Lusin's Theorem: Difference between revisions

From Optimal Transport Wiki
Jump to navigation Jump to search
No edit summary
No edit summary
 
(One intermediate revision by the same user not shown)
Line 1: Line 1:


==Introduction==
==Introduction==
Lusin's Theorem formalizes the measure-theoretic principle that pointwise convergence is "nearly" uniformly convergent. This is the second of Littlewood's famed three principles of measure theory, which he elaborated in his 1944 work "Lectures on the Theory of Functions"[1] as
Lusin's Theorem formalizes the fundamental measure-theoretic principle that every measurable function is "nearly" continuous. This is the second of Littlewood's famed three principles of measure theory, which he elaborated in his 1944 work "Lectures on the Theory of Functions"[1] as


"There are three principles, roughly expressible in the following terms: Every (measurable) set is nearly a finite sum of intervals; every function (of class Lp) is nearly continuous; every convergent sequence of functions is nearly uniformly convergent."  
"There are three principles, roughly expressible in the following terms: Every (measurable) set is nearly a finite sum of intervals; every function (of class Lp) is nearly continuous; every convergent sequence of functions is nearly uniformly convergent."  
Line 8: Line 8:


==Statement==
==Statement==
Let <math> (\mathbb{R}, \mathcal{M}_{\lambda}, \lambda) </math> be the Lebesque measure space on <math> \mathbb{R} </math>, and <math> E </math> a measurable subset of <math> \mathbb{R} </math> satisfying <math> \lambda(E)<+\infty. </math> Let <math> f: E\to \mathbb{R} </math> a be a measurable real-valued function on <math> E </math>. For all <math> \varepsilon>0, </math> there exists a compact set <math> K\subseteq E </math> such that the restriction of <math> f </math> to <math> K </math> is continuous.
Let <math> (\mathbb{R}, \mathcal{M}_{\lambda}, \lambda) </math> be the Lebesque measure space on <math> \mathbb{R} </math>, and <math> E </math> a measurable subset of <math> \mathbb{R} </math> satisfying <math> \lambda(E)<+\infty. </math> Let <math> f: E\to \mathbb{R} </math> be a measurable real-valued function on <math> E </math>. Then for all <math> \varepsilon>0, </math> there exists a compact set <math> K\subseteq E </math> such that the restriction of <math> f </math> to <math> K </math> is continuous.


==Proof==
==Proof==
Line 15: Line 15:
We first prove a lemma: for all <math> \delta>0 </math> and measurable sets <math> B </math>, there exists a compact set <math> F\subseteq B </math> such that <math> \lambda(B\setminus F)<\delta. </math>  
We first prove a lemma: for all <math> \delta>0 </math> and measurable sets <math> B </math>, there exists a compact set <math> F\subseteq B </math> such that <math> \lambda(B\setminus F)<\delta. </math>  


From a standard theorem, we know already the existence of a closed set <math> V\subseteq E </math>
By a standard theorem in measure theory, there exists a closed set <math> V\subseteq E </math>
with <math> \lambda(E\setminus V)<\delta/2. </math> The sequence of measurable sets defined by <math> V_n = V\cap [-n,n]</math> is nested and increasing, and satisfies <math> \bigcup_{n=1}^\infty V_n=V. </math> By continuity of the measure from below, there exists <math> N\in \mathbb{N} </math> such that <math> \lambda(B\setminus [-n,n])<\delta /2.</math> Then <math> F:=V\cap [-N,N]</math> is a closed subset of the compact space <math> [-N,N], </math> and is hence compact. Masurability of the involved sets implies the desired inequality <math> \lambda(E\setminus K)<\delta. </math>
such that <math> \lambda(E\setminus V)<\delta/2. </math> The sequence of measurable sets defined by <math> V_n = V\cap [-n,n]</math> is nested and increasing, and satisfies <math> \bigcup_{n=1}^\infty V_n=V. </math> By continuity of the measure from below, there hence exists <math> N\in \mathbb{N} </math> such that <math> \lambda(B\setminus [-n,n])<\delta /2.</math> Then <math> F:=V\cap [-N,N]</math> is a closed subset of the compact space <math> [-N,N], </math> and is consequently compact. Measurability of the relevant sets implies the desired inequality <math> \lambda(E\setminus K)<\delta. </math>


