A group is a set equipped with an operation that combines any two elements to produce another element of the set, subject to three requirements: associativity, the existence of an identity element, and the existence of an inverse for every element. Groups arise naturally from reversible transformations: two transformations can be performed in succession, doing nothing serves as an identity, and reversing a transformation gives its inverse. They also arise from familiar arithmetic, such as addition of the integers.
Groups are fundamental objects of abstract algebra and the central subject of group theory. They provide a common language for symmetries of geometric figures, permutations of objects, solutions of polynomial equations, continuous transformations, and many structures in physics. Their importance lies partly in abstraction: systems whose elements look entirely different may have the same group structure and therefore satisfy the same theorems.
The theory ranges from small, explicitly calculable groups to infinite groups and families carrying additional geometric or topological structure. Its principal tools include subgroups, homomorphisms, quotient groups, actions, and representations. These tools connect structural questions—such as how a group decomposes—with concrete questions about equations, geometry, computation, and symmetry.
Definition, axioms, and elementary consequences
A group consists of a nonempty set and a binary operation, usually written multiplicatively as . Formally, the operation is a function , so closure is already built into its definition. The defining axioms are:
- Associativity: for all .
- Identity: there is an element such that for every .
- Inverses: for every , there is an element such that .
Associativity allows a product of finitely many elements to be written without specifying parentheses. It does not allow factors to be reordered. A group is called an abelian group when for every pair of elements. For abelian groups, additive notation is common: the operation is written , the identity is , and the inverse of is . An additive notation does not introduce a second operation; it is simply a different notation for the group law.
The axioms imply that the identity and each inverse are unique. If and are both identities, then . If and are both inverses of , then
They also imply cancellation: entails , and entails . Thus the equations and each have unique solutions, namely and .
Inverses reverse the order of multiplication:
This identity expresses a basic feature of reversible processes: undoing two successive operations requires undoing the last operation first. Integer powers are defined by repeated multiplication, with and . For a single element, ; the formula , however, generally requires that and commute.
The number of elements of , its cardinality, is called its order and is denoted . A finite group has finitely many elements; otherwise it is infinite. The one-element group is the trivial group. Groups are distinguished from related algebraic structures by their inverse requirement: a monoid has an associative operation and identity but need not have inverses, while a semigroup requires only associativity.
Arithmetic, permutations, and matrix examples
The integers form a group : addition is associative, is the identity, and the inverse of is . The integers do not form a group under multiplication, because most integers lack multiplicative inverses within . By contrast, the nonzero rational numbers, nonzero real numbers, and nonzero complex numbers are groups under multiplication. Zero must be excluded because it has no multiplicative inverse.
For each positive integer , addition modulo gives the group . Its elements are residue classes, and its law is inherited from modular arithmetic. For example, in , the inverse of the class of is the class of , since . Under multiplication modulo , the invertible residue classes form a different group, denoted . A class is invertible exactly when its representative has greatest common divisor with .
A permutation of a set is a bijection from to itself. All such bijections form the symmetric group , with function composition as multiplication. When has elements, this group is denoted and has elements. In , let exchange and , and let exchange and . With the convention that the rightmost permutation acts first, but . Thus is nonabelian.
Invertible matrices over a field form the general linear group under matrix multiplication. Its identity is the identity matrix, and the inverse of an element is its matrix inverse. Invertibility is equivalent to having nonzero determinant. The subgroup consists of matrices with determinant . These examples connect groups directly to linear algebra.
Geometric symmetry supplies another major class. The distance-preserving transformations of a geometric object form a group because composites and inverses still preserve distances. The symmetries of a regular -gon form a dihedral group of order , including rotations and reflections. Notation varies: this group is called in one convention and in another. Its rotations form an abelian subgroup, but a reflection reverses the direction of a rotation, making the whole group nonabelian when .
Subgroups, generators, and cyclic groups
A subgroup of a group is a subset that is itself a group under the operation inherited from . It is written . A useful test states that a nonempty subset is a subgroup if and only if whenever . This single condition supplies closure, the identity, and inverses. For example, the even integers form a subgroup of the additive integers, but the positive integers do not: they contain neither nor additive inverses.
For a subset , the subgroup generated by , denoted , is the smallest subgroup containing . Equivalently, it consists of finite products of elements of and their inverses. The empty product is the identity. A group is finitely generated if some finite subset generates it. This is distinct from being finite: is infinite but generated by the single element . A finite generating set compresses a potentially infinite group into a finite collection of basic operations.
