Lebesgue Measure on R^n

Lebesgue measure is the canonical way to assign a notion of $n$-dimensional volume to subsets of Euclidean space. One begins with the outer measure ${\lambda }^{\ast }$ obtained by covering a set with countably many boxes and taking an infimum of total volumes; a set is measurable if it splits every other set additively (Carathéodory's criterion). The resulting measure is translation-invariant, regular, and complete, and it agrees with elementary geometry on boxes and with Jordan content on nice sets. The surprising fact: under the Axiom of Choice there exist sets that are not Lebesgue measurable (Vitali sets), so no genuinely useful measure can assign a volume to every subset of ${\mathbb{R}}^n$ while remaining translation-invariant and countably additive.

The Core Insight

Formal Statement

Elementary volume of a box

A box (or rectangle/cell) in ${\mathbb{R}}^n$ is a product $Q={\prod }_{i=1}^n{I}_i$ where each $I_i$ is a bounded interval with endpoints $a_i\le b_i$. Its elementary volume is

Lebesgue outer measure

For an arbitrary set $E\subseteq {\mathbb{R}}^n$, the Lebesgue outer measure is

Carathéodory measurability

A set $E\subseteq {\mathbb{R}}^n$ is Lebesgue measurable if for every test set $A\subseteq {\mathbb{R}}^n$,

On the $\sigma$-algebra $\mathcal{L}$ of measurable sets, Lebesgue measure is the restriction $\lambda ={\lambda }^{\ast }{|}_{\mathcal{L}}$.

Existence, uniqueness, and core properties

There is a unique measure $\lambda$ on a $\sigma$-algebra $\mathcal{L}\supseteq \mathcal{B}({\mathbb{R}}^n)$ (the Borel sets) satisfying:

1. Box normalization: $\lambda (Q)=\mathrm{vol}(Q)$ for every box.
2. Countable additivity: for disjoint measurable $E_k$, $\lambda \left({⨆}_k{E}_k\right)={\sum }_k\lambda (E_k)$.
3. Translation invariance: $\lambda (E+x)=\lambda (E)$ for all $x\in {\mathbb{R}}^n$, $E\in \mathcal{L}$.
4. Completeness: every subset of a null set is measurable with measure $0$.
5. Regularity: for $E\in \mathcal{L}$,

Moreover any translation-invariant Borel measure finite on boxes equals $c\lambda$ for some constant $c\ge 0$.

Linear transformation rule

For an invertible linear map $T\in G{L}_n(\mathbb{R})$ and measurable $E$,

Why It Holds / Derivation

Carathéodory's criterion yields a complete measure

We sketch the construction proving the theorem.

1. ${\lambda }^{\ast }$ is an outer measure. From [2], ${\lambda }^{\ast }(\varnothing )=0$, ${\lambda }^{\ast }$ is monotone, and countably subadditive: given covers of each $E_k$ within $\epsilon 2^{-k}$ of the infimum, their union covers ${\bigcup }_k{E}_k$ and total volume is within $\epsilon$ of ${\sum }_k{\lambda }^{\ast }(E_k)$.

1. Measurable sets form a $\sigma$-algebra $\mathcal{L}$. Equation [3] is symmetric in $E,E^c$, so $\mathcal{L}$ is closed under complements and contains $\varnothing$. For finite unions, apply [3] twice and use subadditivity. For countable unions, reduce to disjoint sets and use that finite measurability already yields, for disjoint $E_1,\dots ,E_m$,

1. ${\lambda }^{\ast }$ is countably additive on $\mathcal{L}$. Take $A={⨆}_k{E}_k$ in the disjoint additivity above and let $m\to \infty$; combined with subadditivity this forces equality.

1. Boxes are measurable and keep their volume. A box $Q$ cuts any cover of a test set $A$ into the part inside $Q$ and outside; comparing box volumes shows [3] holds. A covering/packing argument shows ${\lambda }^{\ast }(Q)=\mathrm{vol}(Q)$ — the nontrivial direction needs compactness (Heine–Borel) to reduce an infinite cover of a closed box to a finite one.

1. Completeness. If ${\lambda }^{\ast }(N)=0$ and $S\subseteq N$, then ${\lambda }^{\ast }(A\cap S)\le {\lambda }^{\ast }(N)=0$, so [3] holds trivially; hence $S\in \mathcal{L}$.

