02.07.01 · analysis / measure-theory

Sigma-algebra, Measurable Space, and the Borel Sigma-algebra

shipped3 tiersLean: partial

Anchor (Master): Halmos, Measure Theory §I; Bogachev, Measure Theory Vol. 1 §1.1-1.4; Rudin, Real and Complex Analysis 3e §1

Intuition Beginner

Ordinary length and area work well for nice shapes. The length of an interval from $2$ to $5$ is $3$ . The area of a rectangle $4$ by $7$ is $28$ . The area of a disk of radius $1$ is $π$ . These numbers behave the way we expect: cutting a shape into two pieces and adding the parts gives back the whole. Moving a shape does not change its size. Bigger shapes have bigger numbers.

The trouble starts when we ask about strange shapes. What is the total length of all rational numbers between $0$ and $1$ ? What is the area of a set that is dense everywhere but contains no rectangle? What if we cut a sphere into a few pieces and rearrange them into two spheres of the same size? That last problem is real, and it shows that length and area cannot be defined for every possible set in a sensible way.

Measure theory steps in to repair this. It picks out which sets are well-behaved enough to have a size, and quietly refuses to assign a size to the pathological ones. The well-behaved collection is called a sigma-algebra.

A sigma-algebra on a set $X$ is a list of subsets of $X$ that you are allowed to measure. The list has three rules. The whole set $X$ is on the list. If a subset is on the list, then its complement is on the list. If you have a countable collection of subsets all on the list, then their union is also on the list. That is it. Those three rules are enough to give a working theory of size.

A pair built from a set $X$ and a sigma-algebra on it is called a measurable space. The sigma-algebra is the rule book; the set is the playing field. Later units will attach an actual number-valued measure to a measurable space, but the rule book has to come first.

On the real line, a special sigma-algebra called the Borel sigma-algebra is the natural choice. It is the smallest sigma-algebra that contains every open interval. Open intervals are the building blocks; closed intervals, half-open intervals, single points, countable unions of intervals, and far more exotic sets can all be assembled from them through the three rules.

The one-sentence takeaway: a sigma-algebra is a careful list of which subsets you trust enough to measure.

Visual Beginner

Two pictures help. The first shows a small set with all of its subsets laid out in a grid, and a few of those subsets coloured to show which ones make up a valid sigma-algebra. The second shows the real line with three open intervals stacked on it, and arrows suggesting that complements, countable unions, and countable intersections of these intervals fill the Borel sigma-algebra.

The grid on the left is finite and easy to enumerate. The line on the right is infinite, but the same three rules apply.

Worked example Beginner

Take the set $X = {1, 2, 3}$ . We want to list every sigma-algebra on $X$ .

Step 1. Find the smallest one. The rules force the empty set and $X$ itself to be present, since the complement of $X$ is the empty set. So ${\emptyset, X}$ is a sigma-algebra. It has $2$ elements.

Step 2. Find the biggest one. The collection of all subsets of $X$ is a sigma-algebra, because every rule is automatically satisfied. The set ${1, 2, 3}$ has $2^{3} = 8$ subsets, so this sigma-algebra has $8$ elements.

Step 3. Find the in-between ones. Pick a single point, say ${1}$ . Once ${1}$ is on the list, its complement ${2, 3}$ must also be on the list. The empty set and $X$ are on the list. That gives four sets: $\emptyset, {1}, {2, 3}, X$ . The same construction with ${2}$ gives a different sigma-algebra of four sets: $\emptyset, {2}, {1, 3}, X$ . And with ${3}$ : $\emptyset, {3}, {1, 2}, X$ .

Step 4. Count. There are five sigma-algebras on ${1, 2, 3}$ : the smallest one, the three middle ones generated by singletons, and the largest one.

What this tells us: even on a set with three points, there are several different valid sigma-algebras. Choosing one is choosing how fine your measurement is allowed to be.

Check your understanding Beginner

Formal definition Intermediate+

Let $X$ be a set. A sigma-algebra on $X$ is a collection $M \subseteq P (X)$ of subsets of $X$ that satisfies the following three axioms ^{[Folland §1.2]}:

$X \in M$ .
If $A \in M$ , then $A^{c} = X ∖ A \in M$ .
If ${A_{n}}_{n \in N}$ is a countable collection with each $A_{n} \in M$ , then $⋃_{n = 1}^{\infty} A_{n} \in M$ .

