Lebesgue Measure on R^n

💡 The Core Insight

Lebesgue measure extends the intuitive notion of $n$ -dimensional volume to a vast class of sets by approximating from the outside with countable coverings of boxes, then restricting to those sets that are "measurable" in the sense that they split every other set additively with respect to outer measure.

Definition

Lebesgue Measure on $\mathbb{R}^n$ : The Lebesgue outer measure $\lambda^*: \mathcal{P}(\mathbb{R}^n) \to [0,\infty]$ is defined by

\lambda^*(E) = \inf\left\{\sum_{j=1}^\infty \text{vol}(R_j) : E \subseteq \bigcup_{j=1}^\infty R_j, \text{ each } R_j \text{ a box}\right\}

where a box is a product $R = [a_1,b_1) \times \cdots \times [a_n,b_n)$ with volume $\text{vol}(R) = \prod_{i=1}^n (b_i - a_i)$ . A set $E \subseteq \mathbb{R}^n$ is Lebesgue measurable (or $\lambda$ -measurable) if for every test set $A \subseteq \mathbb{R}^n$ ,

\lambda^*(A) = \lambda^*(A \cap E) + \lambda^*(A \cap E^c)

The collection $\mathcal{L}(\mathbb{R}^n)$ of all Lebesgue measurable sets forms a sigma-algebra, and Lebesgue measure $\lambda: \mathcal{L}(\mathbb{R}^n) \to [0,\infty]$ is the restriction of $\lambda^*$ to $\mathcal{L}(\mathbb{R}^n)$ .

TheoremFundamental Properties of Lebesgue Measure

Theorem: Lebesgue measure $\lambda$ on $\mathbb{R}^n$ satisfies:

[1] (Countable Additivity) If $\{E_j\}_{j=1}^\infty$ are pairwise disjoint measurable sets, then $\lambda\left(\bigcup_{j=1}^\infty E_j\right) = \sum_{j=1}^\infty \lambda(E_j)$ .

[2] (Translation Invariance) For any measurable set $E$ and any $h \in \mathbb{R}^n$ , the translate $E + h = \{x + h : x \in E\}$ is measurable and $\lambda(E + h) = \lambda(E)$ .

[3] (Normalization) For any box $R = [a_1,b_1) \times \cdots \times [a_n,b_n)$ , we have $\lambda(R) = \prod_{i=1}^n (b_i - a_i)$ .

[4] (Borel Regularity) Every Borel set is Lebesgue measurable. Moreover, for any $E \in \mathcal{L}(\mathbb{R}^n)$ , there exist Borel sets $B_1 \subseteq E \subseteq B_2$ with $\lambda(B_2 \setminus B_1) = 0$ .

[5] (Completeness) If $E \in \mathcal{L}(\mathbb{R}^n)$ with $\lambda(E) = 0$ and $F \subseteq E$ , then $F \in \mathcal{L}(\mathbb{R}^n)$ and $\lambda(F) = 0$ .

Proof

We prove [1] and [2]; the others follow from standard measure-theoretic arguments.

Proof of [1]: First, we show finite additivity for outer measure and disjoint sets at positive distance. If $E_1, E_2$ satisfy $\text{dist}(E_1, E_2) = \delta > 0$ , and $\{R_j\}$ covers $E_1 \cup E_2$ , we can partition $\{R_j\}$ into those that intersect $E_1$ (collecting to volume $V_1$ ) and those intersecting $E_2$ (collecting to volume $V_2$ ), with possible overlap contributing at most the boundary terms. Taking infima carefully yields $\lambda^*(E_1 \cup E_2) \geq \lambda^*(E_1) + \lambda^*(E_2)$ . The reverse inequality follows from subadditivity of outer measure. For the general case, let $\{E_j\}_{j=1}^\infty$ be pairwise disjoint measurable sets, and let $F_n = \bigcup_{j=1}^n E_j$ . By the Carathéodory criterion for measurability and induction, $F_n$ is measurable and $\lambda(F_n) = \sum_{j=1}^n \lambda(E_j)$ . Let $E = \bigcup_{j=1}^\infty E_j$ . For any test set $A$ , we have $A \cap E = \bigcup_{j=1}^\infty (A \cap E_j)$ , and by measurability of $F_n$ :

