Real Analysis

Real analysis is the study of the real number line and of functions defined on it: what it means for a sequence to approach a limit, for a function to be continuous, for a curve to have a length or a region an area, and for a sum of infinitely many terms to settle on a finite value.

9 frameworks, from 1854 to today · Analysis · Mathematics

still a working view ended · each leads to its entry below

Overview

Real analysis is the study of the real number line and of functions defined on it: what it means for a sequence to approach a limit, for a function to be continuous, for a curve to have a length or a region an area, and for a sum of infinitely many terms to settle on a finite value1. It is the branch of mathematics that put the calculus of Newton and Leibniz on secure foundations, and in doing so it created tools—the precise limit, the measure of a set, the space of functions—that spread across mathematics1. Its history is largely a history of making intuition rigorous, and then discovering that the rigor forced on mathematicians ideas no one had imagined.

Calculus without foundations

When Newton and Leibniz developed the calculus in the late seventeenth century, they worked with "infinitesimals"—quantities smaller than any positive number yet not zero. The results were spectacular: planetary orbits, fluid flow, optimization, the whole machinery of derivatives and integrals. The logic was shaky. Bishop George Berkeley's 1734 polemic The Analyst asked pointedly what these vanishing quantities were, and neither Newton's "fluxions" nor Leibniz's differentials answered him. Eighteenth-century analysts, including Euler and the Bernoullis, made progress by treating infinite series and integrals with remarkable boldness and occasional error, trusting algebraic manipulation to carry them. D'Alembert's formulation of the limit as a concept, and Euler's and Lagrange's efforts to base calculus on algebra, pointed toward a fix, but the fix would take a century.

Epsilon-delta: making the limit precise

The decisive turn came in the early nineteenth century, led above all by Augustin-Louis Cauchy. In his Cours d'analyse (1821) and later works, Cauchy defined the limit of a function and used it to define continuity and the derivative, and he gave the first systematic treatment of the definite integral as a limit of sums. Cauchy's own language still leaned on infinitesimals and on "infinitely small" increments, but the structure was there23.

Bernard Bolzano had anticipated some of this, notably in an 1817 proof of the intermediate value theorem, but Bolzano's work was little known in his lifetime24. Cauchy's was read everywhere.

The final polishing into the now-standard "epsilon-delta" form—for every ε > 0 there is a δ > 0 such that…—is associated with Karl Weierstrass, whose Berlin lectures from the 1860s onward (published later by students) recast limits, continuity, and convergence entirely in terms of inequalities between real numbers. The definition of a limit that every student now meets is essentially Weierstrass's. With it, a claim like "f is continuous at a" became a statement about numbers that could be checked and proved, and the vague talk of infinitely small increments could be retired from rigorous arguments.

The rigor immediately paid off in surprising counterexamples. Weierstrass exhibited, in 1872, a function continuous everywhere but differentiable nowhere—a curve with no tangent line at any point. This shattered the intuition, inherited from the eighteenth century, that a continuous curve must be smooth except perhaps at isolated points. It showed that the new definitions described a far stranger world than the founders of calculus had assumed.

Building the real numbers and the strange sets of the line

Epsilon-delta arguments presuppose that the real numbers themselves are well defined. What is a real number? In the 1870s three answers appeared almost at once. Weierstrass's student Eduard Heine and independently Georg Cantor characterized the reals as limits of Cauchy sequences of rationals. Richard Dedekind, in 1872, introduced what we now call Dedekind cuts: a real number is a partition of the rationals into a lower and an upper set with no largest element below. Cantor's approach by Cauchy sequences and Dedekind's by cuts gave the continuum a construction from the rationals, completing the "arithmetization of analysis"—the program of grounding all of analysis in arithmetic rather than geometry.

Cantor then went further. In 1874 he proved that the real numbers cannot be listed in a sequence—there are strictly more reals than natural numbers—and in the 1880s he developed the theory of infinite cardinalities and of point sets: closed sets, perfect sets, and the beginnings of what would become topology. His work on the structure of sets of real numbers, at first controversial, became the seedbed for measure theory.