From these axioms several closure properties follow at once. The empty set $\emptyset = X^{c}$ lies in $M$ . Finite unions are countable unions with all but finitely many terms empty, so $M$ is closed under finite union. By De Morgan, $⋂_{n} A_{n} = (⋃_{n} A_{n}^{c})^{c}$ , so $M$ is closed under countable intersections and hence under finite intersections. Set differences $A ∖ B = A \cap B^{c}$ and symmetric differences $A △ B = (A ∖ B) \cup (B ∖ A)$ are in $M$ whenever $A$ and $B$ are.

A pair $(X, M)$ consisting of a set $X$ together with a sigma-algebra $M$ on $X$ is called a measurable space, and the elements of $M$ are called measurable sets.

Given any family $E \subseteq P (X)$ , the sigma-algebra generated by $E$ , written $σ (E)$ , is the smallest sigma-algebra on $X$ containing $E$ . This is the intersection of all sigma-algebras on $X$ containing $E$ , and that intersection is itself a sigma-algebra (see the Key theorem below).

When $X$ is a topological space 02.01.01 with topology $τ_{X}$ , the Borel sigma-algebra $B (X) = σ (τ_{X})$ is the sigma-algebra generated by the open sets ^{[Rudin §1.11]}. Elements of $B (X)$ are called Borel sets. On the real line $R$ with its standard topology, $B (R)$ contains every open set, every closed set, every countable union of closed sets ( $F_{σ}$ sets), every countable intersection of open sets ( $G_{δ}$ sets), and far more.

A map $f : (X, M) \to (Y, N)$ between measurable spaces is measurable if $f^{- 1} (B) \in M$ for every $B \in N$ . When $Y = R$ with the Borel sigma-algebra, this reduces to requiring $f^{- 1} (U) \in M$ for every open $U \subseteq R$ , by the generating-class argument.

Counterexamples to common slips Intermediate+

Uncountable unions are not allowed. If $M$ is closed under uncountable unions, it has to contain every singleton union, which forces $M = P (X)$ as soon as singletons are in $M$ . The countability bound is load-bearing.
An algebra is not a sigma-algebra. An algebra of sets requires only finite-union closure. The collection of finite unions of half-open intervals $(a, b]$ in $R$ forms an algebra but not a sigma-algebra: the union $⋃_{n} (n, n + 1]$ over all integers $n$ is not a finite union of half-open intervals.
Borel sigma-algebra is not the power set. On $R$ , $∣ B (R) ∣ = 2^{ℵ_{0}}$ (continuum), but $∣ P (R) ∣ = 2^{2^{ℵ_{0}}}$ . Strictly more subsets of $R$ exist than Borel sets exist.
Measurability is about the domain sigma-algebra. The pre-image condition runs from $Y$ into $X$ . A function $f : R \to R$ may be Borel-measurable with respect to one sigma-algebra on the domain and fail to be measurable with respect to another.

Key theorem with proof Intermediate+

Theorem (generated sigma-algebra is well-defined). Let $X$ be a set and let ${M_{α}}_{α \in I}$ be any family of sigma-algebras on $X$ indexed by a set $I$ . Then $M = ⋂_{α \in I} M_{α}$ is a sigma-algebra on $X$ . In particular, given any $E \subseteq P (X)$ , the intersection of all sigma-algebras containing $E$ is the smallest sigma-algebra containing $E$ , and $σ (E)$ is well-defined.

Proof. We check the three axioms for $M$ .

For axiom 1, each $M_{α}$ contains $X$ , so $X \in ⋂_{α} M_{α} = M$ .

For axiom 2, suppose $A \in M$ . Then $A \in M_{α}$ for every $α$ . Each $M_{α}$ is closed under complement, so $A^{c} \in M_{α}$ for every $α$ . Therefore $A^{c} \in ⋂_{α} M_{α} = M$ .

For axiom 3, let ${A_{n}}_{n \in N}$ be a countable family with each $A_{n} \in M$ . For each fixed $α$ and each $n$ , $A_{n} \in M_{α}$ . Since $M_{α}$ is closed under countable union, $⋃_{n} A_{n} \in M_{α}$ . This holds for every $α$ , so $⋃_{n} A_{n} \in ⋂_{α} M_{α} = M$ .

Thus $M$ is a sigma-algebra.

Now fix $E \subseteq P (X)$ . The power set $P (X)$ is a sigma-algebra containing $E$ , so the family ${M_{α} : M_{α} is a sigma-algebra containing E}$ is non-empty. Let $σ (E) = ⋂_{α} M_{α}$ over this family. By the first part, $σ (E)$ is a sigma-algebra. It contains $E$ since every $M_{α}$ does. It is the smallest such sigma-algebra: if $N$ is any sigma-algebra containing $E$ , then $N$ appears in the indexing family, so $σ (E) \subseteq N$ .

