Loading field map
Real analysis is the branch of mathematics that grapples with the rigorous treatment of limits, continuity, differentiation, integration, and infinite series of real numbers and functions. Its history is not a steady accumulation of theorems but a series of profound shifts in what practitioners considered the fundamental objects, acceptable methods, and criteria for proof. The central tension driving these shifts has been the need to justify operations involving infinite processes—infinitesimals, infinite sums, limits—without falling into contradiction. This article traces how thirteen successive and sometimes competing frameworks have defined the field, from the intuitive calculus of the 17th century to the diverse foundational and applied schools active today.
The first framework, Infinitesimal Calculus (1670–1820), developed by Newton and Leibniz, treated derivatives and integrals as ratios of infinitely small quantities. These infinitesimals were not rigorously defined; they were manipulated as if they were ordinary numbers, then discarded at the end of a calculation. The method was extraordinarily productive for physics and geometry, but its logical foundations were shaky. Berkeley famously mocked the "ghosts of departed quantities."
Eulerian-Lagrangian Algebraic Analysis (1748–1821) took a different approach. Euler and Lagrange sought to ground calculus purely in algebraic manipulation of power series, avoiding infinitesimals altogether. For Euler, a function was an analytic expression built from algebraic operations and series expansions. He freely manipulated divergent series, treating them as formal objects, and derived results that later proved correct under more rigorous frameworks. This framework was powerful—it unified much of 18th-century mathematics—but it lacked a clear criterion for when algebraic manipulations were legitimate. Both early frameworks shared a common weakness: they relied on intuitive or formal procedures that could not reliably distinguish valid from invalid reasoning.
The first decisive break came with Cauchy-Weierstrass Epsilon-Delta Analysis (1817–present). Cauchy redefined limits, continuity, and convergence using inequalities rather than infinitesimals or series expansions. A limit was no longer a process but a static condition: for every epsilon there exists a delta. Weierstrass later refined this into the modern epsilon-delta definition, eliminating all geometric intuition. This framework superseded infinitesimal calculus by providing a purely arithmetic foundation for limits and continuity. It made possible the precise definition of the derivative and the Riemann integral.
Riemann Integration Theory (1854–1901) grew directly out of this epsilon-delta framework. Riemann defined the integral as the limit of sums of rectangles, using upper and lower sums to handle functions with infinitely many discontinuities. This definition coexisted with the classical theory of real functions, which was being developed by Dirichlet, Riemann, and others. The Classical Theory of Real Functions (1829–1905) studied pathological functions—nowhere differentiable continuous functions, functions discontinuous on dense sets—that earlier frameworks could not handle. Riemann's integral was adequate for many purposes, but it could not integrate functions with too many discontinuities, and it did not interact well with limits of sequences of functions. The classical theory revealed the need for a more powerful integration concept.
The Weierstrassian Arithmetization Program (1858–1900) went beyond epsilon-delta analysis by demanding that all concepts of analysis be reduced to properties of integers and rational numbers, eliminating any reliance on geometric intuition or spatial continuity. Weierstrass insisted that a continuous function be defined purely in terms of arithmetic conditions, not as a curve. This program superseded the earlier epsilon-delta framework by requiring a rigorous foundation for the real numbers themselves.
The Construction of the Real Numbers (1872–present) provided that foundation. Dedekind (via Dedekind cuts) and Cantor (via Cauchy sequences of rationals) independently constructed the real numbers as a complete ordered field. The key property that the rationals lack is completeness: every nonempty set bounded above has a least upper bound. This property is essential for proving the Intermediate Value Theorem, the Bolzano-Weierstrass Theorem, and the convergence of Cauchy sequences. The construction completed the arithmetization program by showing that the continuum could be built from the rationals without any geometric assumptions.
Lebesgue Integration Theory (1901–present) revolutionized integration by replacing Riemann's partition of the domain with a partition of the range. Instead of slicing the x-axis into intervals, Lebesgue sliced the y-axis and measured the set of points where the function lies between two values. This allowed integration of functions that are highly discontinuous, as long as the sets of points where the function takes certain values are measurable. The Lebesgue integral is more powerful than the Riemann integral: every Riemann-integrable function is Lebesgue-integrable, and the Lebesgue integral behaves better under limits (the Dominated Convergence Theorem).
