Measure Theory

The Lebesgue integral extends Riemann integration to a vastly larger class of functions, including all pointwise limits of integrable functions. Instead of partitioning the domain (as Riemann does), it partitions the range: the integral of f is built up by asking 'how much measure does f assign to values near y?' This reorientation, due to Henri Lebesgue (1902), resolved deep deficiencies of Riemann integration and became the foundation of modern analysis and probability.

The three classical convergence theorems — the Monotone Convergence Theorem (MCT), Fatou's Lemma, and the Dominated Convergence Theorem (DCT) — are the most powerful and frequently used results in integration theory. They provide conditions under which the limit of integrals equals the integral of the limit, resolving the fundamental question of when limit and integral can be interchanged.

Given two sigma-finite measure spaces, the product measure is the unique measure on the product sigma-algebra that assigns to each measurable rectangle the product of the factor measures. Fubini's theorem then justifies interchanging the order of integration: if a function is integrable (or non-negative measurable), the double integral equals the iterated integrals in either order. Tonelli's theorem handles the non-negative case without integrability, while Fubini's theorem requires absolute integrability.

Borel sets form the smallest sigma-algebra containing all open sets of a topological space. On the real line they are generated by the open intervals and include virtually every set one encounters in analysis: open and closed sets, countable unions and intersections thereof, and their iterated combinations. Borel sets are the natural domain for Borel measures and are central to real analysis, probability theory, and descriptive set theory.

The Carathéodory Extension Theorem provides a rigorous method to extend a pre-measure defined on an algebra (or semi-ring) of sets to a full countably additive measure on the generated sigma-algebra. It is the foundational result that justifies the construction of Lebesgue measure from the length function on intervals, and more generally allows measures to be defined via their values on simple sets and then extended uniquely to a complete measure space.

An outer measure is a set function defined on all subsets of a space (not just measurable ones) that satisfies monotonicity and countable sub-additivity. Outer measures are the starting point for Carathéodory's construction: from any outer measure one extracts a sigma-algebra of 'measurable' sets on which the outer measure restricts to a genuine measure. This machinery produces Lebesgue measure, Hausdorff measure, and many others.

Lebesgue measure is the canonical extension of length, area, and volume to a far larger class of subsets of ℝⁿ. On ℝ it assigns to each interval (a, b) the value b − a, and it extends this consistently to all Borel sets and beyond (to the full Lebesgue σ-algebra). The construction via outer measure and Carathéodory's criterion is one of the most elegant in modern mathematics.

The L^p spaces (1 <= p <= ∞) are Banach spaces of measurable functions on a measure space (X, F, mu), equipped with the L^p norm. They provide the natural function spaces for Fourier analysis, PDEs, probability theory, and functional analysis. The cases p = 1 (integrable functions), p = 2 (square-integrable, a Hilbert space), and p = ∞ (essentially bounded) are most frequently encountered.

A measurable function is a map between measurable spaces that respects their σ-algebraic structure: the pre-image of every measurable set is measurable. This is the natural notion of 'structure-preserving map' in measure theory, analogous to continuous functions in topology. Measurable functions are the integrands of the Lebesgue integral and the random variables of probability theory.

A measure space is the fundamental structure of measure theory: a set X equipped with a σ-algebra ℱ and a function μ : ℱ → [0, ∞] that assigns a 'size' to every measurable set in a consistent way. This framework generalises length on ℝ, area in ℝ², counting on discrete sets, and probability on sample spaces, all within a single axiomatic system.

A null set (or measure-zero set) is a measurable set whose measure is zero. The phrase 'almost everywhere' (abbreviated a.e.) means 'except possibly on a null set'. These concepts are ubiquitous in analysis and probability: two functions that agree a.e. are identified in L^p spaces, convergence theorems hold a.e., and properties that fail only on null sets are treated as universally true for most analytic purposes.

A σ-algebra (sigma-algebra) is a collection of subsets of a given set that is closed under complementation and countable unions. σ-algebras form the backbone of measure theory: they precisely identify which subsets of a space are 'measurable', making it possible to assign consistent sizes (measures) to those sets. Without the σ-algebra framework, paradoxes like the Banach–Tarski paradox show that naive notions of size lead to contradictions.