1. Borel sets are measurable. Half-spaces $\{x_i<c\}$ satisfy [3], they generate the Borel $\sigma$-algebra, and $\mathcal{L}$ is a $\sigma$-algebra; thus $\mathcal{B}({\mathbb{R}}^n)\subseteq \mathcal{L}$. $\square$

Approximating a planar region by an inner packing and outer covering of dyadic squares: inner squares (lower estimate) and outer squares (upper estimate) trap the Lebesgue area between them.

Henri Léon Lebesgue | 1875–1941

A French mathematician whose 1902 thesis "Intégrale, longueur, aire" introduced both the measure and integral now bearing his name. Lebesgue reversed Riemann's recipe: instead of partitioning the domain, he partitioned the range of a function and asked how large the preimage of each level set is — a question that demands a flexible notion of size, i.e. measure. His integral exchanges limits and integration far more freely (dominated and monotone convergence) and integrates many functions Riemann cannot. Lebesgue himself was initially skeptical of the heavy set-theoretic machinery that Carathéodory, Borel, and others built around his ideas.

Building Intuition

Think of measuring a puddle with tiles. Jordan content lets you use only finitely many tiles, so a "puddle" like the rationals in $[0,1]$ — dense but with empty interior — cannot be tiled: any finite inner tiling captures nothing, any finite outer tiling captures the whole interval. Lebesgue's countable covers let you place a tiny tile of width $\epsilon 2^{-k}$ over the $k$-th rational, capturing all of them with total width $\le \epsilon$. Letting $\epsilon \to 0$ shows the rationals have measure zero. The deeper Carathéodory idea is to stop describing the set and instead test it: a set is measurable if, for every conceivable probe set $A$, slicing $A$ by $E$ never loses or creates volume at the boundary. Non-measurable sets fail precisely because their "boundary" is so pathologically diffuse that the outer measure of the inside plus the outside exceeds the whole.

Worked Examples

Measure of the Smith–Volterra–Cantor "fat Cantor" set

• Total length removed: at step $k$ we remove $2^{k-1}$ intervals each of length $4^{-k}$:

• By countable additivity, $\lambda (C)=1-\frac{1}{2}=\frac{1}{2}$.
• $C$ has empty interior, so its Jordan inner content is $0$ while its Lebesgue measure is $1/2$ — a set Riemann/Jordan theory cannot measure.

Volume of a parallelepiped via the linear rule

• $\lambda ([0,1{]}^3)=1$.
• $\det T=2(3\cdot 2-1\cdot 0)-1(0\cdot 2-1\cdot 1)+0=2\cdot 6-1\cdot (-1)=12+1=13.$
• By [5], $\lambda (T([0,1{]}^3))=|\det T|\cdot 1=13.$

Common Misconceptions

• "Measure zero means countable." False: the standard middle-thirds Cantor set is uncountable yet has measure $0$. Conversely, a measure-zero set can be dense (the rationals) or can have full topological "fatness" failing — measure and cardinality/topology are independent.
• "Every subset of ${\mathbb{R}}^n$ is measurable." False under the Axiom of Choice (Vitali). It is consistent with ZF + Dependent Choice (Solovay's model) that all sets are measurable, so non-measurability is genuinely tied to strong choice.
• "Lebesgue measurable = Borel." False: $\mathcal{L}$ is strictly larger. By cardinality, $|\mathcal{B}(\mathbb{R})|=2^{{\aleph }_0}$ but $|\mathcal{L}|=2^{2^{{\aleph }_0}}$ because every subset of the Cantor null set is measurable.
• "${\lambda }^{\ast }(A\cup B)={\lambda }^{\ast }(A)+{\lambda }^{\ast }(B)$ for disjoint $A,B$." Only for measurable sets; outer measure is merely subadditive, and additivity can fail for disjoint non-measurable pieces.

A non-measurable Vitali set

Define $x\sim y$ iff $x-y\in \mathbb{Q}$ on $[0,1]$. By the Axiom of Choice pick one representative from each class, forming $V\subseteq [0,1]$. Enumerate $\mathbb{Q}\cap [-1,1]=\{q_k\}$. The translates $V+q_k$ are pairwise disjoint and satisfy

Connections

Lebesgue measure is the foundation on which Lebesgue Integration is built: $\int fd\lambda$ first defines integrals of simple functions via measures of level sets, then extends by monotone limits. It is the prototype produced by the general Carathéodory Extension Theorem from a premeasure on the algebra of finite box unions. Its decomposition theory connects to Radon–Nikodym Theorem (densities of measures absolutely continuous w.r.t. $\lambda$) and to Lebesgue Decomposition (splitting into absolutely continuous, singular continuous, and atomic parts). The product structure ${\lambda }_n={\lambda }_1\otimes \cdots \otimes {\lambda }_1$ is the model case of Product Measure and underlies Fubini–Tonelli Theorem. Translation invariance generalizes to the Haar Measure on locally compact groups, of which $({\mathbb{R}}^n,+)$ is the abelian prototype, and the scaling rule [5] previews Hausdorff Measure and fractal dimension.