Measure-Theoretic Analysis (1902–present) absorbed Lebesgue's ideas into a general theory of measures on sets. A measure assigns a nonnegative number to certain subsets (measurable sets) and satisfies countable additivity. This framework superseded the classical theory of real functions by providing a unified language for probability, integration, and functional analysis. It made possible the study of function spaces like Lp spaces, where functions are identified if they differ on a set of measure zero. The Lebesgue integral and measure theory together became the standard toolkit for real analysis in the 20th century.
Intuitionism (1907–present), founded by Brouwer, rejected the classical law of excluded middle and the existence of infinite sets as completed totalities. For an intuitionist, a mathematical object exists only if it can be constructed in a finite number of steps. This led to a radical revision of analysis: the real numbers are not a fixed set but a species of choice sequences, and the Intermediate Value Theorem fails in its classical form. Intuitionism was a philosophical challenge to the entire arithmetization program, arguing that its notion of existence was meaningless.
Constructive Analysis (1930–present), derived from intuitionism but less radical, was developed by Bishop and others. It retains the demand that existence proofs be algorithmic: to prove that a real number with a certain property exists, one must provide a method to compute it to any desired accuracy. Constructive analysis does not reject the law of excluded middle entirely but restricts its use to decidable properties. It preserves many classical theorems in a modified form, but it cannot prove the least upper bound property for arbitrary bounded sets. Constructive analysis coexists with measure-theoretic analysis as a minority school, valued for its algorithmic content and its connections to computer science.
Calderón-Zygmund Theory (1950–present) extends measure-theoretic analysis to the study of singular integral operators. A singular integral operator is an integral operator whose kernel has a singularity along the diagonal, such as the Hilbert transform or the Riesz transforms. These operators arise naturally in partial differential equations and harmonic analysis. Calderón and Zygmund developed a method to prove Lp boundedness for such operators using a decomposition of the function into a "good" part and a "bad" part, combined with estimates on the measure of level sets. This theory sits within measure-theoretic analysis but adds new techniques for handling singularities that classical measure theory alone could not address. It remains an active research area with applications to elliptic PDEs and signal processing.
Nonstandard Analysis (1961–present), developed by Robinson, reacted against the Weierstrassian arithmetization program by reintroducing infinitesimals in a rigorous way. Using model theory, Robinson constructed an extension of the real numbers, the hyperreals, which contains infinitesimal and infinite numbers. A function is continuous at a point if its hyperreal extension maps infinitesimally close points to infinitesimally close values. This framework retains all the theorems of classical analysis (by the transfer principle) but allows proofs that resemble the original infinitesimal calculus. Nonstandard analysis does not replace epsilon-delta proofs; it provides an alternative, often more intuitive, route to the same results. It coexists with measure-theoretic analysis as a minority approach, valued in certain areas of probability theory and economics.
Today, measure-theoretic analysis, built on the Lebesgue integral and the construction of the real numbers, is the dominant framework in real analysis. It is the default language in graduate textbooks, research in harmonic analysis, partial differential equations, and probability. The Cauchy-Weierstrass epsilon-delta framework remains the standard for introductory analysis courses, providing the rigorous foundation for limits and continuity. The construction of the real numbers is taught as the bedrock of the subject.
The leading frameworks agree on the core results: the completeness of the real numbers, the convergence of Cauchy sequences, the properties of the Lebesgue integral, and the basic theorems of measure theory. They disagree on foundational questions: whether existence proofs must be constructive, whether infinitesimals are a legitimate tool, and whether the law of excluded middle is universally valid. These disagreements are not settled; they reflect different philosophical commitments and different practical needs. Constructive analysis and nonstandard analysis remain active minority schools, each offering insights that the mainstream framework does not. Intuitionism, while less active as a research program in analysis, continues to influence the philosophy of mathematics. The history of real analysis is thus not a story of a single victorious framework but of a dominant paradigm that coexists with persistent alternatives, each addressing aspects of the subject that the others leave unexplored.