A cyclic group is generated by one element. If , every element is for some integer , and all elements commute because powers of the same element commute. The order of an element is the least positive integer such that , if such an integer exists; otherwise has infinite order. This order equals the order of the cyclic subgroup .
Every infinite cyclic group is isomorphic to , and every cyclic group of finite order is isomorphic to . A cyclic group of order has exactly one subgroup of order for each positive divisor of . If , that subgroup is . The order of is
Consequently, generates exactly when is relatively prime to .
Subgroups can reveal structure without determining the entire group. The integers have exactly the subgroups for nonnegative integers . In a general group, intersections of subgroups are again subgroups, but unions usually are not. The subgroup generated by a union is therefore needed when combining subgroups. All subgroups of a fixed group form a partially ordered collection under inclusion, called its subgroup lattice; this lattice records much of the group's internal organization, though nonisomorphic groups can have isomorphic subgroup lattices.
Homomorphisms, cosets, and quotient groups
A group homomorphism is a map preserving multiplication:
It automatically preserves identities, inverses, and integer powers. Its image is a subgroup of , and its kernel is
A homomorphism is injective exactly when its kernel is trivial. A bijective homomorphism is an isomorphism; isomorphic groups have the same algebraic structure, even when their elements are represented differently.
For a subgroup , a left coset is a subset ; right cosets are defined by . Left cosets partition , and each has the same cardinality as . Their number is the index . For finite groups this gives Lagrange's theorem:
Thus a subgroup's order divides the group's order. In particular, the order of every element divides . The converse is false in general: a divisor of need not occur as a subgroup order.
A subgroup is normal if for every , equivalently if its left and right cosets agree. Normality is exactly the condition needed to define multiplication on cosets by
independently of representatives. The resulting quotient group treats all elements of as indistinguishable from the identity. Every kernel is normal, and every normal subgroup is the kernel of the natural projection .
The first isomorphism theorem states that
For example, the determinant map is surjective and has kernel , so their quotient is isomorphic to . Reduction modulo similarly identifies with the quotient of by . These are instances of one mechanism: a homomorphism forgets precisely the distinctions encoded by its kernel.
Other isomorphism theorems describe how subgroups and quotients interact. If and , then . If are normal in , then . Subgroups of correspond to subgroups of containing , with normality preserved. These correspondences allow many questions about a quotient to be translated into questions inside the original group.
Group actions and the mathematics of symmetry
A group action makes the idea of a group acting as transformations precise. A left action of on a set assigns to each and an element , satisfying
Equivalently, an action is a homomorphism . The action is faithful when this homomorphism is injective, so distinct group elements induce distinct transformations.
The orbit of is , and its stabilizer is the subgroup . Orbits partition . The orbit–stabilizer theorem identifies the orbit with the coset space ; in the finite case,
For example, the rotation group of a cube has elements. It acts transitively on the cube's six faces, and the stabilizer of a face has four elements, corresponding to rotations through multiples of about the axis through that face and its opposite.
Every group acts on itself by left multiplication. This action is faithful and proves Cayley's theorem: every group is isomorphic to a subgroup of a symmetric group. If is finite of order , it embeds in . This does not make permutation groups merely one example among others; it shows that arbitrary groups can always be realized concretely as permutations, although the resulting realization may be much larger than necessary.
Another central action is conjugation, . Its orbits are conjugacy classes. The stabilizer of is its centralizer , the elements commuting with . Elements fixed by all conjugations form the center . For a finite group, partitioning into conjugacy classes yields the class equation
where the sum runs over representatives of noncentral conjugacy classes. This equation translates information about commutation into divisibility statements.
Actions also solve counting problems. Burnside's lemma states that the number of orbits of a finite group acting on a finite set is
It counts configurations up to symmetry, such as colorings of a polygon considered equivalent under rotation. The stabilizers matter because highly symmetric configurations have smaller orbits; simply dividing the total number of configurations by the group order usually gives an incorrect answer.
Finite groups and Sylow theory
A finite group's order places strong restrictions on its structure, but usually does not determine it. A group of prime order is cyclic, because any nonidentity element generates a subgroup whose order divides . At order , however, there are already two possibilities: the cyclic group and the Klein four-group . Both are abelian, but one contains an element of order , while the other does not.
