0
$\begingroup$

Let $(X, \mathcal{A}, \mu)$ be a complete, $\sigma$-finite measure space and $(E,|\cdot|)$ a Banach space.

  • We say $f \in E^{X}$ is $\boldsymbol{\mu}\textbf{-simple}$ if $f(X)$ is finite, $f^{-1}(e) \in \mathcal{A}$ for every $e \in E,$ and $\mu\left(f^{-1}(E \backslash\{0\})\right)<\infty$.

  • Suppose $f_n, f \in E^{X}$ for $n \in \mathbb{N} .$ Then $(f_n)_{n \in \mathbb N}$ converges to $f$ $\boldsymbol{\mu}\textbf{-almost everywhere}$ if and only if there is a $\mu$-null set $N$ such that $f_{n}(x) \rightarrow f(x)$ for all $x \in N^{c}$.

  • In the theory of integration, it is useful to consider not only real-valued functions but also maps into the extended number line $\overline{\mathbb{R}}$. Such maps are called $\overline{\mathbb{R}}$-valued functions.

  • An $\overline{\mathbb{R}}$-valued function $f: X \rightarrow \overline{\mathbb{R}}$ is said to be $\boldsymbol{\mu}\textbf{-measurable}$ if $\mathcal{A}$ contains $f^{-1}(-\infty), f^{-1}(\infty)$, and $f^{-1}(O)$ for every open subset $O$ of $\mathbb{R}$.


After dicussing with @Thorgott, I came up with the following theorem. I've tried but to no avail. Could you please leave me some hints to finish it?

Theorem $f: X \to \overline{\mathbb{R}}$ is $\mu$-measurable if and only if there is a sequence of $\mu$-simple functions $f_n: X \to \mathbb R$ such that $f_n \to f$ $\mu$-almost everywhere.

$\endgroup$
4
  • $\begingroup$ Do you have ideas for either direction? Or would you like help with both? What are some things you've tried? Also - which theorems of measure theory are you familiar with? $\endgroup$ – HallaSurvivor Jan 26 '20 at 19:46
  • 1
    $\begingroup$ One direction is trivial. The other follows by splitting $f$ into a finite valued part and an indicator on the set where $f$ is $+\infty$ or $-\infty$. It is straightforward to construct a simple function converging to the latter ($n 1_{A_\infty \cap [-n,n]} $ for example if the space is $\mathbb{R}$, use the $\sigma$ finite part to do for general Banach space). $\endgroup$ – copper.hat Jan 26 '20 at 20:08
  • $\begingroup$ Thank you so much @copper.hat! I will try your suggestion. $\endgroup$ – LE Anh Dung Jan 26 '20 at 20:20
  • $\begingroup$ Hi @copper.hat! From your hints, I've figured a proof and posted it as an answer below. I'm not sure if the part I prove $f^{-1}(+\infty) \in \mathcal{A}$ (at the end of my proof) is correct or not. Could you please verify it? $\endgroup$ – LE Anh Dung Jan 30 '20 at 16:02
0
$\begingroup$

From @copper.hat's hints and my textbook, I've figured out a proof. I would be great if someone helps verify it. Thank you so much!


$\textbf{My attempt}$

$\Longrightarrow$

(i) We consider first the case $\mu(X)<\infty$. Let $(a_k)_{k \in \mathbb N}$ be an enumeration of $\mathbb Q$ and $A_{k,n} = f^{-1} [ \mathbb B (a_k, 1/(n+1))]$. Let $A_{+} = f^{-1}(+\infty)$ and $A_{-} = f^{-1}(-\infty)$. Then $\{A_{k,n},A_{+},A_{-}\} \subseteq \mathcal A$ for all $(k,n) \in \mathbb N^2$. The continuity of $\mu$ from above and the assumption $\mu(X)<\infty$ implies there are $m_n$ and $B_n \in \mathcal A$ such that $$B^c_n =A_+ \cup A_- \cup \bigcup_{k=0}^{m_n} A_{k,n} \quad \text{and} \quad \mu(B_n) < \frac{1}{2^{n+1}}$$

Now define $\varphi_{n} \in {\mathbb R}^{X}$ by $$\varphi_{n}(x) = \begin{cases} {a_{0}} & {\text {if} \quad x \in A_{0,n}} \\ {a_{k}} & {\text {if} \quad x \in A_{k, n} \setminus \bigcup_{p=0}^{k-1} A_{p, n} \quad \text {for} \quad 1 \le p \leq m_{n}} \\ n & {\text {if}} \quad x \in A_{+} \\ -n & {\text {if}} \quad x \in A_{-} \\ {0} & {\text {otherwise}} \end{cases}$$

Clearly, $\varphi_{n}$ is $\mu$-simple and $\|\varphi_{n}(x) - f(x)\| < 1/(n+1)$ for all $x \in B_n^c$. Define a decreasing sequence $(C_n)_{n \in \mathbb N}$ by $C_n = \bigcup_{p=0}^{\infty} B_{n+p}$. Then $C_n^c \subseteq B_n^c$ and $\mu(C_n) \le \sum_{p=0}^\infty \mu(B_{n+p}) < 1/2^n$. It therefore follows from the continuity of $\mu$ from above that $C = \bigcap_{n=0}^\infty C_{n}$ is $\mu$-null. We now set $$\psi_{n}(x) = \begin{cases} {\varphi_{n}(x)} & {\text {if} \quad x \in C_{n}^{c}} \\ n & {\text {if}} \quad x \in A_{+} \\ -n & {\text {if}} \quad x \in A_{-} \\ {0} & {\text {otherwise}}\end{cases}$$

Clearly, $\psi_{n}$ is $\mu$-simple. For $x \in C^c$, there exists $n \in \mathbb N$ such that $x \in C_n^c$. Then $x \in C_{n+p}^c$ for all $p \in \mathbb N$. So $\|\psi_{n+p}(x) - f(x)\| =\|\varphi_{n+p}(x) - f(x)\| < 1/(n+p+1)$ for all $p \in \mathbb N$. Hence $\psi_{n} (x) \to f(x)$ for all $x \in A_+ \cup A_- \cup C^c$.