We now proceed with the main proof. Let <math>\varepsilon>0 </math> be given. By cardinality considerations, we may enumerate the collection of open intervals in <math>\mathbb{R} </math> with rational endpoints, say by <math> \{R_n\}_{n\in \mathbb{N}. </math> By the measurability of <math> f </math> and the closure of the <math> \sigma</math>-algebra under complements and countable intersections, for each <math> n\in \mathbb{N} </math>both <math> f^{-1}(R_n)</math> and <math> E\setminus f^{-1}(R_n) </math> are measurable. From the lemma, there exist compact sets <math> K_n \subseteq f^{-1}(R_n) </math> and <math> K_n'\subseteq E\setminus f^{-1}(R_n)  </math> satisfying <math> \lambda(f^{-1}(R_n)\setminus (K_n\cup K_n')<\varepsilon/2^n </math>.  
We now proceed with the main proof. Let <math>\varepsilon>0 </math> be given. By cardinality considerations, we may enumerate the collection of open intervals in <math>\mathbb{R} </math> with rational endpoints, say by <math> \{R_n\}_{n\in \mathbb{N}} .</math> By the measurability of <math> f </math> and the closure of the <math> \sigma</math>-algebra under complements and countable intersections, for each <math> n\in \mathbb{N} </math>, both <math> f^{-1}(R_n)</math> and <math> E\setminus f^{-1}(R_n) </math> are measurable. From the lemma, there exist compact sets <math> K_n \subseteq f^{-1}(R_n) </math> and <math> K_n'\subseteq E\setminus f^{-1}(R_n)  </math> satisfying <math> \lambda(f^{-1}(R_n)\setminus (K_n\cup K_n'))<\varepsilon/2^n </math>.  


Note that a finite union of compact sets remains compact, and an arbitrary intersection of compact sets remains compact. Define <math> K=\bigcap_{n=1}^{\infty} (K_n\cup K_n'), </math> so that <math> K </math> is compact and satisfies <math> \lambda(E\setminus K)<\varepsilon.</math>  
Note that a finite union of compact sets, as well as an arbitrary intersection of compact sets, remains compact. Define <math> K=\bigcap_{n=1}^{\infty} (K_n\cup K_n'), </math> so that <math> K </math> is compact and satisfies <math> \lambda(E\setminus K)<\varepsilon.</math>  


Let <math> x\in K</math>, and by density of the rationals and the axiom of choice fix an <math> n\in \mathbb{N} </math> such that <math> x\in R_n.</math> Then <math> x\in \mathcal{O}:=\mathbb{R}\setminus K_n' </math> and f(\mathcal{O}\cap K)\subseteq R_n. From the definition, this implies that the restriction of <math> f </math> to <math> K </math> is continuous.
Let <math> x\in K</math>, and by density of the rationals fix an <math> n\in \mathbb{N} </math> such that <math> x\in R_n.</math> Then <math> x\in \mathcal{O}:=\mathbb{R}\setminus K_n' </math> and <math> f(\mathcal{O}\cap K)\subseteq R_n.</math> From the definition, this implies that the restriction of <math> f </math> to <math> K </math> is continuous.


==References==
==References==
[1] Littlewood, J. E. "Lectures on the Theory of Functions." Oxford University Press. 1944.
[1] Littlewood, J. E. "Lectures on the Theory of Functions." Oxford University Press. 1944.
[2] Talvila, Erik; Loeb, Peter. "Lusin's Theorem and Bochner Integration." arXiv. 2004. https://arxiv.org/abs/math/0406370
[2] Talvila, Erik; Loeb, Peter. "Lusin's Theorem and Bochner Integration." arXiv. 2004. https://arxiv.org/abs/math/0406370
[3] https://conf.math.illinois.edu/~mjunge/54004-lusin.pdf
[3] https://conf.math.illinois.edu/~mjunge/54004-lusin.pdf

Latest revision as of 02:22, 18 December 2020

Introduction

Lusin's Theorem formalizes the fundamental measure-theoretic principle that every measurable function is "nearly" continuous. This is the second of Littlewood's famed three principles of measure theory, which he elaborated in his 1944 work "Lectures on the Theory of Functions"[1] as

"There are three principles, roughly expressible in the following terms: Every (measurable) set is nearly a finite sum of intervals; every function (of class Lp) is nearly continuous; every convergent sequence of functions is nearly uniformly convergent."

Lusin's theorem is hence a key tool in working with measurable functions, often enabling one to reduce measurable, yet intractible functions to the consideration of a continuous approximation.

Statement

Let be the Lebesque measure space on , and a measurable subset of satisfying Let be a measurable real-valued function on . Then for all there exists a compact set such that the restriction of to is continuous.

Proof

We present an adaptation of the proofs found in references [2] and [3].

We first prove a lemma: for all and measurable sets , there exists a compact set such that

By a standard theorem in measure theory, there exists a closed set such that The sequence of measurable sets defined by is nested and increasing, and satisfies By continuity of the measure from below, there hence exists such that Then is a closed subset of the compact space and is consequently compact. Measurability of the relevant sets implies the desired inequality

We now proceed with the main proof. Let be given. By cardinality considerations, we may enumerate the collection of open intervals in with rational endpoints, say by By the measurability of and the closure of the -algebra under complements and countable intersections, for each , both and are measurable. From the lemma, there exist compact sets and satisfying .

Note that a finite union of compact sets, as well as an arbitrary intersection of compact sets, remains compact. Define so that is compact and satisfies

Let , and by density of the rationals fix an such that Then and From the definition, this implies that the restriction of to is continuous.

References

[1] Littlewood, J. E. "Lectures on the Theory of Functions." Oxford University Press. 1944.

[2] Talvila, Erik; Loeb, Peter. "Lusin's Theorem and Bochner Integration." arXiv. 2004. https://arxiv.org/abs/math/0406370

[3] https://conf.math.illinois.edu/~mjunge/54004-lusin.pdf