Measurable space

If $\mathrm{\Omega }$ is a given set, then a $\sigma$-algebra $\mathcal{F}$ on $\mathrm{\Omega }$ is a family $\mathcal{F}$ of subsets (called measurable sets) of $\mathrm{\Omega }$ with the following properties:

1. Empty Set:
   $\mathrm{\varnothing }\in \mathcal{F}$
2. Closed under Complements:
   $F\in \mathcal{F}\Rightarrow F^C\in \mathcal{F}$, where $F^C=\mathrm{\Omega }\setminus F$ is the complement of $F$ in $\mathrm{\Omega }$
3. Closed under Countable Unions:
   $A_1,A_2,\dots \in \mathcal{F}\Rightarrow A:=\bigcup _{i=1}^{\mathrm{\infty }}{A}_i\in \mathcal{F}$

The pair $(\mathrm{\Omega },\mathcal{F})$ is called a measurable space.

Here we can define a measure and obtain a measure space.

Key examples:

• A discrete set and its power set.
• See The Borel sets and the Lebesgue measurable sets.

Motivation

The motivation behind introducing a $\sigma$-algebra lies at the very heart of measure theory and modern probability. To understand why we need it, we have to look at what happens when we try to assign a "size," "length," or "probability" to every possible collection of items in a space.

In short: We introduce $\sigma$-algebras because it is mathematically impossible to assign a meaningful "measure" to every possible subset of a continuous space without breaking the rules of geometry and logic. Here is the breakdown of the motivation, step-by-step.

1. The Goal: We Want to Measure Things

Imagine the real number line, $\mathbb{R}$. We want to create a function—let's call it a measure, $m$—that gives us the "length" of any subset of $\mathbb{R}$. For this measure to make logical sense, we want it to follow three basic, highly intuitive rules: see measure.

2. The Problem: The Paradox of the Power Set

Ideally, we would apply this measure to the power set. For discrete, finite sets (like rolling a standard 6-sided die), measuring the power set works perfectly. However, for infinite, continuous spaces like the real number line ($\mathbb{R}$) or 3D space (${\mathbb{R}}^3$), the power set is too large and chaotic. If you assume you can measure every subset of $\mathbb{R}$ while maintaining the three rules above, you inevitably run into severe mathematical paradoxes.
The most famous examples are:

• The Vitali Set: A specifically constructed subset of the interval $[0,1]$ that, if you try to assign it a length, creates a logical contradiction. Its length mathematically cannot be zero, but it also cannot be greater than zero.
• The Banach-Tarski Paradox: In 3D space, you can take a solid sphere, cut it into a finite number of highly irregular, jagged subsets, and reassemble those exact same subsets to form two solid spheres of the exact same size as the original.

These paradoxes happen because some sets are so fractal-like, scattered, and infinitely complex that the concept of "volume" or "length" completely breaks down. We call these non-measurable sets.

3. The Solution: A VIP Club for Sets

Since we cannot measure everything without breaking mathematics, we must restrict our measure to a smaller domain. We need a "VIP club" of sets that are well-behaved enough to be measured. This collection of measurable sets is exactly what a $\sigma$-algebra is.

Instead of asking, "What is the length of any subset?", we first define a space $(X,\mathrm{\Sigma })$, where $X$ is our universe and $\mathrm{\Sigma }$ (the $\sigma$-algebra) is our collection of measurable sets. We only allow our measure function to look at sets inside $\mathrm{\Sigma }$.

4. Why the Specific Rules of a $\sigma$-algebra?

For this "club" of measurable sets to be practically useful for calculus, limits, and probability, it needs certain structural guarantees. A collection of sets $\mathrm{\Sigma }$ is a $\sigma$-algebra if it satisfies three specific properties. Here is the motivation behind each one:

• Property 1: It contains the whole space ($X\in \mathrm{\Sigma }$).

  • Motivation: We must be able to measure the universe we are working in. In probability, this means the probability of "anything happening at all" is 100%, or $1$. If we can't measure the whole space, the system is useless.

• Property 2: It is closed under complements (If $A\in \mathrm{\Sigma }$, then $A^c\in \mathrm{\Sigma }$).

  • Motivation: If we know the measure of a set, we must mathematically be able to know the measure of its opposite. If we can measure the probability of it raining tomorrow, we must also be able to measure the probability of it not raining.

• Property 3: It is closed under countable unions (If $A_1,A_2,A_3...\in \mathrm{\Sigma }$, then $\bigcup _{i=1}^{\mathrm{\infty }}{A}_i\in \mathrm{\Sigma }$).

  • Motivation: This is the most crucial property and the reason for the "$\sigma$" (which stands for sum, or countable limits, in mathematics). In analysis and probability, we constantly take limits as things go to infinity. If we are measuring a sequence of events, and we combine an infinite (but countable) sequence of them, the resulting combined event must still be measurable. If we only allowed finite unions (an ordinary algebra), we couldn't do modern calculus, limits, or advanced probability.

Intuition (in probability theory)

A $\sigma$-algebra $\mathcal{F}$ is like a "system of holes" or a "viewing window" that you get to look through to see the outcome of a experiment.
Suppose we roll a the die ($\mathrm{\Omega }=\{1,2,3,4,5,6\}$):

• The "Experiment" ($\mathrm{\Omega }$): This is a dark room where the die has been rolled. The outcome (e.g., a "4") is in the room, but you're outside.

• The "Wall": This is a barrier for information between you and the outcome.

• The $\sigma$-Algebra ($\mathcal{F}$): This is the set of holes you're allowed to look through.

Scenario 1: The "Full Info" Wall (${\mathcal{F}}_{\text{full}}$)

• Holes: The wall is made of clear glass. You can see the die perfectly.

• What you can observe: You can answer any question. "Was it a 4?" (Yes). "Was it even?" (Yes). "Was it less than 3?" (No).

• This $\sigma$-algebra is the "power set"—it contains all possible subsets as "viewable" events. Your observation is perfect.

Scenario 2: The "Even/Odd" Wall (${\mathcal{F}}_{\text{even}}$)

• Holes: The wall is solid, but it has two peepholes.

  • One is labeled "ODD" ($\{1,3,5\}$).

  • One is labeled "EVEN" ($\{2,4,6\}$).

• What you can observe: A light flashes in the room. You can't see the die itself, but you can see which peephole the light is coming from.

• Measurable Events ("Questions you can answer"):

  • "Was the roll even?" Yes, you can answer this. If you see the light in the "EVEN" hole, the answer is yes. This event ($\{2,4,6\}$) is measurable with this $\sigma$-algebra.

  • "Was the roll a 4?" No, you cannot answer this. The light in the "EVEN" hole tells you the outcome was in the set $\{2,4,6\}$, but you can't distinguish which of those three it was. The event $\{4\}$ is not measurable with this $\sigma$-algebra.

Scenario 3: The "No Info" Wall (${\mathcal{F}}_{\text{none}}$)

• Holes: The wall is solid. There are no holes.

• What you can observe: You know a die was rolled (this is $\mathrm{\Omega }$, the whole room), but you see nothing.

• Measurable Events: The only question you can answer is "Did something happen?" (Yes, $\mathrm{\Omega }$). This is the trivial $\sigma$-algebra $\{\mathrm{\varnothing },\mathrm{\Omega }\}$.