$□$

Bridge. The intersection construction here builds toward 02.07.02 (Lebesgue outer measure and the Carathéodory extension theorem), where the same well-defined-generated-object pattern produces the Lebesgue sigma-algebra as the completion of the Borel sigma-algebra under outer-measure null sets. The foundational reason this works is that sigma-algebras are closed under arbitrary intersection, dual to the way topologies are closed under arbitrary union. This is exactly the structure that lets us define a sigma-algebra by what we want to put in rather than by an explicit list. The bridge is between the descriptive theory (Borel hierarchy, generated structure) and the constructive theory (Carathéodory measurability), and the central insight is that both rest on this single closure property. Putting these together with the topology prerequisite 02.01.01, we see why Borel sigma-algebras on a topological space are intrinsic objects: they depend only on the topology, not on a chosen basis.

Exercises Intermediate+

Exercise 3 (medium, symbolic).

Show that the Borel sigma-algebra on $R$ is generated by the family of half-open intervals $E = {(a, b] : a, b \in R, a < b}$ .

Hint

Show inclusions both ways. For one direction, write each half-open interval as a countable intersection of open intervals. For the other, write each open interval as a countable union of half-open intervals.

Answer

Write $σ (E)$ for the sigma-algebra generated by half-open intervals and $B (R)$ for the Borel sigma-algebra.

For $σ (E) \subseteq B (R)$ : each $(a, b] = ⋂_{n = 1}^{\infty} (a, b + 1/ n)$ is a countable intersection of open intervals, hence Borel.

For $B (R) \subseteq σ (E)$ : each open interval $(a, b) = ⋃_{n = 1}^{\infty} (a, b - 1/ n]$ is a countable union of half-open intervals (for $n$ large enough that $b - 1/ n > a$ ). Every open subset of $R$ is a countable union of disjoint open intervals (a standard topological fact), hence lies in $σ (E)$ . Since $σ (E)$ is a sigma-algebra containing the topology, it contains $B (R) = σ (τ)$ by minimality.

Exercise 6 (hard, symbolic).

Let $(X_{1}, M_{1})$ and $(X_{2}, M_{2})$ be measurable spaces. The product sigma-algebra is $M_{1} \otimes M_{2} = σ ({A_{1} \times A_{2} : A_{i} \in M_{i}})$ . Prove that the projection maps $π_{i} : X_{1} \times X_{2} \to X_{i}$ are measurable when the product is equipped with $M_{1} \otimes M_{2}$ , and that $M_{1} \otimes M_{2}$ is the smallest sigma-algebra on $X_{1} \times X_{2}$ making both projections measurable.

Hint

For measurability of projections, compute $π_{i}^{- 1} (A_{i})$ as a measurable rectangle. For minimality, use Exercise 4 to show any sigma-algebra making projections measurable must contain measurable rectangles.

Answer

For each $A_{1} \in M_{1}$ , $π_{1}^{- 1} (A_{1}) = A_{1} \times X_{2}$ , which is a measurable rectangle in $M_{1} \otimes M_{2}$ . By Exercise 4 with the generating family $M_{1}$ , $π_{1}$ is measurable. Symmetrically for $π_{2}$ .

For minimality: suppose $N$ is any sigma-algebra on $X_{1} \times X_{2}$ making both $π_{1}$ and $π_{2}$ measurable. Then $π_{1}^{- 1} (A_{1}) = A_{1} \times X_{2}$ and $π_{2}^{- 1} (A_{2}) = X_{1} \times A_{2}$ lie in $N$ for all measurable $A_{i}$ . The intersection $A_{1} \times A_{2} = (A_{1} \times X_{2}) \cap (X_{1} \times A_{2})$ lies in $N$ . So $N$ contains every measurable rectangle. Since $M_{1} \otimes M_{2}$ is generated by measurable rectangles, $M_{1} \otimes M_{2} \subseteq N$ .

Exercise 7 (hard, symbolic).

Prove that $B (R)$ is generated by the family of rays ${(a, \infty) : a \in Q}$ . Conclude that $∣ B (R) ∣ \leq 2^{ℵ_{0}}$ .

Hint

First show generation by rays with arbitrary real endpoints, then use that every real is a limit of rationals. For the cardinality bound, count by transfinite induction or by encoding each Borel set as a countable tree of countable operations on a countable generating family.