Cauchy's theorem supplies a partial converse to Lagrange's theorem: if a prime divides , then contains an element of order . The more extensive Sylow theorems, proved by Ludwig Sylow in 1872, concern the largest prime-power subgroups. If
then has subgroups of order , called Sylow -subgroups. Every subgroup of -power order lies in a Sylow -subgroup, and all Sylow -subgroups are conjugate.
If denotes the number of Sylow -subgroups, then
These arithmetic restrictions often force , in which case the unique Sylow -subgroup is normal. For a group of order , the number of Sylow -subgroups divides and is modulo , so it is . The number of Sylow -subgroups divides and is modulo , so it too is . The two normal subgroups commute, giving .
Groups of order , called finite -groups, have nontrivial centers. The class equation proves this because every noncentral conjugacy class has size divisible by , forcing to be divisible by . This observation leads to inductive arguments through successive central quotients. Every group of order is abelian, whereas nonabelian groups already appear at order .
Finite group theory combines such local information with global constraints. Sylow subgroups reveal the structure associated with each prime divisor, but the ways these subgroups interact can be highly intricate. Groups with identical orders and similar Sylow subgroups may still be nonisomorphic. Consequently, order calculations are most effective when combined with actions, normal subgroups, conjugacy classes, and representations rather than treated as a complete classification method.
Abelian groups and their classification
Abelian groups admit particularly strong structural theorems because commutativity makes it possible to manipulate generators by methods resembling linear algebra. Every vector space is an abelian group under addition, but an abelian group need not support scalar multiplication by a field. More generally, abelian groups are exactly modules over the ring : multiplication by an integer means repeated addition and its inverse. This places their classification within the theory of modules.
The fundamental theorem of finitely generated abelian groups states that every such group is isomorphic to
where , each , and . The integer and the invariant factors are uniquely determined. The component is free abelian, while the finite summands form the torsion subgroup, consisting of elements of finite order.
An alternative decomposition splits the finite part into cyclic groups of prime-power order. For example,
This follows from the Chinese remainder theorem, since and are coprime. In contrast, is not isomorphic to . For a finite abelian group, the possible decompositions are controlled by partitions of the exponents in its prime factorization. At order , the abelian possibilities are , , and .
The classification can be computed from generators and relations using Smith normal form. A matrix of integer relations is transformed by invertible integer row and column operations into a diagonal matrix whose nonzero diagonal entries divide the following entries. Those entries determine the finite cyclic factors, while missing pivots determine the free rank. Unlike ordinary row reduction over a field, the permitted transformations must preserve the integer lattice, not merely the rational span.
The finitely generated hypothesis is essential. The additive group is torsion-free but is not a free abelian group, and it is not finitely generated. The group is torsion but infinite. The structure theory of arbitrary infinite abelian groups is therefore substantially more complicated than the finite or finitely generated theory.
Abelian groups also arise naturally from nonabelian groups. The subgroup generated by all commutators is normal, and quotienting by it gives the largest abelian quotient, called the abelianization. This construction extracts the part of a group visible to every homomorphism into an abelian group, making abelian invariants useful even when the original group is highly noncommutative.
Products, free groups, and presentations
New groups can be constructed from existing ones in several distinct ways. The direct product has ordered pairs as elements and componentwise multiplication:
Each factor appears as a normal subgroup, and elements from the two factors commute. For finitely many abelian groups, direct product and direct sum describe the same group. For an infinite family, the direct sum permits only tuples with finitely many nonidentity coordinates, whereas the direct product allows arbitrary tuples.
A semidirect product allows one factor to act nontrivially on another. Given a homomorphism , the group has multiplication
The subgroup is normal, while need not be. The symmetries of a regular -gon give , where the nonidentity element of sends each rotation to its inverse. This construction records not just two component groups but also the interaction between them.
A free group on a set has no relations beyond those forced by the group axioms. Its elements can be represented by reduced words in symbols and formal inverses , with adjacent inverse pairs cancelled. Its defining universal property says that every function from to a group extends uniquely to a homomorphism from the free group. The free group on one generator is ; on two or more generators it is nonabelian.
A presentation
specifies generators and relators , each relator being a word required to equal the identity. Formally, is the free group on divided by the smallest normal subgroup containing the relators. Examples include
and
A finite presentation is a finite description, not a guarantee that the group is finite.
Presentations are especially effective when a group is obtained from geometric operations or paths. They also reveal a computational limitation: the same group may have very different presentations, and a short presentation can conceal complicated structure. The free product , another basic construction, combines two groups without adding commutation relations between them. Thus direct products, semidirect products, and free products encode three genuinely different ways of assembling groups.