\lambda^*(A \cap E) \geq \lambda^*(A \cap F_n) = \sum_{j=1}^n \lambda(E_j)

Taking $n \to \infty$ gives $\lambda^*(A \cap E) \geq \sum_{j=1}^\infty \lambda(E_j)$ . The reverse inequality follows from countable subadditivity of $\lambda^*$ , giving $\lambda(E) = \sum_{j=1}^\infty \lambda(E_j)$ .

Proof of [2]: We first show translation invariance of outer measure. For any covering $E \subseteq \bigcup_{j=1}^\infty R_j$ by boxes, we have $E + h \subseteq \bigcup_{j=1}^\infty (R_j + h)$ , and each translated box has the same volume. Thus $\lambda^*(E + h) \leq \lambda^*(E)$ . Applying the same argument to $E + h$ and translation by $-h$ gives $\lambda^*(E) \leq \lambda^*(E + h)$ , so $\lambda^*(E + h) = \lambda^*(E)$ . Now suppose $E$ is measurable. For any test set $A$ :

\lambda^*(A) = \lambda^*(A \cap E) + \lambda^*(A \cap E^c)

Translating by $h$ : $\lambda^*((A - h)) = \lambda^*((A - h) \cap E) + \lambda^*((A - h) \cap E^c)$ . By translation invariance of outer measure:

\lambda^*(A - h) = \lambda^*((A - h) \cap E) + \lambda^*((A - h) \cap E^c)

Substituting $A' = A - h$ and using $(A' \cap E) + h = A \cap (E + h)$ :

\lambda^*(A') = \lambda^*(A' \cap (E + h)) + \lambda^*(A' \cap (E + h)^c)

Thus $E + h$ is measurable. □

∎

⚠️ Common mistake

A common error is assuming that every subset of $\mathbb{R}^n$ is Lebesgue measurable. In fact, the Axiom of Choice guarantees the existence of non-measurable sets (e.g., Vitali sets). The Carathéodory measurability criterion is not merely a technicality—it's essential for ensuring countable additivity. Without restricting to measurable sets, one cannot have both translation invariance and countable additivity simultaneously for all subsets.

mathematician

Henri Lebesgue (1902) — In his doctoral thesis, Lebesgue introduced the construction of outer measure and the criterion for measurability, revolutionizing integration theory and laying the foundation for modern measure theory.

visualization

(E) ≤</text> <text x="270" y="120" font-size="11" fill="#82aaff" font-family="serif">Σ vol(Rⱼ)</text> <text x="287" y="132" font-size="8" fill="#82aaff" font-family="serif">j=1</text> <text x="305" y="132" font-size="10" fill="#82aaff" font-family="serif">⁵</text>  <text x="200" y="190" font-size="10" fill="#c9d1d9" text-anchor="middle">λ(E) = inf { Σ vol(Rⱼ) : E ⊆ ⋃Rⱼ, countable covering }

Example1

Computing the measure of a closed interval in $\mathbb{R}$ Let $E = [a,b] \subseteq \mathbb{R}$ . We claim $\lambda(E) = b - a$ . Upper bound: The interval $[a,b]$ is covered by the single box $[a,b)$ (plus the single point $\{b\}$ which has outer measure 0), so $\lambda^*(E) \leq (b-a) + 0 = b-a$ . Lower bound: Suppose $E \subseteq \bigcup_{j=1}^\infty [a_j, b_j)$ is a countable covering. By compactness of $[a,b]$ , there exists a finite subcover $E \subseteq \bigcup_{j=1}^N [a_j, b_j)$ . Order these intervals and observe that to cover $[a,b]$ completely, we must have $a_1 \leq a$ and the intervals must have sufficient overlap. By a careful analysis (or by the Heine-Borel theorem), we get $\sum_{j=1}^N (b_j - a_j) \geq b - a$ , hence $\sum_{j=1}^\infty (b_j - a_j) \geq b - a$ . Taking the infimum over all coverings: $\lambda^*(E) \geq b - a$ . Therefore $\lambda([a,b]) = b - a$ . Since closed intervals are measurable (as closed sets, they're Borel, hence Lebesgue measurable), this is the Lebesgue measure of $[a,b]$ .