Answer

Generation: each ray $(a, \infty)$ with $a \in R$ is open, hence Borel. For arbitrary $a \in R$ , pick a sequence $a_{n} \in Q$ with $a_{n} ↓ a$ ; then $(a, \infty) = ⋃_{n} (a_{n}, \infty)$ , so the rational-ray family generates every real ray. Half-open intervals $(a, b] = (a, \infty) ∖ (b, \infty) = (a, \infty) \cap (b, \infty)^{c}$ are in the generated sigma-algebra. By Exercise 3, half-open intervals generate $B (R)$ .

Cardinality bound: the rational rays form a countable family of cardinality $ℵ_{0}$ . Build $B (R)$ by transfinite induction: $B_{0} = {(a, \infty) : a \in Q}$ , then for $α < ω_{1}$ set $B_{α}$ = closure under countable union and complement of $⋃_{β < α} B_{β}$ . Each step takes a set of size $\leq 2^{ℵ_{0}}$ to a set of size $\leq 2^{ℵ_{0}}$ . Since $ω_{1}$ has $ℵ_{1} \leq 2^{ℵ_{0}}$ stages, the total cardinality is at most $2^{ℵ_{0}} \cdot ℵ_{1} = 2^{ℵ_{0}}$ .

Exercise 8 (hard, symbolic).

Let $f, g : (X, M) \to R$ be measurable functions where $R$ carries the Borel sigma-algebra. Prove that ${x \in X : f (x) < g (x)}$ is measurable. Then prove $f + g$ is measurable.

Hint

Use that the rationals are dense: $f (x) < g (x)$ iff there exists $q \in Q$ with $f (x) < q < g (x)$ . For $f + g$ , study ${x : f (x) + g (x) < c}$ similarly.

Answer

For the first claim: $f (x) < g (x)$ iff $\exists q \in Q$ with $f (x) < q < g (x)$ . Hence ${x : f (x) < g (x)} = q \in Q ⋃ {x : f (x) < q} \cap {x : g (x) > q} .$ Each ${x : f (x) < q} = f^{- 1} ((- \infty, q))$ and ${x : g (x) > q} = g^{- 1} ((q, \infty))$ is measurable since both rays are Borel and $f, g$ are measurable. A countable union of measurable sets is measurable.

For $f + g$ : ${x : f (x) + g (x) < c} = {x : f (x) < c - g (x)}$ . Apply the first claim with $g$ replaced by the function $c - g$ , which is measurable since $c - g = c \cdot 1 - g$ and measurability is preserved under linear combinations (immediate from the ray-generation criterion). The sets ${x : (f + g) (x) < c}$ for $c \in R$ generate the Borel sigma-algebra on $R$ via the ray characterization, so $f + g$ is measurable.

Lean formalization Intermediate+

lean_status: partial — Mathlib provides the typeclass MeasurableSpace X which encodes the sigma-algebra axioms internally, the predicate MeasurableSet for membership, and borel for the Borel sigma-algebra of a topological space. The companion module records the unit's predicate names; the deeper structural theorems (Borel hierarchy classification, existence of non-measurable sets via choice) remain in the prose, not the Lean module.

import Mathlib.MeasureTheory.MeasurableSpace.Basic
import Mathlib.MeasureTheory.Constructions.BorelSpace.Basic

variable (X : Type*) [MeasurableSpace X]

abbrev IsSigmaAlgebra : Prop := True
-- Mathlib's `MeasurableSpace` typeclass already encodes the σ-algebra
-- axioms (`MeasurableSet.univ`, `MeasurableSet.compl`, `MeasurableSet.iUnion`).

variable (Y : Type*) [TopologicalSpace Y]

abbrev CodexBorel : MeasurableSpace Y := borel Y

Advanced results Master

The advanced theory of sigma-algebras splits into three strands: descriptive set theory (which Borel sets exist?), measure-theoretic completion (what does adding null sets do?), and the role of set-theoretic axioms (what non-measurable sets must exist?).

Theorem 1 (Borel hierarchy). The Borel sigma-algebra $B (R)$ admits a transfinite stratification into classes $Σ_{α}^{0}, Π_{α}^{0}, Δ_{α}^{0}$ for $1 \leq α < ω_{1}$ , where $Σ_{1}^{0}$ is the open sets, $Π_{1}^{0}$ is the closed sets, $Σ_{2}^{0} = F_{σ}$ , $Π_{2}^{0} = G_{δ}$ , and for limit ordinals each class is built from countable unions or intersections of sets in lower-rank classes. Every Borel set sits in some $Σ_{α}^{0}$ with $α < ω_{1}$ , and the hierarchy is strictly increasing in $α$ : each level contains sets not in the previous level ^{[Bogachev §1.1-1.4]}.

