Background:
Let (Ω,Σ,μ) be a σ-finite measure space, and X be a Banach space.
A function f:Ω→X is simple if it assumes only finitely many values. That is, there exists subsets E1,...,En of Ω and scalars x1,...,xn∈X such that f=∑ni=1χEixi.
If the sets Ei can be chosen from Σ, then f is μ-measurable simple.
A function f:Ω→X is measurable if it is the limit of a sequence of μ-measurable simple functions (almost everywhere).
Problem Statement:
A function f:Ω→X is measurable ⇔ χEf is measurable for all E∈Σ such that μ(E)<∞.
Partial Solution:
The (⇒) direction is easy. Write f=lim for \mu-measurable simple f_{n}. Then \chi_{E}f_{n} is \mu-measurable simple and converges \chi_{E}f almost everywhere.
The (\Leftarrow) direction is giving me problems. I thought at first to let (as in the definition of \sigma-finite) (E_{n})_{n=1}^{\infty} be a sequence in \Sigma such that \mu(E_{n}) < \infty and \Omega = \cup_{n=1}^{\infty}E_{n}, and defining
F_{n} = \cup_{j=1}^{n}E_{j}. Then for each n\geq 1, take a sequence (g^{n}_{m})_{m=1}^{\infty} of \mu-measurable simple functions which converges to \chi_{F_{n}}f, by assumption. I thought I might be able to show that g_{n}^{n}\to f almost everywhere, but I couldn't figure out the trick. My idea is below.
I claim that \{\lambda\in \Omega : g^{n}_{n}(\lambda)\not\to f(\lambda)\}\subset\bigcup_{n=1}^{\infty}\{\lambda\in\Omega : g^{n}_{m}(\lambda)\not\to \chi_{F_{n}}(\lambda)f(\lambda)\}, the latter set being \mu-measure 0.
Is am I on the right track? I can't prove my claim.
Answer
A diagonalization argument would work, we just need to take care how to approximate our functions. For ease of notation, define
\begin{align}H_1 &= F_1\\ H_{n+1} &= F_{n+1}\setminus F_n\end{align}
so that the H_n's are pairwise disjoint and \bigcup_{k=1}^n H_k = F_n.
For all n\geq 1 we can take measurable simple functions \{g^{(n)}_m\}_{m\geq 1} such that g^{(n)}_m\mathop{\longrightarrow}_{m\to\infty} f^{(n)} := \chi_{H_n}f. W.l.o.g. we can assume that g^{(n)}_m is supported in H_n.
Define the functions
h_n := \sum_{k=1}^n g^{(k)}_n.
Clearly, for all n\geq 1 we have \chi_{F_n}h_m\mathop{\longrightarrow}_{m\to\infty} \sum_{k=1}^n f^{(k)}=\chi_{F_n} f almost everywhere (since for every sufficiently large m we have \chi_{F_n}h_m = \sum_{k=1}^n g^{(k)}_m), and since \Omega=\bigcup F_n we're done.
No comments:
Post a Comment