Loading field map
Geometric group theory is the study of groups—algebraic structures that encode symmetry—through the geometric and topological properties of spaces on which they act. Its central move is to trade an abstract algebraic object for a concrete geometric one: given a finitely generated group, one constructs a metric space, the Cayley graph, whose vertices are the group elements and whose edges connect elements that differ by a generator. The group then acts on this graph by left multiplication, preserving distances. The insight that launched the field is that many algebraic properties of the group are reflected in the large-scale geometry of this space, and conversely, that geometric features of the space can be translated back into algebraic statements about the group.
The field's founding question is deceptively simple: what can be learned about a group from the shape of its Cayley graph, when that shape is viewed only at large scales? Because the choice of generating set changes the Cayley graph's local details—edge lengths, angles, small cycles—one must ignore such fine structure. The correct equivalence relation is quasi-isometry: two metric spaces are quasi-isometric if, roughly, they look the same when viewed from very far away, with distances distorted by at most a bounded multiplicative factor and an additive constant. Different generating sets for the same group yield quasi-isometric Cayley graphs, so quasi-isometry type is an invariant of the group itself, not of the generating set. The field's program is to determine which group-theoretic properties are quasi-isometry invariants, and to classify groups up to quasi-isometry.
The roots of geometric group theory lie in late nineteenth-century work on discrete groups of motions of Euclidean and hyperbolic space, particularly the study of fundamental domains and the realization that a group acting properly discontinuously on a space could be studied through the quotient space. But the modern field crystallized in the 1980s, when several strands converged. The work of Mikhael Gromov on hyperbolic groups provided a unified framework for understanding groups whose Cayley graphs resemble negatively curved spaces. Simultaneously, the development of the theory of automatic groups and the growing use of computer science concepts like regular languages to describe group structure gave the subject computational tools. The field also absorbed deep influences from low-dimensional topology, especially through the study of 3-manifold groups and the mapping class groups of surfaces.
A key early result that illustrates the field's method is Gromov's polynomial growth theorem: a finitely generated group has polynomial growth—meaning the number of elements within distance n of the identity grows at most polynomially—if and only if it is virtually nilpotent, that is, contains a nilpotent subgroup of finite index. This theorem is striking because it shows that a purely geometric quantity, the growth rate of balls in the Cayley graph, completely determines a purely algebraic class of groups. The proof uses the geometry of the Cayley graph to derive algebraic structure, and the result remains a cornerstone of the field.
The most influential class of groups in the field is that of hyperbolic groups, introduced by Gromov. A finitely generated group is hyperbolic if its Cayley graph is δ-hyperbolic for some constant δ: every geodesic triangle in the graph has its sides within distance δ of each other, meaning that triangles are uniformly thin. This condition is a coarse version of negative curvature. Hyperbolic groups include free groups, fundamental groups of compact hyperbolic manifolds, and many small cancellation groups. They have a rich theory: they have solvable word problem, their boundary at infinity carries a natural topology, and they exhibit exponential growth unless they are virtually cyclic.
The importance of hyperbolic groups lies partly in their tractability and partly in their role as a testing ground for general conjectures. Many questions that remain open for general groups have been answered in the hyperbolic case. For instance, the isomorphism problem—deciding whether two presentations define the same group—is solvable for hyperbolic groups, a result due to Zlil Sela. The theory of hyperbolic groups also introduced the concept of a boundary: the set of geodesic rays emanating from a basepoint, modulo the relation of staying within bounded distance of each other. This boundary is a topological space that encodes information about the group's large-scale structure, and its homeomorphism type is a quasi-isometry invariant.
Hyperbolic groups are too restrictive for many purposes. Many natural groups, such as fundamental groups of finite-volume hyperbolic manifolds with cusps, are not hyperbolic but have a clear geometric structure: they contain peripheral subgroups that are "parabolic" and the rest of the group behaves hyperbolically away from them. This led to the theory of relatively hyperbolic groups, developed by Gromov and refined by others. A group is relatively hyperbolic with respect to a collection of subgroups if its Cayley graph, after coning off each coset of a peripheral subgroup, becomes hyperbolic. This framework has been essential for studying lattices in semisimple Lie groups and for understanding the structure of groups acting on trees of hyperbolic spaces.
A different generalization is provided by CAT(0) groups: groups that act properly and cocompactly by isometries on a CAT(0) space, a geodesic metric space in which triangles are no thicker than Euclidean triangles. CAT(0) spaces include Euclidean and hyperbolic spaces, as well as products of trees and many other spaces. The theory of CAT(0) groups, developed by Gromov and by Bruce Kleiner, Michah Sageev, and others, provides a framework for groups with nonpositive curvature that is more flexible than hyperbolicity. CAT(0) groups have solvable word problem, and their geometry can be studied through the structure of their boundaries and through the action on the space at infinity. However, the theory is less complete than that of hyperbolic groups: it is not known whether all CAT(0) groups are biautomatic, and the classification of CAT(0) groups up to quasi-isometry remains far from settled.