The strictness of the hierarchy uses a universal-set construction: there exists a single Borel set $U \subseteq R \times R$ whose sections enumerate all $Σ_{α}^{0}$ sets, and a diagonal argument then shows the next level is strictly bigger. The Borel hierarchy continues into the projective hierarchy $Σ_{n}^{1}, Π_{n}^{1}$ , where $Σ_{1}^{1}$ is the class of analytic (Souslin) sets, characterised as continuous images of Borel sets or projections of Borel subsets of $R^{2}$ .

Theorem 2 (cardinality of $B (R)$ ). $∣ B (R) ∣ = 2^{ℵ_{0}}$ .

The lower bound is immediate from the embedding $R ↪ B (R)$ via $x \mapsto {x}$ (singletons are closed, hence Borel). The upper bound is the transfinite induction from Exercise 7: $ℵ_{1}$ stages of countable operations applied to a countable generating family yield at most $ℵ_{1} \cdot ℵ_{0}^{ℵ_{0}} = 2^{ℵ_{0}}$ sets.

Theorem 3 (existence of non-Borel sets). There exists a subset of $R$ that is not Borel.

Proof outline. By Theorem 2, $∣ B (R) ∣ = 2^{ℵ_{0}}$ but $∣ P (R) ∣ = 2^{2^{ℵ_{0}}}$ . Cantor's theorem gives $2^{ℵ_{0}} < 2^{2^{ℵ_{0}}}$ , so most subsets of $R$ are not Borel. A constructive example: the projection of a closed subset of $R^{2}$ is in general analytic but not Borel (Souslin 1917).

Theorem 4 (Lebesgue sigma-algebra is strictly larger than Borel). Let $L$ denote the sigma-algebra of Lebesgue-measurable subsets of $R$ . Then $B (R) ⊊ L$ and $∣ L ∣ = 2^{2^{ℵ_{0}}}$ . Concretely, $L$ is the completion of $B (R)$ under Lebesgue null sets: $A \in L$ iff $A = B △ N$ for some Borel $B$ and some $N$ contained in a Borel null set ^{[Lebesgue Annali 7 1902]}.

The cardinality jump comes from the Cantor set $C \subseteq R$ : $C$ is Borel and Lebesgue-null, $∣ C ∣ = 2^{ℵ_{0}}$ , and every subset of $C$ is Lebesgue-measurable (with measure zero), so $P (C) \subseteq L$ contributes $2^{2^{ℵ_{0}}}$ sets.

Theorem 5 (Vitali's non-measurable set). Assuming the axiom of choice, there exists a subset $V \subseteq [0, 1]$ that is not Lebesgue-measurable ^{[Vitali 1905]}.

Proof. Define an equivalence relation on $[0, 1]$ by $x \sim y$ iff $x - y \in Q$ . Use the axiom of choice to pick one representative from each equivalence class, forming $V \subseteq [0, 1]$ . The translates $V_{q} = (V + q) \cap [0, 1]$ for $q \in Q \cap [- 1, 1]$ are pairwise disjoint and their union covers $[0, 2]$ with bounded overlap. If $V$ were Lebesgue-measurable with measure $m (V) = 0$ , then countable additivity gives $m ([0, 2]) = 0$ , a contradiction. If $m (V) > 0$ , then $\sum_{q} m (V_{q}) = \infty$ but $\sum_{q} m (V_{q}) \leq m ([0, 2]) = 2$ , also a contradiction. Hence $V$ is not Lebesgue-measurable.

Theorem 6 (composition of measurable and continuous). If $f : (X, M) \to (Y, B (Y))$ is measurable and $g : Y \to Z$ is continuous (where $Y, Z$ are topological spaces with their Borel sigma-algebras), then $g \circ f : (X, M) \to (Z, B (Z))$ is measurable.

Proof. For any open $U \subseteq Z$ , $g^{- 1} (U)$ is open in $Y$ by continuity 02.01.02, hence Borel. So $(g \circ f)^{- 1} (U) = f^{- 1} (g^{- 1} (U)) \in M$ by measurability of $f$ . Open sets generate $B (Z)$ , so $g \circ f$ is measurable by the generating-class criterion of Exercise 4.

Theorem 7 (Solovay's consistency result). The theory ZF + DC (dependent choice) + "every subset of $R$ is Lebesgue-measurable" is consistent, assuming the consistency of ZFC plus an inaccessible cardinal ^{[Solovay 1970]}.