Meanwhile Weierstrass's and Cantor's circle produced functions and sets that defeated old assumptions about integration. Riemann had defined the integral in the 1850s by dividing the domain into small intervals and taking a limit of rectangular sums; this Riemann integral handles a broad class of functions but fails badly on wilder ones. Around 1875, Vito Volterra and others produced functions that are bounded and not Riemann-integrable, and Cantor's sets of measure—intuitively "length"—zero raised the question of which sets could sensibly be measured at all.

Measure and the Lebesgue integral

The answer came from Henri Lebesgue. In his 1902 thesis Intégrale, longueur, aire, Lebesgue built a theory of measure: a systematic assignment of "length," "area," or "volume" to sets, obeying natural rules (the measure of a union of disjoint pieces is the sum of their measures, the measure of a subpiece is no larger, and so on)5. He then defined the integral not by chopping the horizontal axis (the domain) into intervals as Riemann had, but by chopping the vertical axis (the range) into levels and measuring the sets where the function lies between consecutive levels.

The payoff was large. The Lebesgue integral integrates every nonnegative function for which the area makes any sense, not just those Riemann could handle; it behaves well under limits (the theorems of Fatou, and of the monotone and dominated convergence, tell exactly when the integral of a limit is the limit of the integrals); and the resulting space of integrable functions became the prototype of a complete normed space. Later work, notably by Émile Borel (whose measure theory Lebesgue built on), Constantin Carathéodory (who gave an abstract criterion for measurability in 1914), and in the 1910s and 1920s by the Russian school led by Dmitri Egorov and Nikolai Luzin, extended and deepened it.

The theory of measure also forced the discovery that some sets cannot be measured consistently. Using a nonconstructive argument based on choosing one representative from each class of reals differing by a rational, Giuseppe Vitali showed in 1905 that there exist sets to which no length can be assigned without contradiction. This "Vitali set" made explicit a trade-off, sharpened by the 1924 Banach–Tarski paradox: a domain of mathematics in which everything can be measured is not freely available.

The great generalization: abstraction and function spaces

A central lesson of Lebesgue's work was that many arguments depend less on the specific line than on a general notion of "space" and "size." Maurice Fréchet's 1906 thesis introduced metric spaces, where distance is abstracted to a function satisfying a few axioms; Stefan Banach and Hans Hahn's work in the 1920s and early 1930s built the theory of complete normed vector spaces (Banach spaces), and Stefan Banach's 1932 monograph Théorie des opérations linéaires became the reference. These ideas—of treating a whole function as a single point in a space, of measuring distance between functions—are the beginnings of functional analysis, which real analysis and its descendants fed directly.

Real-variable methods then produced spectacular tools. Hardy and Littlewood in the 1910s and 1920s developed the theory of maximal functions and the Hardy–Littlewood maximal inequality; this was the seed of real-variable harmonic analysis, where results that had been proved for Fourier series by complex-analytic or very special methods were reproved and extended by covering arguments, singular integral operators, and measure theory. Antoni Zygmund and his student Alberto Calderón in the 1950s and 1960s pushed the singular integral method far—the Calderón–Zygmund theory of what are now called singular integrals—and Charles Fefferman's work in the 1970s extended it to several variables. What had been a subject about Fourier series became a general theory of operators on function spaces.

Other currents: intuitionism, constructivism, and the infinities

Not everyone accepted the classical framework. L. E. J. Brouwer, from 1907 onward, founded mathematical intuitionism, rejecting the principle of excluded middle ("every statement is either true or false") and nonconstructive existence proofs. In Brouwer's view a theorem asserting that something exists must provide a method for constructing it; he accordingly denied results proved by contradiction alone. A major consequence is that, in intuitionistic analysis, the intermediate value theorem fails in general: one cannot always construct the root whose existence the classical theorem asserts. The tradition survived largely through Arend Heyting's formalization of intuitionistic logic in 1930 and, later, through the school of Errett Bishop, whose 1967 Foundations of Constructive Analysis rebuilt large parts of real analysis dwelling on what could be proved constructively. Per Martin-Löf's type theory in the 1970s and the rise of computer proof assistants later gave constructivism a practical life it had lacked.