Commutators, solvability, and extensions
The failure of two elements to commute can be measured by a commutator. Using one common convention,
The equation holds exactly when . Conventions differ by inverses or reordered variables, but the subgroup generated by all commutators is the same. This commutator subgroup, denoted or , is normal and gives the abelianization .
Iterating the construction yields the derived series:
A solvable group is one for which this series eventually reaches the trivial subgroup. Every abelian group is solvable, and is solvable because its commutator subgroup is , whose own commutator subgroup is trivial. By contrast, is nonabelian and simple, so its commutator subgroup is all of ; it is not solvable.
A stronger condition is nilpotence. The lower central series is defined by
The group is nilpotent if this series reaches the identity after finitely many steps. Every nilpotent group is solvable, but the converse fails: is solvable and not nilpotent. Every finite -group is nilpotent, and a finite group is nilpotent exactly when it is the direct product of its Sylow subgroups.
A group extension expresses a group through a normal subgroup and quotient:
Exactness means that is identified with the kernel of the surjection onto . An extension is split when the quotient map admits a homomorphic section ; in that case is a semidirect product . Not every extension splits. Both and have a normal subgroup and quotient , but only the latter realizes the corresponding extension as a direct product.
Extensions explain why knowledge of constituent groups does not uniquely reconstruct the whole. The action of the quotient on the normal subgroup and additional compatibility data matter. Group cohomology provides a systematic language for such data; with an abelian kernel and a specified quotient action, second cohomology classifies equivalence classes of extensions. This perspective links elementary structural questions to topology and homological algebra.
Simple groups and the finite classification
A simple group is a nontrivial group whose only normal subgroups are the identity subgroup and the whole group. The finite abelian simple groups are precisely the cyclic groups of prime order. Nonabelian simple groups have no nontrivial abelian quotients and cannot be decomposed through a proper nontrivial normal subgroup. They therefore serve as basic constituents of finite groups, though reconstructing a group from those constituents still requires extension data.
Every finite group admits a composition series
whose successive quotients are simple. Each subgroup need only be normal in the next subgroup, not necessarily in all of . The Jordan–Hölder theorem states that the resulting simple factors are independent of the chosen series, up to isomorphism and order. Different groups can have the same composition factors, so these factors are invariants rather than a complete description.
The classification of finite simple groups states that every finite simple group belongs to one of four classes:
- cyclic groups of prime order;
- alternating groups for ;
- finite simple groups of Lie type;
- sporadic simple groups.
The alternating group consists of the even permutations in . The smallest nonabelian simple group is , of order , which is also the rotation group of a regular icosahedron. Groups of Lie type are constructed using algebraic and geometric structures over finite fields; small parameters require exclusions or produce familiar coincidences between families.
The sporadic groups do not belong to the infinite families. The largest, the Monster group, has order approximately . Their discovery and analysis involved permutation representations, lattices, and sophisticated local subgroup methods. The classification proof developed across thousands of pages and many decades. A missing case in the treatment of quasithin groups was completed by Michael Aschbacher and Stephen D. Smith in 2004.
A major ingredient is the Feit–Thompson theorem of 1963: every finite group of odd order is solvable. It follows that every nonabelian finite simple group has even order. Such results illustrate how an elementary numerical condition can impose a deep structural restriction. The classification has consequences far beyond listing simple groups, including results on permutation groups, maximal subgroups, finite geometries, and algorithms. It does not, however, classify all finite groups individually; extension problems and the proliferation of prime-power groups remain substantial.
Representations and characters
Representation theory studies groups by realizing their elements as linear transformations. A representation of on a vector space over a field is a homomorphism
Choosing a basis turns each into a matrix. The representation need not be faithful: its kernel consists of group elements acting as the identity on all vectors. Its dimension is the dimension of .
A subspace is invariant if for every . A nonzero representation is irreducible when its only invariant subspaces are and . For a finite group over a field whose characteristic does not divide , Maschke's theorem states that every finite-dimensional representation is a direct sum of irreducible representations. The proof averages a projection over the group to produce an invariant complementary subspace. If the field characteristic divides , this averaging is unavailable, and complete reducibility can fail.
For a complex representation, its character is the function
where the trace is the sum of diagonal entries. Characters are constant on conjugacy classes because similar matrices have the same trace. For a finite group, irreducible complex characters form an orthonormal basis for the space of complex-valued functions constant on conjugacy classes, using
Consequently, the number of irreducible complex representations, up to equivalence, equals the number of conjugacy classes.