This places the existence of non-measurable sets at the same logical strength as choice principles strictly beyond ZF+DC. Vitali's construction is not avoidable under full choice, but it is unavoidable: any model where every set is measurable must drop the axiom of choice in its standard form.

Theorem 8 (monotone class theorem). Let $A$ be an algebra of subsets of $X$ (closed under finite union and complement but not necessarily countable union). A monotone class $C$ is a family of subsets closed under countable monotone unions ( $⋃_{n} A_{n}$ for $A_{n} ↑$ ) and countable monotone intersections ( $⋂_{n} A_{n}$ for $A_{n} ↓$ ). The monotone class generated by $A$ equals $σ (A)$ ^{[Halmos §6]}.

This theorem is the working tool for proving statements about all measurable sets: prove the statement holds for an algebra, prove it is preserved under monotone limits, conclude it holds on the generated sigma-algebra. It anchors the standard machinery of Fubini's theorem and the uniqueness of extension in the Carathéodory construction.

Synthesis. The descriptive theory of Borel sets, the cardinality calculation, the existence of non-measurable sets, and the monotone class theorem fit together as a single picture. The foundational reason that measure theory needs a sigma-algebra rather than the full power set is that the power set is too large: at cardinality $2^{2^{ℵ_{0}}}$ it admits a Vitali partition that no countably additive measure can survive, and the central insight of Borel and Lebesgue was to identify the largest sub-collection on which countable additivity can be enforced. The Borel hierarchy is exactly the constructive image of this sub-collection: every set we can write down using countable operations on intervals lives somewhere in the hierarchy, and the hierarchy stops at $ω_{1}$ because countable iteration of countable operations closes off there. The Lebesgue completion identifies the Borel sigma-algebra with its enlargement by null sets, and this is exactly the structure that turns Carathéodory's outer-measure construction into a sigma-algebra rather than a mere algebra.

Putting these together with the Solovay consistency result, the picture sharpens: the existence of non-measurable sets is a precise measure of how much choice the ambient set theory uses. The bridge is between the descriptive (which sets are Borel?) and the foundational (which sets exist at all?), and the pattern recurs throughout twentieth-century analysis. The monotone class theorem generalises the inductive proof strategy used for sigma-algebras to a wider class of inductive proofs, and identifies the algebra-of-rectangles approach with the sigma-algebra-of-Borel-sets approach in product spaces, an identification that appears again in 02.07.02 as the rectangular generation of the Lebesgue sigma-algebra and feeds forward into Fubini's theorem.

Full proof set Master

Proposition 1 (Borel sigma-algebra is generated by closed sets). $B (R) = σ ({C \subseteq R : C closed})$ .

Proof. Let $S = σ ({C : C closed})$ . Each closed set $C$ has open complement $C^{c}$ , which is Borel. Since sigma-algebras are closed under complement, $C = (C^{c})^{c} \in B (R)$ . Thus the generating family of $S$ lies in $B (R)$ , so $S \subseteq B (R)$ . Conversely, every open set $U \subseteq R$ has closed complement $U^{c} \in S$ , hence $U = (U^{c})^{c} \in S$ . The open sets generate $B (R)$ , so $B (R) \subseteq S$ .

$□$

Proposition 2 (Borel sigma-algebra is countably generated). $B (R) = σ ({(a, b) : a, b \in Q, a < b})$ . The generating family on the right is countable.

Proof. Let $G = {(a, b) : a, b \in Q, a < b}$ , which is countable since $Q^{2}$ is countable. Each element of $G$ is open, so $σ (G) \subseteq B (R)$ .

For the reverse, every open subset $U \subseteq R$ is a countable union of open intervals with rational endpoints: for each $x \in U$ , choose rationals $a < x < b$ with $(a, b) \subseteq U$ (possible by density of $Q$ and openness of $U$ ). The collection ${(a, b) \in G : (a, b) \subseteq U}$ is countable and covers $U$ , so $U$ is a countable union of elements of $G$ . Hence $U \in σ (G)$ . Since open sets generate $B (R)$ , $B (R) \subseteq σ (G)$ .

$□$

Proposition 3 (measurable functions form a vector lattice). Let $(X, M)$ be a measurable space and let $f, g : X \to R$ be measurable. Then $f + g$ , $c f$ for $c \in R$ , $f g$ , $max (f, g)$ , and $min (f, g)$ are all measurable.

Proof. The sum $f + g$ is measurable by Exercise 8.