Example2

The Cantor set has measure zero The Cantor set $C \subseteq [0,1]$ is constructed by iteratively removing middle thirds. At stage $n$ , we have $2^n$ closed intervals, each of length $3^{-n}$ , with total length $\sum_{k=1}^{2^n} 3^{-n} = (2/3)^n$ . Since $C$ is the intersection of these sets, and outer measure is monotone:

\lambda^*(C) \leq (2/3)^n \text{ for all } n

Taking $n \to \infty$ , we get $\lambda^*(C) = 0$ . This is remarkable: $C$ is uncountable (having the cardinality of $\mathbb{R}$ ), yet has Lebesgue measure zero. This shows that "measure zero" is a geometric, not merely cardinality-based, notion. The Cantor set is measurable (being a Borel set) with $\lambda(C) = 0$ .

Counterexample

Why we cannot measure all subsets Suppose we could extend Lebesgue measure to all subsets of $[0,1)$ while maintaining translation invariance (modulo 1) and countable additivity. Define an equivalence relation: $x \sim y$ if $x - y \in \mathbb{Q}$ . Using the Axiom of Choice, select one representative from each equivalence class to form a set $V$ (a Vitali set). For each $q \in \mathbb{Q} \cap [0,1)$ , the translates $V + q \pmod{1}$ are pairwise disjoint and their union is $[0,1)$ . If $V$ were measurable:

1 = \lambda([0,1)) = \sum_{q \in \mathbb{Q} \cap [0,1)} \lambda(V + q) = \sum_{q \in \mathbb{Q} \cap [0,1)} \lambda(V)

If $\lambda(V) = 0$ , the sum is 0 (contradiction). If $\lambda(V) > 0$ , the sum is $+\infty$ (contradiction). Thus $V$ cannot be measurable. This shows that the Carathéodory measurability criterion is necessary—we genuinely cannot measure all sets while preserving the properties we want.

🧠 Intuition

Think of Lebesgue measure as answering: "If I threw darts randomly at $\mathbb{R}^n$ , what's the probability of hitting set $E$ ?" For this to make sense, we need to approximate $E$ from the outside using simple shapes (boxes) whose volumes we understand intuitively. The outer measure $\lambda^*(E)$ is the "tightest wrapping" we can achieve with countably many boxes. Not all sets play nicely with this wrapping process. A set is measurable if it doesn't have pathological boundaries that prevent us from consistently determining how it splits other sets. Geometrically, measurable sets are those where the "boundary effects" don't accumulate in measure—the set and its complement interact cleanly with outer measure. The key conceptual shift from Riemann to Lebesgue: instead of partitioning the domain and asking "what is the function doing on each piece?", we partition the range and ask "where does the function take each value?" This reversal makes limit operations natural and powerful.

connection

Quantum Mechanics and Spectral Theory: Lebesgue measure is the canonical spectral measure for the position operator in quantum mechanics. In the spectral theorem for self-adjoint operators on Hilbert spaces, the position operator $\hat{x}$ on $L^2(\mathbb{R}^n, \lambda)$ is represented as multiplication by the identity function, and its spectral decomposition is given by integration with respect to Lebesgue measure. The Born rule, which states that $|\psi(x)|^2 dx$ gives the probability density for finding a particle at position $x$ , explicitly uses Lebesgue measure as the reference measure. More abstractly, the Gelfand-Naimark theorem and the theory of von Neumann algebras show that the measure-theoretic structure of quantum observables is deeply tied to integration theory built on measures like Lebesgue's.

Lebesgue Measure on R^n

Overview

Core Theory

Examples and Applications

Historical Development

Connections to Other Areas

What You Need First

What This Opens Up

See Also