(ii) We next consider the case $\mu(X)=\infty$. Because $\mu$ is $\sigma$-finite, there is a sequence $(A_k)_{k \in \mathbb N}$ of pairwise disjoint subsets in $\mathcal{A}$ such that $\bigcup_{k=0}^\infty A_{k}=X$ and $\mu (A_{k}) < \infty$. As in (i), for each $A_k$, there is a sequence $(\psi^k_{n})_{n \in \mathbb N}$ of $\mu$-simple functions and a $\mu$-null set $C_k$ such that $\psi^k_{n} (x) \to f(x)$ for all $x \in A_k \setminus C_k$. Moreover, $C=\bigcup_{k=0}^\infty C_{k}$ is $\mu$-null. We define a sequence $(\psi_{n})_{n \in \mathbb N}$ by $$\psi_{n}(x) = \begin{cases} {\psi^k_{n}(x)} & {\text {if} \quad x \in \bigcup_{k=0}^n A_{k}} \\ {0} & {\text {otherwise}}\end{cases}$$

Clearly, $(\psi_{n})_{n \in \mathbb N}$ is a sequence of $\mu$-simple functions such that $\psi_{n} (x) \to f(x)$ for all $x \in \bigcap_{k=0}^\infty C_k^c$.

$\Longleftarrow$

Assume there exist a sequence $(\psi_{n})_{n \in \mathbb N}$ of $\mu$-simple functions and a $\mu$-null set $N$ such that $\psi_{n} (x) \to f(x)$ for all $x \in N^c$.

Let $O$ be open in $\mathbb R$. We define a sequence $(O_k)_{k \in \mathbb N^*}$ by $O_{k} = \{y \in O \mid d(y, O^{c})>1 / k \}$. Then $O_{k}$ is open and $\overline{O}_{k} \subseteq O$. Let $x \in N^{c}$. We have $x \in O \iff \exists k \in \mathbb{N}^{*}: x \in O_k$. Therefore, $f(x) \in O$ if and only if there exists $(k,m_k) \in \mathbb{N}^{*} \times \mathbb{N}$ such that $\forall n \ge m_k: \varphi_{n}(x) \in O_{k}$. Consequently, $x \in f^{-1}(O)$ if and only if there exists $(k,m_k) \in \mathbb{N}^{*} \times \mathbb{N}$ such that $\forall n \ge m_k: x \in \varphi^{-1}_{n}(O_{k})$. As a result, $$f^{-1}(O) \cap N^{c} = \left ( \bigcup_{(k,m_k) \in \mathbb{N}^{*} \times \mathbb{N}} \bigcap_{n \ge m_k} \varphi_{n}^{-1} (O_{k}) \right ) \cap N^{c} = \bigcup_{(k,m_k) \in \mathbb{N}^{*} \times \mathbb{N}} \bigcap_{n \ge m_k} \left ( \varphi_{n}^{-1} (O_{k}) \cap N^{c}\right )$$

Because $\varphi_{n}$ is $\mu$-simple, $\varphi_{n}^{-1}(O_{k}) \in \mathcal{A}$ for all $(n,k) \in \mathbb{N} \times \mathbb{N}^{*}$. Hence $f^{-1}(O) \cap N^{c} \in \mathcal{A}$. Furthermore, the completeness of $\mu$ implies $f^{-1}(O) \cap N$ is a $\mu$-null set. Altogether, we obtain $$f^{-1}(O)=\left(f^{-1}(O) \cap N\right) \cup\left(f^{-1}(O) \cap N^{c}\right) \in \mathcal{A}$$

Let $x \in N^c$. We have $f(x) = +\infty \iff \forall M \in \mathbb N, \exists N \in \mathbb N,\forall n \ge N: \varphi_n(x) \ge M$. Consequently, $x \in f^{-1}(+\infty) \iff \forall M \in \mathbb N, \exists N \in \mathbb N,\forall n \ge N: x \in \varphi^{-1}_n ([M, \infty))$. As a result, $$\begin{aligned} f^{-1}(+\infty) \cap N^c &= \left( \bigcap_{M=0}^\infty \bigcup_{N=0}^\infty \bigcap_{n=N}^\infty \varphi^{-1}_n ([M, \infty)) \right) \cap N^c \\ &= \bigcap_{M=0}^\infty \bigcup_{N=0}^\infty \bigcap_{n=N}^\infty \left( \varphi^{-1}_n ([M, \infty)) \cap N^c \right) \end{aligned}$$

Because $\varphi_{n}$ is $\mu$-simple, $\varphi_{n}^{-1}([M, \infty)) \in \mathcal{A}$ for all $(n,M) \in \mathbb{N} \times \mathbb{N}$. Hence $f^{-1}(+\infty) \cap N^{c} \in \mathcal{A}$. Furthermore, the completeness of $\mu$ implies $f^{-1}(+\infty) \cap N$ is a $\mu$-null set. Altogether, we obtain $$f^{-1}(+\infty)=\left(f^{-1}(+\infty) \cap N\right) \cup\left(f^{-1}(+\infty) \cap N^{c}\right) \in \mathcal{A}$$

With similar reasoning, we have $f^{-1}(-\infty) \in \mathcal{A}$.

$\endgroup$

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.