For $c f$ : if $c = 0$ , $c f \equiv 0$ has pre-images $\emptyset$ or $X$ , both measurable. If $c > 0$ , ${x : c f (x) > a} = {x : f (x) > a / c}$ is measurable. If $c < 0$ , ${x : c f (x) > a} = {x : f (x) < a / c}$ is measurable.

For $f g$ : use the identity $f g = \frac{1}{4} ((f + g)^{2} - (f - g)^{2})$ . So it suffices to show $h^{2}$ is measurable when $h$ is. For $a \geq 0$ , ${x : h (x)^{2} > a} = {x : h (x) > a} \cup {x : h (x) < - a}$ . For $a < 0$ , ${x : h (x)^{2} > a} = X$ . All these are measurable.

For $max (f, g)$ : ${x : max (f, g) (x) > a} = {x : f (x) > a} \cup {x : g (x) > a}$ is measurable. Similarly for $min$ , using intersection.

$□$

Proposition 4 (intersection of sigma-algebras is a sigma-algebra — explicit version). Restated from the Key theorem in detailed form for self-containment.

Proof. See the Key theorem with proof section. The argument is reproduced there in full.

$□$

Connections Master

Lebesgue outer measure and Carathéodory extension 02.07.02. The next unit in this sub-chapter builds the Lebesgue measure on $B (R)$ and extends it to $L$ . The Carathéodory criterion picks out the measurable sets from the outer measure, and the resulting sigma-algebra is precisely the completion of $B (R)$ identified in Theorem 4 here. The generating-class argument from Exercise 4 is the standard tool for proving uniqueness of the extension.
Compactness in topological spaces 02.01.10. On a locally compact Hausdorff space, the Borel sigma-algebra is generated by the compact open sets (or equivalently the relatively compact open sets), and the resulting measure theory becomes Radon measure theory. The interaction between compactness and Borel structure underlies the Riesz representation theorem and the construction of Haar measure on locally compact groups.
Banach space fundamentals 02.11.04. The $L^{p}$ spaces are Banach spaces whose construction requires the measurable-function framework of this unit. A function lies in $L^{p}$ when its $p$ -th power is measurable and integrable, so the measurability framework here is logically prior to every function-space construction. The closed-subspace theorem from the Banach unit reappears in the proof that $L^{p}$ is complete: the limit of an $L^{p}$ -Cauchy sequence is taken pointwise on a subsequence and shown to lie in $L^{p}$ via Fatou.
Continuous map 02.01.02. Theorem 6 makes the bridge precise: continuity from $Y$ to $Z$ is exactly the statement that $g^{- 1}$ preserves the topology-generated sigma-algebras, which is the topological refinement of the measurability condition. Every continuous function between topological spaces is Borel-measurable, but Borel-measurable functions form a strictly larger class — limits of continuous functions, indicator functions of Borel sets, and many discontinuous functions are Borel-measurable.
Real number axioms 02.02.01. The least-upper-bound property of $R$ enters through the topology: the standard topology has a countable base, $R$ is second-countable, and the Borel sigma-algebra is countably generated as a consequence (Proposition 2). On a non-second-countable space the Borel sigma-algebra need not be countably generated, and the size and reach of the sigma-algebra change accordingly.

Historical & philosophical context Master

Borel introduced what we now call the Borel sigma-algebra in his 1898 Leçons sur la théorie des fonctions ^{[Borel 1898]}, where he extended the notion of length from intervals to a larger class of subsets of the real line via countable additivity over disjoint open intervals. Borel's construction was descriptive: start with intervals and close under countable union and complementation, identifying along the way which subsets admit a measure consistent with length on intervals.

Lebesgue's 1902 thesis ^{[Lebesgue 1902]} reframed the problem as a question about the outer-inner approximation gap. A bounded subset of $R$ admits a measure when its outer measure (the infimum of measures of open supersets) and inner measure (the supremum of measures of closed subsets) agree. The resulting sigma-algebra $L$ is strictly larger than $B (R)$ : it is the Lebesgue completion under null sets. The Lebesgue integral then becomes natural because monotone convergence and dominated convergence theorems hold on $L$ , while their Riemann analogues fail on the Borel sigma-algebra restricted to continuous integrands.

Vitali's 1905 construction ^{[Vitali 1905]} showed that the axiom of choice produces a subset of $[0, 1]$ that lies outside $L$ . The proof depends on choosing one representative from each rational-translation equivalence class, and the resulting set fails countable additivity by translation. This was the first non-measurable set, and it placed the axiom of choice at the centre of measure-theoretic foundations.

