Loading field map
Topology is the branch of mathematics that studies the properties of spaces that are preserved under continuous deformations—stretching, bending, twisting, and crumpling—but not tearing, cutting, or gluing. A classic way to picture the subject is to imagine shapes made of infinitely stretchable rubber: a coffee mug and a doughnut are topologically the same because one can be deformed into the other without breaking, while a sphere and a doughnut are different because turning a sphere into a doughnut would require punching a hole. This informal image captures the spirit, but the actual discipline is far more general and rigorous, dealing with spaces that have no notion of distance, angle, or straightness at all.
The central question of topology is: When are two spaces equivalent? The answer depends on what kind of equivalence one cares about. The most fundamental notion is homeomorphism: a bijective map between two spaces that is continuous in both directions. Two spaces are homeomorphic if one can be continuously deformed into the other and back again. But topology also studies weaker or stronger invariants—properties that are preserved under homeomorphism—such as connectedness, compactness, and the number of holes in a space. Much of the subject is devoted to finding ways to distinguish spaces that are not homeomorphic, and to classifying spaces up to homeomorphism or up to some coarser equivalence.
Topology did not begin as a self-conscious discipline. Its roots lie in the eighteenth and nineteenth centuries in problems from analysis, particularly the study of functions and their behavior near points. The modern definition of continuity, formulated in the nineteenth century, made precise the idea that a function is continuous if small changes in the input produce small changes in the output. This definition does not require a notion of distance in the abstract; it only requires a way to say which points are "near" each other. The realization that continuity could be studied independently of any particular metric—any particular way of measuring distance—was a crucial step.
The term "topology" itself was coined in the nineteenth century, but the subject's early development was driven by specific problems. One important thread came from the study of graphs and polyhedra, where mathematicians asked combinatorial questions about vertices, edges, and faces. Leonhard Euler's solution to the Königsberg bridge problem in the eighteenth century—showing that a certain walk crossing each of seven bridges exactly once was impossible—is often cited as an early topological result, because it depends only on the arrangement of connections, not on distances. Another thread came from the study of surfaces in three-dimensional space, where mathematicians began to ask whether two surfaces could be continuously deformed into each other.
A decisive moment came in the late nineteenth century with the work of Henri Poincaré, who is often regarded as the founder of algebraic topology. Poincaré introduced the idea of associating algebraic objects—such as groups—to topological spaces, so that topological questions could be translated into algebraic ones. His work on the fundamental group assigned to each space a group that records the ways loops in the space can be deformed into one another. For a sphere, every loop can be shrunk to a point, so the fundamental group is trivial. For a doughnut, a loop that goes around the hole cannot be shrunk away, and the fundamental group is the infinite cyclic group. Poincaré's insight was that these algebraic invariants could distinguish spaces that were otherwise hard to tell apart.
The first major branch of topology to be systematized is general topology, also called point-set topology. This is the study of topological spaces themselves, defined axiomatically. A topological space is a set of points together with a collection of subsets, called open sets, that satisfy a few axioms: the empty set and the whole space are open; the union of any collection of open sets is open; and the intersection of finitely many open sets is open. This definition, crystallized in the early twentieth century, is astonishingly general. It includes familiar spaces like the real line and Euclidean space, but also spaces with no metric, spaces with infinitely many dimensions, and spaces whose points are functions or sets.
General topology addresses foundational questions: What does it mean for a sequence to converge? When is a space connected? When is it compact? Compactness is a topological generalization of the idea of being "finite in extent"; in Euclidean space, a set is compact exactly when it is closed and bounded, but in general spaces the definition is more subtle. Connectedness captures whether a space can be split into two disjoint nonempty open pieces. These properties are not just abstract curiosities; they underpin much of analysis. For instance, a continuous function on a compact space attains its maximum and minimum, and a continuous function on a connected space has the intermediate value property.
General topology also studies separation axioms, which describe how finely a topology can distinguish points. At one extreme, a space may have the property that any two distinct points can be separated by disjoint open sets (the Hausdorff property); at the other extreme, a space may be so coarse that no two points can be separated at all. These distinctions matter because many theorems in analysis require a space to be Hausdorff, while other constructions—such as the Zariski topology used in algebraic geometry—deliberately violate it.
