Caratheodory's Theorem: Difference between revisions

From Optimal Transport Wiki
Jump to navigation Jump to search
No edit summary
(Added an introductory section. Cleaned up the proof a bit.)
 
(2 intermediate revisions by one other user not shown)
Line 1: Line 1:
Let <math> X </math> be a set and <math> \mu^*: 2^X \to [0, \infty] </math> an [[outer measure]] on <math> X </math>, we call a subset <math> A \subseteq  X </math> to be '''<math>\mu^*</math>-measurable''' if
<math> \mu^*(E) = \mu^*(E \cap A) + \mu^*(E \cap A^C) </math>
for all <math>E \in 2^X</math>. This should remind the reader of a full [[measure]] since this is similar to disjoint additivity. Thus, the natural question arises of the significance of these <math>\mu^*-</math>measurable sets. Caratheodory's Theorem provides and answer by proving that if <math>\mathcal{M}</math> is the collection of <math>\mu^*-</math>measurable sets, then <math>(X, \mathcal{M}, \mu^*)</math> is indeed a Measure Space.
== Statement ==
== Statement ==
Consider an outer measure <math> \mu^* </math> on <math> X </math>.


Consider an out measure <math> \mu </math> on <math> X </math>. Define
<math> \mathcal{M} = \{ A \subseteq X : A \ \text{is} \  \mu^*-\text{measurable} \} </math>.
 
<math> \mathcal{M} = \{ A \subseteq X : A \ \text{is} \  \mu-\text{measurable} \} </math>.


Then <math> \mathcal{M} </math> is a <math>\sigma</math>-algebra and <math> \mu^* </math> is a measure on <math> \mathcal{M} </math>.
Then <math> \mathcal{M} </math> is a <math>\sigma</math>-algebra and <math> \mu^* </math> is a measure on <math> \mathcal{M} </math><ref>Craig, Katy. ''MATH 201A Lectures 6,7''. UC Santa Barbara, Fall 2020.</ref>.


== Proof ==
== Proof ==


First, observe that <math>\mathcal{M}</math> is closed under complements due to symmetry in the meaning of <math> \mu </math>-measurability. Now, we show if <math> A, B </math> then <math> A \cup B \in \mathcal{M} </math><ref name="Folland"> Folland, Gerald B., ''Real Analysis: Modern Techniques and Their Applications, second edition'' </ref>.  
First, observe that <math>\mathcal{M}</math> is closed under complements due to symmetry in the meaning of <math> \mu^* </math>-measurability. Now, we show if <math> A, B </math> then <math> A \cup B \in \mathcal{M} </math><ref name="Folland"> Folland, Gerald B., ''Real Analysis: Modern Techniques and Their Applications, second edition'', §1.4 </ref> to conlcude that <math> \mathcal{M} </math> is an [[algebra]].  


Suppose <math>E \subseteq X </math>. Then
Suppose <math>E \subseteq X </math>. Then
Line 25: Line 31:
<math> E =  \mu^*(E \cap (A \cup B)^c) + \mu^*(E \cap (A \cup B) </math>
<math> E =  \mu^*(E \cap (A \cup B)^c) + \mu^*(E \cap (A \cup B) </math>


hence <math> A \cup B \in \mathcal{M} </math> and we have <math> \mathcal{M} </math> is an algebra.
hence <math> A \cup B \in \mathcal{M} </math> and we have shown that <math> \mathcal{M} </math> is an [[algebra]].




Now, suppose <math> A, B \in \mathcal{M} </math> and <math> A, B \in M </math> are disjoint. Then
Now, suppose <math> A, B \in \mathcal{M} </math> are disjoint. Then


<math> \mu^*(A \cup B) = \mu^*((A \cup B) \cap A) + \mu^*((A \cup B) \cap A^c) = \mu^*(A) + \mu^*(B) </math>
<math> \mu^*(A \cup B) = \mu^*((A \cup B) \cap A) + \mu^*((A \cup B) \cap A^c) = \mu^*(A) + \mu^*(B) </math>
Line 47: Line 53:
<math> = \sum_{i=1}^{n} \mu^*( A \cap B_i) + \mu^*(A \cap (\cup_{i=1}^{\infty} B_i)^c) .</math>
<math> = \sum_{i=1}^{n} \mu^*( A \cap B_i) + \mu^*(A \cap (\cup_{i=1}^{\infty} B_i)^c) .</math>


Since this holds for all <math> n </math>,
Since this holds for all <math> n \in \mathbb{N} </math>,


<math> \mu^*(A) \geq  \sum_{i=1}^{\infty}  \mu^*( A \cap B_i) + \mu^*(A \cap (\cup_{i=1}^{\infty} B_i)^c) </math>
<math> \mu^*(A) \geq  \sum_{i=1}^{\infty}  \mu^*( A \cap B_i) + \mu^*(A \cap (\cup_{i=1}^{\infty} B_i)^c) </math>
Line 65: Line 71:
<math> D_1 = C_1, D_2 = C_2 \ C_1, . . . ., D_n = C_n \ (\cup_{i=1}^{n-1} C_i) .</math>
<math> D_1 = C_1, D_2 = C_2 \ C_1, . . . ., D_n = C_n \ (\cup_{i=1}^{n-1} C_i) .</math>


Their unions are equal and <math> \cup_{i=1}^{\infty} D_i </math> is disjoint, hence contained in <math> \mathcal{M} </math>.
Their unions are equal and <math> \cup_{i=1}^{\infty} D_i </math> is disjoint, hence contained in <math> \mathcal{M} </math>, proving it to be a <math>\sigma-</math>algebra.


This finishes all parts to the proof.
This finishes all parts to the proof.


<math> \square </math>
<math> \square </math>

Latest revision as of 21:27, 18 December 2020

Let be a set and an outer measure on , we call a subset to be -measurable if

for all . This should remind the reader of a full measure since this is similar to disjoint additivity. Thus, the natural question arises of the significance of these measurable sets. Caratheodory's Theorem provides and answer by proving that if is the collection of measurable sets, then is indeed a Measure Space.

Statement

Consider an outer measure on .

.

Then is a -algebra and is a measure on [1].

Proof

First, observe that is closed under complements due to symmetry in the meaning of -measurability. Now, we show if then [2] to conlcude that is an algebra.

Suppose . Then

and by subadditivity

But certainly, since the inequality in the other direction also holds, and we conclude

hence and we have shown that is an algebra.


Now, suppose are disjoint. Then

so is finitely additive.


Next, we show is closed under countable disjoint unions. Given a disjoint sequence of sets , for all , by countable subadditivity,

It remains to show the other inequality direction. Since is closed under finite unions, . By using the definition of -measurability,

and by monotonicity,

Since this holds for all ,

which proves

Similarly as before, the other inequality direction is proved by monotonicity, so we conclude equality and we have that .

The only thing that remains to be shown is closure under countable unions. Consider a sequence of sets not necessarily disjoint. Define

Their unions are equal and is disjoint, hence contained in , proving it to be a algebra.

This finishes all parts to the proof.

  1. Craig, Katy. MATH 201A Lectures 6,7. UC Santa Barbara, Fall 2020.
  2. Folland, Gerald B., Real Analysis: Modern Techniques and Their Applications, second edition, §1.4