A very different challenge came from Abraham Robinson. In 1960 he showed that the infinitesimals of the founders could be made logically respectable: one can construct a "nonstandard" model of the reals containing infinitesimal and infinitely large elements, in which the familiar theorems of analysis all hold. This is nonstandard analysis6. It vindicated part of the old intuition—Leibniz's differentials could be treated as genuine objects, not fictions—and has proved a useful technique in some corners of analysis and beyond.

The present

Real analysis today is less a single subject than a broad toolkit. A student meeting it learns the Wiener-era canon: sequences and series, continuity, the Riemann integral and then the Lebesgue integral, metric spaces, and perhaps some Fourier analysis. Working analysts use many of the branches above: measure theory is foundational almost everywhere; functional analysis supplies the setting for differential equations, probability, and operator theory; real-variable harmonic analysis, from the Hardy–Littlewood maximal theorem to Calderón–Zygmund theory, is a living field in its own right; and constructive and nonstandard methods continue as specialist currents. The questions that remain open are often technical versions of the old wonder—how irregular can a function or a set be, how much can be measured, which limits survive under which operations—and much of modern analysis (partial differential equations, geometric measure theory, probability) is real analysis extended to new arenas.

References

  1. Real analysis (introduction, Limits and convergence, Relation to complex analysis), Wikipedia
  2. Limit of a function (History), Wikipedia
  3. Continuous function (History), Wikipedia
  4. Intermediate value theorem (History), Wikipedia
  5. Lebesgue measure, Wikipedia
  6. Hyperreal number, Wikipedia

The frameworks of Real Analysis

Each framework is a way of working the field has used: when it held, and the question it answers.

  1. 1854-1950

    Riemann integral

    How do we define the area under a curve?

    Fly to its tower in Nousfarer

  2. 1860-Present

    Epsilon-Delta Rigorization

    How can limits be defined without infinitesimals?

    Fly to its tower in Nousfarer

  3. 1870-1930
  4. 1874-Present

    Point-Set Theory and Descriptive Set Theory

    What structure do sets of points on a line have?

    Fly to its tower in Nousfarer

  5. 1902-Present

    Measure Theory and the Lebesgue Integral

    Which sets have a well-defined length or area?

    Fly to its tower in Nousfarer

  6. 1906-Present

    Metric and Banach Space Theory

    What can be said about spaces of functions as wholes?

    Fly to its tower in Nousfarer

  7. 1907-Present

    Constructive analysis

    Which existence proofs actually construct their objects?

    Fly to its tower in Nousfarer

  8. 1910-Present

    Real-Variable Harmonic Analysis

    How can Fourier-type results be proved by real-variable methods?

    Fly to its tower in Nousfarer

  9. 1960-Present

    Nonstandard Analysis

    Can infinitesimals be made logically rigorous?

    Fly to its tower in Nousfarer

How the ideas connect

Each arrow runs from an earlier idea to the later one it shaped. The list below says what passed between them.

Riemann integralEpsilon-Delta RigorizationConstruction of the real numbersPoint-Set Theory and Descriptive Set …Measure Theory and the Lebesgue Integ…Metric and Banach Space TheoryConstructive analysisReal-Variable Harmonic AnalysisNonstandard Analysis

Beyond the field

Where Real Analysis's ideas came from in other fields, and what they led to there: the strongest 12 of 66 links.

A city in Nousfarer: a red-carpeted stair climbing terraces to a crowned tower on an acropolis, more towers round it, forest and snowy mountains beyond.
A city in Nousfarer, the game played in the browser.

Real Analysis is a planet in Nousfarer

Its city has a tower for each of these 9 frameworks. Land, walk in, and climb one concept at a time.

Land on Real Analysis

Related fields

Fields its ideas are linked to

More in Analysis

See where Real Analysis sits on the Map of Human Knowledge