A central theme in geometric group theory is the use of boundaries and ends to classify groups. The number of ends of a group—the number of connected components of the complement of a large ball in the Cayley graph—is a quasi-isometry invariant. A classical theorem of Hopf and Freudenthal states that a finitely generated group has 0, 1, 2, or infinitely many ends, and Stallings' theorem characterizes groups with more than one end as those that split over finite subgroups. This result connects geometric group theory to the algebraic theory of group splittings and has been generalized in many directions.
For hyperbolic groups, the boundary is a more refined invariant than the number of ends. The boundary of a hyperbolic group is a compact metrizable space, and its topological type is a quasi-isometry invariant. For example, a hyperbolic group is virtually free if and only if its boundary is a Cantor set, and it is the fundamental group of a closed hyperbolic surface if and only if its boundary is a circle. The classification of hyperbolic groups by their boundaries has been a major project, culminating in results such as the Cannon conjecture—still open—which states that a hyperbolic group with sphere boundary acts properly cocompactly on hyperbolic 3-space.
A powerful tool in geometric group theory is the study of group actions on trees. A group acting on a tree without global fixed point gives rise to a splitting of the group as a graph of groups, a decomposition that encodes how the group is built from simpler pieces. This theory, developed by Jean-Pierre Serre, provides a bridge between geometry and algebra: the tree is a geometric object, but the action encodes algebraic information about the group's structure. Bass–Serre theory, as it is known, has been used to prove theorems about group splittings, to study groups acting on buildings, and to analyze the structure of groups acting on more general complexes.
The theory of actions on trees has been extended to actions on R-trees, where the tree is a metric space in which every point is a branch point. R-tree actions arise naturally as limits of actions on simplicial trees, and they play a central role in the theory of stable actions and in the proof of the accessibility theorem. The Rips machine, developed by Eliyahu Rips, analyzes actions of finitely generated groups on R-trees and has been instrumental in proving results about the structure of groups acting freely on R-trees.
A major research direction is quasi-isometry rigidity: determining when a group's quasi-isometry type determines its algebraic structure. Some groups are rigid in the strongest sense: any group quasi-isometric to them is virtually isomorphic to them. For example, lattices in higher-rank semisimple Lie groups are quasi-isometrically rigid, a result due to Richard Schwartz and to Alex Eskin and Benson Farb. In contrast, many groups are not rigid: for instance, all nonabelian free groups of finite rank are quasi-isometric to each other, as are all surface groups of genus at least two.
The field today is characterized by a rich interplay between several approaches. One approach focuses on the geometry of specific classes of spaces, such as CAT(0) cube complexes, which have become central through the work of Sageev and Ian Agol on the virtual Haken conjecture. Another approach emphasizes the study of group boundaries and their topological and dynamical properties. A third approach, sometimes called measured group theory, studies groups through their measure-preserving actions on probability spaces and has connections to ergodic theory and descriptive set theory. These approaches are not mutually exclusive; many researchers combine them, and results from one area often have implications for another.
The field also maintains deep connections to topology, particularly through the study of 3-manifold groups and the mapping class group. The proof of the virtual Haken conjecture by Agol, building on work of Dani Wise, used geometric group theory techniques—specifically, the theory of special cube complexes—to resolve a long-standing question in 3-manifold topology. This result exemplifies the field's outward reach: geometric group theory provides tools that solve problems in other areas, while those areas in turn generate new questions and examples for the theory.
Several major open problems define the current frontier. The Cannon conjecture, mentioned above, asks whether a hyperbolic group with a 2-sphere boundary must act properly cocompactly on hyperbolic 3-space. The quasi-isometry classification of groups remains incomplete even for relatively well-understood classes: for example, it is not known whether all finitely generated groups quasi-isometric to a given nonuniform lattice in a rank-one Lie group are themselves lattices. The Hopf problem—whether a group with one end can have a nontrivial homomorphism to the integers—was solved affirmatively for many classes but remains open in general. The relationship between the algebraic structure of a group and the possible topologies of its boundary is only partially understood.
A particularly active area is the study of random groups, where one considers groups given by random presentations with a fixed number of generators and relators of increasing length. Gromov's work on random groups showed that, for a wide range of parameters, a random group is hyperbolic, and more recent work has explored the boundary between hyperbolic and non-hyperbolic behavior. This probabilistic perspective provides a way to understand "generic" behavior in the space of all groups, complementing the study of specific examples.
The field's methods have also expanded to include tools from analysis, such as the study of coarse embeddings into Hilbert space and the associated Kazhdan's property (T) and a-T-menability. These properties, which concern the existence or nonexistence of certain unitary representations, have geometric formulations in terms of actions on Hilbert space and have become important in the study of group actions on Banach spaces and in the theory of expander graphs.
Geometric group theory is thus not a single method but a family of interconnected approaches united by a common conviction: that the large-scale geometry of spaces on which groups act is a faithful mirror of group structure. The field's history shows a steady expansion from the study of hyperbolic groups to a broader theory encompassing relative hyperbolicity, CAT(0) geometry, cube complexes, and measured group theory, with each new framework addressing limitations of its predecessors while building on their insights. The result is a mature discipline with a robust toolkit, a rich stock of examples and counterexamples, and a clear sense of the questions that remain open.