Real-World Applications

• Probability theory: $\lambda$ restricted (or reweighted by a density) to $[0,1]$ or ${\mathbb{R}}^n$ provides the uniform and continuous distributions; Kolmogorov's axioms are exactly a normalized measure space.
• Signal processing & Fourier analysis: $L^2({\mathbb{R}}^n,\lambda )$ is the Hilbert space in which the Fourier/Plancherel theory, sampling, and energy ($\int |f{|}^{2}d\lambda$) live.
• Statistical mechanics: the microcanonical and Liouville measures on phase space ${\mathbb{R}}^{6N}$ are Lebesgue measure, whose invariance under Hamiltonian flow (Liouville's theorem) is the linear rule [5] applied to volume-preserving symplectic maps.
• Ergodic theory & dynamical systems: invariant measures, mixing, and Birkhoff's theorem are stated for measure-preserving transformations modeled on $\lambda$.
• Image processing & computer graphics: pixel/voxel "area" and volume estimates, anti-aliasing, and Monte Carlo integration approximate $\lambda$-integrals.
• Economics & game theory: continua of agents are modeled by atomless (Lebesgue) measure spaces, giving "negligible individual" results (Aumann).
• Numerical analysis: quasi-Monte Carlo and quadrature error bounds quantify how well discrete samples approximate $\int fd\lambda$.

Historical Development

• 1883 — Cantor & Stolz: early notions of content; Cantor constructs his ternary set, foreshadowing measure-zero pathologies.
• 1887–1892 — Peano, Jordan: Peano–Jordan content formalizes finite-cover "measurability," but fails for many sets (e.g. $\mathbb{Q}\cap [0,1]$).
• 1898 — Émile Borel: introduces countably additive measure on what become the Borel sets, insisting on countable operations.
• 1901–1902 — Henri Lebesgue: Comptes Rendus note (1901) and thesis Intégrale, longueur, aire (1902) define outer/inner measure and the Lebesgue integral.
• 1905 — Giuseppe Vitali: constructs the first non-measurable set using the Axiom of Choice.
• 1914 — Constantin Carathéodory: abstract outer-measure criterion [3], freeing the theory from inner measure and topology.
• 1918 — Felix Hausdorff: Hausdorff measures and dimension generalize $\lambda$ to fractional dimension; Hausdorff paradox prefigures Banach–Tarski.
• 1924 — Banach & Tarski: the paradoxical decomposition shows no finitely additive, isometry-invariant, total measure on ${\mathbb{R}}^{\ge 3}$ exists — explaining why countable additivity must coexist with non-measurable sets.
• 1970 — Robert Solovay: constructs a model of ZF+DC in which every set of reals is Lebesgue measurable, locating non-measurability precisely within strong choice.

Subtleties & Limits

• Invariant extensions: Can $\lambda$ be extended to a translation-invariant countably additive measure on all subsets of ${\mathbb{R}}^n$? No — Vitali/Banach–Tarski forbid it. But finitely additive, isometry-invariant extensions to all of $\mathcal{P}(\mathbb{R})$ and $\mathcal{P}({\mathbb{R}}^2)$ exist (Banach), and fail for $n\ge 3$ — amenability of the isometry group is the dividing line.
• Real-valued measurable cardinals: The question of whether $\lambda$ can be extended to a countably additive measure on all subsets (dropping translation invariance) is equivalent to the existence of a real-valued measurable cardinal — a large-cardinal axiom independent of ZFC (Banach–Ulam, Ulam).
• Regularity from determinacy: Under the Axiom of Determinacy (AD), every set of reals is Lebesgue measurable; AD contradicts AC, so the universal measurability of sets is a choiceless luxury. Solovay's model achieves it relative to an inaccessible cardinal.
• Constructive content: The non-constructive nature of Vitali sets means non-measurable sets cannot be exhibited explicitly; the boundary between definable and measurable sets is governed by the projective hierarchy and the theory of ${\mathbf{\Sigma }}_n^1$ measurability.

See Also