Loading field map
Knot theory is the branch of topology that studies the embeddings of a circle in three-dimensional space. Its central object is the knot: a closed, non-self-intersecting curve in space that cannot be untangled to a plain circle without cutting it. Two knots are considered equivalent if one can be continuously deformed into the other without passing the curve through itself—a process mathematicians call an ambient isotopy. The field's fundamental question is deceptively simple: given two knot diagrams, how can one tell whether they represent the same knot?
The core difficulty of knot theory is that the same knot can be drawn in infinitely many ways. A trefoil knot—the simplest nontrivial knot, resembling a three-leaf clover—can be drawn with three crossings or with dozens, with loops pulled tight or stretched loosely across the page. The equivalence relation of ambient isotopy is geometrically intuitive but computationally intractable to check directly. The field therefore develops around two complementary strategies: invariants and moves.
An invariant is any mathematical quantity or structure assigned to a knot that does not change under ambient isotopy. If two knots have different invariants, they are certainly different knots. The simplest invariants are numerical: the crossing number (the minimum number of crossings in any diagram of the knot), the unknotting number (the minimum number of crossing changes needed to turn the knot into the unknot), and the bridge number (related to the number of local maxima in a projection). These are easy to define but often hard to compute, and they are rarely powerful enough to distinguish all knots.
The first genuinely useful invariant was the Alexander polynomial, introduced in the 1920s. It is a polynomial associated to a knot that can be computed from any diagram using a simple set of rules. The Alexander polynomial distinguishes many pairs of knots, but it fails to distinguish others—most famously, it cannot tell a knot from its mirror image. For decades, the Alexander polynomial, together with its generalizations, formed the backbone of the subject.
For most of its history, knot theory worked primarily with knot diagrams: planar drawings of a knot with small gaps indicating which strand passes over another. The field's foundational result in this tradition is the Reidemeister theorem (1920s), which states that two diagrams represent the same knot if and only if they can be connected by a finite sequence of three local moves—the Reidemeister moves. These moves correspond to the simplest ways a strand can be pulled over, under, or past another in a planar projection.
The Reidemeister theorem transforms a geometric problem into a combinatorial one: instead of thinking about continuous deformations of curves in space, one can work with diagrams and local rewriting rules. This shift made knot theory accessible to computation and to algebraic methods. The theorem also underlies the definition of many invariants: if a quantity computed from a diagram is unchanged by all three Reidemeister moves, it is automatically a knot invariant.
The diagrammatic tradition reached a high point with the Jones polynomial (1984), discovered by Vaughan Jones while working on operator algebras. The Jones polynomial is defined by a simple skein relation: a rule that relates the polynomial of a knot to the polynomials of two knots obtained by changing a single crossing. This relation, together with a normalization for the unknot, determines the polynomial for every knot. The Jones polynomial proved remarkably powerful, distinguishing knots that the Alexander polynomial could not, and it revealed unexpected connections to statistical mechanics and quantum field theory. Its discovery triggered an explosion of new polynomial invariants—the HOMFLY-PT polynomial and the Kauffman bracket being the most prominent—all defined through skein relations.
A second major tradition treats knots not as diagrams but as objects embedded in a specific three-dimensional space, usually the 3-sphere (the one-point compactification of ordinary space). This approach, which became dominant in the second half of the twentieth century, uses the tools of geometric topology and hyperbolic geometry.
The key figure in this tradition is William Thurston, whose geometrization conjecture (proved by Grigori Perelman in 2003) implies that every knot complement—the space obtained by removing the knot from the 3-sphere—can be decomposed into pieces, each of which admits one of eight geometric structures. For the vast majority of knots, the relevant structure is hyperbolic geometry: the knot complement can be given a complete, finite-volume hyperbolic metric. This is a stunning fact: a purely topological object (a knot) carries a canonical geometric structure.
The hyperbolic structure provides powerful invariants. The hyperbolic volume of a knot complement is a real number that is a knot invariant, and it is often effective at distinguishing knots. More refined invariants, such as the canonical decomposition of the complement into ideal tetrahedra, can completely determine the knot. The geometric approach also yields the Geometrization Theorem for knots, which gives a complete classification of knots whose complements are not hyperbolic: these are precisely the torus knots (which lie on the surface of a standard torus) and the satellite knots (which are constructed by wrapping one knot around another). This trichotomy—hyperbolic, torus, satellite—is one of the deepest structural results in the field.