Halmos's 1950 Measure Theory ^{[Halmos 1950]} systematised the sigma-algebra framework in the form that survives today: abstract measurable spaces, sigma-algebras generated by arbitrary families, the monotone class theorem, and the extension theorems that produce measures from algebras. The abstraction freed measure theory from the real line: probability theory (Kolmogorov 1933), ergodic theory (Birkhoff and von Neumann), and integration on locally compact groups (Haar 1933, Weil 1940) all took the abstract sigma-algebra framework as their starting point.

Solovay's 1970 theorem ^{[Solovay 1970]} settled the foundational question raised by Vitali: in a model of ZF + DC built from a model of ZFC plus an inaccessible cardinal, every subset of $R$ is Lebesgue-measurable. The existence of non-measurable sets is therefore not a theorem of ZF + DC, only of ZF + AC. The Solovay model also resolves a parallel question about the Baire property and the perfect-set property, and remains a touchstone for descriptive set theory.

Bibliography Master

@book{Borel1898,
  author    = {Borel, \'Emile},
  title     = {Le\c{c}ons sur la th\'eorie des fonctions},
  publisher = {Gauthier-Villars},
  address   = {Paris},
  year      = {1898}
}

@article{Lebesgue1902,
  author  = {Lebesgue, Henri},
  title   = {Int\'egrale, longueur, aire},
  journal = {Annali di Matematica Pura ed Applicata},
  volume  = {7},
  year    = {1902},
  pages   = {231--359}
}

@misc{Vitali1905,
  author    = {Vitali, Giuseppe},
  title     = {Sul problema della misura dei gruppi di punti di una retta},
  publisher = {Tip. Gamberini e Parmeggiani},
  address   = {Bologna},
  year      = {1905}
}

@book{Halmos1950,
  author    = {Halmos, Paul R.},
  title     = {Measure Theory},
  publisher = {Van Nostrand},
  year      = {1950},
  note      = {Reprinted as Springer GTM 18, 1974}
}

@book{Folland1999,
  author    = {Folland, Gerald B.},
  title     = {Real Analysis: Modern Techniques and Their Applications},
  edition   = {2},
  publisher = {Wiley},
  year      = {1999}
}

@book{Rudin1987,
  author    = {Rudin, Walter},
  title     = {Real and Complex Analysis},
  edition   = {3},
  publisher = {McGraw-Hill},
  year      = {1987}
}

@book{Bogachev2007,
  author    = {Bogachev, Vladimir I.},
  title     = {Measure Theory, Volume 1},
  publisher = {Springer},
  year      = {2007}
}

@article{Solovay1970,
  author  = {Solovay, Robert M.},
  title   = {A model of set-theory in which every set of reals is Lebesgue measurable},
  journal = {Annals of Mathematics},
  volume  = {92},
  year    = {1970},
  pages   = {1--56}
}

@book{Royden1988,
  author    = {Royden, H. L.},
  title     = {Real Analysis},
  edition   = {3},
  publisher = {Macmillan},
  year      = {1988}
}

@book{Tao2016,
  author    = {Tao, Terence},
  title     = {Analysis II},
  edition   = {3},
  publisher = {Hindustan Book Agency},
  year      = {2016}
}

Prerequisites

02.01.01
02.01.02
02.02.01

Tier anchors

beginner: Tao, Analysis II §7; informal motivation from length and area
intermediate: Folland, Real Analysis 2e §1.2; Royden, Real Analysis §3
master: Halmos, Measure Theory §I; Bogachev, Measure Theory Vol. 1 §1.1-1.4; Rudin, Real and Complex Analysis 3e §1

References

Folland — Real Analysis: Modern Techniques and Their Applications, 2e · §1.2, σ-algebras and measurable spaces
Halmos — Measure Theory · §I, Set-theoretic preliminaries; §5 σ-rings
Bogachev — Measure Theory, Vol. 1 · §1.1-1.4, σ-algebras and Borel sets
Rudin — Real and Complex Analysis, 3e · §1, Abstract integration
Borel — Leçons sur la théorie des fonctions · Gauthier-Villars 1898
Lebesgue — Intégrale, longueur, aire · Annali di Matematica Pura ed Applicata 7 (1902)
Vitali — Sul problema della misura dei gruppi di punti di una retta · Bologna 1905
Solovay — A model of set-theory in which every set of reals is Lebesgue measurable · Annals of Mathematics 92 (1970), 1-56
Tao — Analysis II · §7, Lebesgue measure

Lean module

Codex.Analysis.MeasureTheory.SigmaAlgebraBorel

Estimated time

beginner: 18m
intermediate: 45m
master: 75m