A probability measure is a measure P on a measurable space (Omega, F) satisfying P(Omega) = 1. The triple (Omega, F, P) is called a probability space. This is Kolmogorov's axiomatic foundation for probability theory (1933), which unified all of classical probability within the measure-theoretic framework. Random variables become measurable functions, expectations become Lebesgue integrals, and convergence concepts (almost sure, in probability, in L^p) are precisely defined.

The Radon–Nikodym theorem establishes when one measure can be expressed as an integral with respect to another. If nu is absolutely continuous with respect to mu (nu << mu), then there exists a measurable function f — the Radon–Nikodym derivative — such that nu(E) = integral_E f dmu for all measurable E. This function f = dnu/dmu generalises the classical derivative and is fundamental to probability theory (conditional expectations), statistics (likelihood ratios), and functional analysis (dual spaces of L^p).

A signed measure generalises a positive measure by allowing sets to have negative measure. This is essential for representing differences of measures, for defining charges in electrostatics, and for the Radon–Nikodym theorem. Every signed measure decomposes uniquely (Jordan decomposition) into the difference of two positive measures supported on disjoint sets (Hahn decomposition).

Brownian motion (or the Wiener process) is the canonical continuous-time stochastic process with independent Gaussian increments. It is the scaling limit of random walks, serves as the driving noise in stochastic differential equations, and provides the model for diffusion, heat flow, and financial price movements. Despite almost surely being nowhere differentiable, Brownian paths are Hölder-continuous of any order less than 1/2 — a perfect instantiation of the fractal roughness of nature.

A martingale is a stochastic process whose future expected value, given all past information, equals its current value. Martingales model fair games: knowing the full history of a fair coin game gives no advantage for predicting future net winnings. Doob's theory of martingales provides powerful convergence theorems and inequalities that underpin modern probability, stochastic calculus, and mathematical finance.

Conditional expectation is one of the central constructions in modern probability theory. Given a probability space (Ω, F, P) and a sub-σ-algebra G ⊆ F, the conditional expectation E[X | G] of an integrable random variable X is the essentially unique G-measurable random variable that agrees with X on every G-measurable set in the sense of integration. This generalises the elementary notion of conditioning on an event to a fully measure-theoretic framework, and is the foundation of martingale theory, Bayesian statistics, and stochastic filtering.

A stochastic process is a family of random variables {X_t}_{t ∈ T} indexed by a parameter set T (usually time) and defined on a common probability space. The subject provides the rigorous framework for modelling random evolution: from discrete Markov chains to continuous-time diffusions, from queuing systems to financial models. Key concepts include sample paths, filtrations, stopping times, and the distinction between finite-dimensional distributions and the full process.

Ergodic theory studies the long-run average behaviour of measure-preserving dynamical systems. Its central result — Birkhoff's pointwise ergodic theorem — asserts that the time average of an observable along almost every orbit equals the space average, provided the system is ergodic. The subject connects statistical mechanics, number theory (equidistribution), combinatorics (Furstenberg's proof of Szemerédi's theorem), and information theory.

Itô calculus is the extension of classical calculus to functions of Brownian motion and, more generally, semimartingales. Because Brownian paths have infinite first-order variation but finite quadratic variation [B]_t = t, the classical Riemann–Stieltjes integral ∫ H dB fails. Itô (1944) defined a stochastic integral ∫ H_s dB_s as a limit of non-anticipating Riemann sums, and derived the cornerstone Itô formula — the stochastic analogue of the chain rule — which contains an additional second-order correction term arising from the non-zero quadratic variation.

Integration on manifolds generalises the Riemann and Lebesgue integrals to smooth manifolds. The correct objects to integrate are not functions but differential forms: an n-form on an oriented n-dimensional manifold can be integrated coordinate-independently. The fundamental theorem of calculus generalises to Stokes' theorem, which unifies the classical theorems of Green, Gauss, and Stokes. For Riemannian manifolds, the metric also gives a canonical volume form enabling integration of functions.