The geometric and diagrammatic traditions are not rivals but complementary tools. The Jones polynomial, for instance, was later shown to be related to the hyperbolic volume through the volume conjecture, which states that a certain limit of evaluations of the colored Jones polynomial recovers the hyperbolic volume. The conjecture remains open in general but is supported by extensive evidence. This kind of cross-fertilization—where a combinatorial invariant encodes geometric information—is characteristic of modern knot theory.
A third major approach, which emerged in the late twentieth century, treats knot invariants as algebraic or categorical constructions. The central idea is that a knot invariant can be defined not by a formula but by a functor: a structure-preserving map from a category of tangles (generalizations of knots with free ends) to a category of vector spaces or modules.
The most influential framework here is quantum topology, which constructs invariants from representations of quantum groups—deformations of classical Lie algebras. The Jones polynomial, it turns out, is the first in an infinite family of quantum invariants (the colored Jones polynomials), each associated to a different representation of the quantum group associated to the Lie algebra sl(2). These invariants are defined through Reshetikhin–Turaev constructions, which assign to a knot diagram a linear map between vector spaces, with the Reidemeister moves corresponding to natural isomorphisms.
A related but distinct development is Khovanov homology (2000), which categorifies the Jones polynomial. Instead of assigning a polynomial to a knot, Khovanov homology assigns a chain complex—a sequence of vector spaces connected by linear maps—whose graded Euler characteristic recovers the Jones polynomial. The homology itself is a strictly stronger invariant: it distinguishes knots that have the same Jones polynomial. Khovanov homology opened the door to a family of homological invariants (including knot Floer homology, defined by Ozsváth and Szabó and independently by Rasmussen) that have proved extraordinarily powerful. These invariants are not merely stronger versions of polynomials; they carry additional algebraic structure, such as spectral sequences relating different theories, and they have deep connections to symplectic geometry and to the topology of 4-manifolds.
The categorical approach has also led to the concept of a knot invariant as a TQFT (topological quantum field theory). A TQFT assigns vector spaces to surfaces and linear maps to cobordisms between them, and it provides a systematic way to construct invariants of knots embedded in 3-manifolds. The Witten–Reshetikhin–Turaev invariants of 3-manifolds, which generalize the Jones polynomial, are the most famous examples. These constructions reveal that knot theory is not an isolated subject but a special case of a much larger framework connecting topology, algebra, and physics.
Contemporary knot theory is characterized by the productive interaction of all these traditions. The diagrammatic, geometric, and algebraic approaches are not historical stages but active research programs that inform one another. A typical modern result might use hyperbolic geometry to compute a geometric invariant, then show that this invariant is determined by a Khovanov-theoretic quantity, then use the algebraic structure to extract new information.
Several enduring questions organize the field. The unknotting problem—deciding whether a given diagram represents the unknot—is known to be decidable but is not known to be in the complexity class P; the best known algorithms are exponential in the crossing number. The slice-ribbon conjecture asks whether every knot that bounds a smoothly embedded disk in the 4-ball also bounds a ribbon disk (a disk with only certain simple singularities); this remains open and is one of the central problems connecting knot theory to 4-dimensional topology. The volume conjecture and the related AJ conjecture seek to connect quantum invariants to geometric ones. And the classification problem itself—finding a complete, computable set of invariants that distinguishes all knots—remains unresolved, though the combination of hyperbolic invariants and homological invariants comes close in practice.
The field also maintains a practical dimension. Knot tables—systematic lists of knots ordered by crossing number—have been compiled since the late nineteenth century and now include all knots up to 16 crossings (over 1.7 million of them). These tables serve as a testing ground for new invariants and as a reference for applications. Knot theory has found unexpected uses in DNA topology (where enzymes change the linking of circular DNA molecules), in statistical mechanics (where knot invariants arise as partition functions of lattice models), and in quantum computing (where braid group representations, closely related to knot invariants, provide a model for topological quantum computation).
Throughout its history, knot theory has been driven by a single underlying tension: the objects it studies are simple to draw and describe, yet the equivalence relation that defines them is subtle enough that no complete invariant is known. Each major approach—diagrammatic, geometric, algebraic—has provided partial answers, and their combination has produced a rich and interconnected theory. The field's durability comes from this very plurality: no single method has proven sufficient, and the interplay between them continues to generate both new questions and new tools.