 15:51
  • 1
    $\begingroup$ For almost every $x\in E$, you have $f_{k_{i}}(x)\rightarrow f(x)$. $(f_{k}(x))$ is a monotone increasing sequence of real numbers, and $(f_{k_{i}}(x))$ converges to $f(x)$, and thus is bounded above by $f(x)$. It follows that $f(x)$ is also an upper bound for $(f_{k}(x))$, and so $(f_{k}(x))\rightarrow f(x)$. $\endgroup$ – catfish Aug 8 '15 at 16:06
  • 1
    $\begingroup$ Yup, since $(f_{k_{i}}(x))\rightarrow f(x)$, $f(x)$ is a least upper bound for $(f_{{k}_{i}}(x))$, so it will have to be a least upper bound for $(f_{k}(x))$. $\endgroup$ – catfish Aug 8 '15 at 16:14
  • 1
    $\begingroup$ In general, any convergent sequence in a metric space is bounded, so if $(f_{k_{i}}(x))$ consists of real numbers and converges, it is bounded in $\mathbb{R}$. It might be a little trickier here if we're working with the extended real line, $[-\infty,\infty]$ which is not a metric space on its own (it is homeomorphic to one however), but we're given that each $f_{k}$ is finite almost everywhere, and countable unions of measure 0 sets still have measure 0, so I think we should still be able to conclude via the monotone convergence theorem. Have you been able to solve it yet? $\endgroup$ – catfish Aug 8 '15 at 18:27
  • 1
    $\begingroup$ I mean as topological spaces; both $[-\infty,\infty]$ and $[0,1]$ are topological spaces under the order topology, and $[-\infty,\infty]$ is homeomorphic to $[0,1]$. $[-\infty,\infty]$ is not a metric space but $[0,1]$ is (and the standard metric on $[0,1]$ induces the same topology as the order on $[0,1]$ does). This would allow us to consider the image of a sequence in $[-\infty,\infty]$ by the homemorphism which would then be a sequence in $[0,1]$ with the same topological properties. $\endgroup$ – catfish Aug 8 '15 at 19:26

For each $x\in E$ we have $\lim_{i\to\infty } f_{k_i} (x) =\lim_{k\to\infty} f_k (x)$ since $(f_n)$ is monotone. Analogously for the monotone sequence $a_k =\int_E f_k (t) dt $ we get $\lim_{i\to\infty} a_{k_i} =\lim_{k\to \infty } a_k .$


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.