The development of general topology was closely tied to the rise of set theory and the axiomatic method. It provided a common language for analysis, geometry, and later for functional analysis, where spaces of functions themselves become topological spaces. Its methods are primarily descriptive and classificatory: one defines a property, proves that it is preserved under certain operations, and uses it to distinguish or identify spaces.
If general topology asks what a space is, algebraic topology asks how spaces can be distinguished and classified. Its strategy is to assign to each topological space a sequence of algebraic objects—groups, rings, vector spaces—in a way that is invariant under homeomorphism. The fundamental group is the first and most intuitive example, but it is only the beginning.
The central tools of algebraic topology are homotopy groups and homology groups. The idea of homotopy is to consider continuous deformations of maps. Two maps are homotopic if one can be continuously deformed into the other. The fundamental group classifies loops up to homotopy; higher homotopy groups classify maps from spheres of higher dimension into the space. Homology groups, by contrast, are computed from a space by breaking it into simple pieces—triangles, tetrahedra, and their higher-dimensional analogues—and then counting the "holes" of each dimension. The zeroth homology group counts connected components, the first homology group counts one-dimensional holes (like the hole in a doughnut), the second counts two-dimensional voids (like the inside of a hollow sphere), and so on.
Homology and homotopy are related but not identical. Homology is generally easier to compute and is a functor: it turns topological maps into algebraic homomorphisms in a way that respects composition. Homotopy groups are more sensitive but notoriously difficult to compute; even the homotopy groups of spheres are not fully known. The relationship between the two is governed by the Hurewicz theorem, which states that the first nonzero homotopy group of a simply connected space agrees with the corresponding homology group under a natural map.
A major achievement of algebraic topology in the mid-twentieth century was the development of cohomology, a dual theory that assigns to a space a sequence of algebraic objects with a multiplicative structure. Cohomology is not just a mirror image of homology; it carries additional information—the cup product—that allows one to distinguish spaces that homology alone cannot. Cohomology also has a natural interpretation in terms of "obstructions": it measures the extent to which certain constructions on a space fail to exist.
Algebraic topology is not a single method but a family of related techniques. Simplicial complexes break a space into combinatorial pieces; CW complexes are a more flexible generalization that allows cells of various dimensions to be attached. The Eilenberg–Steenrod axioms of the 1940s codified what a homology theory must satisfy, showing that many different constructions all yield the same answer on the spaces for which they are defined. This axiomatic viewpoint clarified the subject and made it possible to prove theorems about all homology theories at once.
A third major branch, differential topology, studies spaces that carry a smooth structure—a way of doing calculus on them. These spaces, called smooth manifolds, are the natural setting for much of geometry and physics. A smooth manifold is a topological space that locally looks like Euclidean space, with transition maps between overlapping local descriptions that are infinitely differentiable. The real line, the circle, the sphere, and the doughnut are all smooth manifolds; so are higher-dimensional analogues and more exotic spaces.
Differential topology asks questions that are topological in spirit but use smooth tools. The most famous example is the Poincaré conjecture, proved by Grigori Perelman in the early 2000s, which states that every simply connected closed three-dimensional manifold is homeomorphic to the three-dimensional sphere. The proof used Ricci flow, a differential equation that deforms the metric on a manifold, showing how analytic methods can resolve topological questions.
A central concept in differential topology is transversality: two submanifolds intersect transversally if their tangent spaces together span the tangent space of the ambient manifold at each intersection point. Transversal intersections are "generic"—they can be achieved by small perturbations—and they behave well under deformation. This idea underlies Morse theory, which studies a smooth function on a manifold by examining its critical points. Morse theory shows that the topology of a manifold can be reconstructed from the critical points of a generic smooth function, providing a powerful bridge between analysis and topology.
Differential topology also studies cobordism: two manifolds are cobordant if their disjoint union is the boundary of a higher-dimensional manifold. Cobordism is a much coarser equivalence than homeomorphism, but it has a rich algebraic structure and is deeply connected to stable homotopy theory. The classification of manifolds up to cobordism was achieved in the mid-twentieth century and is one of the great successes of the field.
The relationship between differential topology and algebraic topology is intimate. Every smooth manifold has an underlying topological space, and the algebraic invariants of that space are often computable using smooth tools. Conversely, algebraic invariants can obstruct the existence of smooth structures: some topological manifolds admit no smooth structure at all, and some admit many inequivalent ones. The discovery in the 1950s that spheres of high dimension can carry multiple distinct smooth structures was a shock and remains a central theme.
Geometric topology is the study of manifolds in low dimensions—primarily dimensions two, three, and four—with an emphasis on understanding their actual shape and structure rather than just their algebraic invariants. Low dimensions are special because they are low enough to be visualized and studied by hand, yet high enough to be genuinely complicated.
In dimension two, the classification of closed surfaces is a classical result: every closed surface is homeomorphic to a sphere with some number of handles (orientable) or a sphere with some number of crosscaps (non-orientable). This classification, known since the nineteenth century, is one of the oldest and most satisfying theorems in topology. It shows that the fundamental group, or equivalently the first homology group, completely determines a closed surface up to homeomorphism.
Dimension three is far richer. The geometrization conjecture, proved by Perelman as part of his work on the Poincaré conjecture, states that every closed three-manifold can be decomposed into pieces, each of which admits one of eight geometric structures. This result gives a complete classification of three-manifolds in a sense, though the classification is far more complex than in dimension two. The eight geometries include the familiar Euclidean, spherical, and hyperbolic geometries, as well as five more exotic ones. The hyperbolic case is particularly important: most three-manifolds are hyperbolic, and the Mostow rigidity theorem states that a hyperbolic structure on a closed three-manifold is unique, so topological and geometric information coincide.
Dimension four is the most mysterious. The classification of smooth four-manifolds is wide open, and the subject is marked by striking phenomena. In dimensions five and higher, the h-cobordism theorem and related results provide a relatively complete classification of simply connected manifolds, but these methods fail in dimension four. The discovery of exotic smooth structures on four-dimensional Euclidean space—smooth structures that are homeomorphic but not diffeomorphic to the standard one—shows that dimension four is genuinely different. The tools used to study four-manifolds, such as Donaldson theory and Seiberg–Witten theory, come from mathematical physics and involve deep connections to gauge theory.
Geometric topology is not a separate tradition from algebraic or differential topology; rather, it is a field of application where the full power of all methods is brought to bear on specific spaces. Its results often have a concrete, almost tangible character: one can draw pictures of knots and surfaces, and the questions are often motivated by physical intuition about the shape of space.
These branches—general, algebraic, differential, and geometric topology—are not rival schools but complementary approaches to a common set of questions. A working topologist typically moves freely among them. General topology provides the foundational language; algebraic topology provides the most powerful invariants; differential topology provides tools for studying smooth spaces; and geometric topology focuses attention on the low-dimensional cases where the most detailed understanding is possible.
The boundaries between topology and neighboring fields are also porous. Algebraic geometry uses topological methods to study the shapes defined by polynomial equations, and its étale cohomology is a topological invariant adapted to algebraic settings. Functional analysis uses topological ideas constantly, and the theory of operator algebras has deep connections to topology through the classification of certain infinite-dimensional spaces. Mathematical physics has been a major source of topological ideas since the late twentieth century, from gauge theory in four dimensions to the study of topological phases of matter, where physical systems are classified by topological invariants.
One of the most striking features of modern topology is the way it has absorbed ideas from outside mathematics. The Jones polynomial, discovered in the 1980s, is a knot invariant that arose from operator algebras and statistical mechanics; it led to a burst of activity in knot theory and to the discovery of quantum invariants of three-manifolds. The Floer homology theories, developed in the 1980s and 1990s, use infinite-dimensional Morse theory to define invariants of three- and four-manifolds. These developments have not replaced the classical tools but have enriched them, creating a web of connections that spans much of mathematics.
The present landscape of topology is characterized by both depth and breadth. On the foundational side, homotopy type theory and the univalence axiom have proposed a new foundation for mathematics based on topological intuition, connecting topology to logic and computer science. On the applied side, topological data analysis uses persistent homology to extract shape information from large data sets, a development that has brought topological ideas into statistics and machine learning. These applications are not the core of the subject, but they illustrate the reach of topological thinking.
Topology remains a living field because its central question—what is the shape of a space, and how can we tell?—is both simple and inexhaustible. The subject has grown from a collection of isolated problems in analysis and geometry into a vast and interconnected discipline, but its core insight has remained constant: the properties of a space that survive continuous deformation are often the most fundamental ones, and they can be studied with remarkable precision.