If their dimensions are , then
For , the irreducible dimensions are , and . The two one-dimensional representations are the trivial representation and the sign representation. The two-dimensional representation can be obtained from the permutation action on by restricting to the plane whose coordinates sum to zero.
For a finite abelian group, all irreducible complex representations are one-dimensional. Its characters therefore generalize the complex exponentials appearing in the discrete Fourier transform. Nonabelian Fourier analysis replaces scalar frequency coefficients with matrix-valued data. Representations also organize symmetry in differential equations and quantum mechanics, where commuting with a symmetry action constrains an operator's invariant subspaces and possible degeneracies. Character tables encode considerable structural information, although identical character tables need not imply that two groups are isomorphic.
Topological groups, Lie groups, and continuous symmetry
A topological group combines a group with a topology such that multiplication and inversion are continuous. This compatibility allows algebraic and limiting processes to interact. Examples include the additive real numbers, the nonzero complex numbers under multiplication, and invertible real matrices with their usual matrix-entry topology. Every abstract group can also be given the discrete topology, though this may discard useful geometric information.
A Lie group is a group that is also a finite-dimensional smooth manifold, with smooth multiplication and inversion. Important examples are , the orthogonal group , the special orthogonal group , and the unitary group . The circle group
describes planar rotations through multiplication of unit complex numbers. The group describes rotations of three-dimensional Euclidean space about the origin.
The tangent space at a Lie group's identity carries an additional operation, the Lie bracket, forming its Lie algebra. For matrix Lie groups, the bracket is
The matrix exponential links infinitesimal data to group elements:
For , the Lie algebra consists of real skew-symmetric matrices. Exponentiating such matrices produces orthogonal matrices of determinant , connecting infinitesimal rotations with finite rotations.
The Lie algebra determines much local structure but not all global structure. The groups and have isomorphic Lie algebras, yet is a double cover of . Their distinction is topological: is simply connected, whereas is not. Thus a description of infinitesimal symmetry may leave global choices unresolved.
Locally compact groups form another major setting. They admit Haar measure, a translation-invariant measure unique up to positive scaling. It allows integration and harmonic analysis on groups, generalizing integration on the real line and circle. For compact groups, Haar measure can be normalized to total mass . Profinite groups provide a contrasting type of topological group: they are inverse limits of finite groups, and are compact and totally disconnected. They arise naturally when infinitely many compatible finite symmetries must be treated together, especially in arithmetic.
Groups in geometry and topology
Groups connect geometric spaces with algebra by recording transformations, paths, and identifications. In 1872, Felix Klein proposed studying geometries through their transformation groups and the properties those groups preserve. Euclidean geometry studies properties invariant under Euclidean motions; affine geometry uses affine transformations; projective geometry uses projective transformations. Different transformation groups preserve different data: distance, parallelism, or incidence, respectively.
In topology, the fundamental group records loops based at a point , considered equivalent when one can be continuously deformed into another while keeping the base point fixed. Multiplication is concatenation of loops, and inversion reverses direction. Concatenation is not literally associative as a parametrized path operation, but it becomes associative on homotopy classes. The circle has fundamental group , while the torus has fundamental group .
The fundamental group detects features that elementary measurements may miss. A sphere of dimension at least is simply connected, so its fundamental group is trivial, whereas a torus has loops that cannot be contracted to a point. For a path-connected space, changing the base point produces isomorphic fundamental groups, although the isomorphism depends on the chosen connecting path. The first homology group is the abelianization of the fundamental group, showing precisely how one familiar topological invariant loses noncommutative information.
Groups also act as deck transformations of covering spaces. Under standard local hypotheses, connected covering spaces of a space correspond to subgroups of its fundamental group. A torus, for example, is obtained from the plane by identifying points differing by an integer translation, so acts on and the quotient is a torus. Similar constructions connect discrete groups with non-Euclidean geometry and manifolds.
Geometric group theory reverses the direction by treating a finitely generated group as a geometric object. A Cayley graph has group elements as vertices and edges corresponding to multiplication by generators. The resulting word metric measures the shortest expression for . Different finite generating sets change exact distances but preserve the large-scale geometry up to quasi-isometry. Growth, curvature-like properties, and boundary constructions then become tools for studying abstract groups.
For example, has polynomial growth of degree , whereas a free group on at least two generators has exponential growth. Gromov's theorem characterizes finitely generated groups of polynomial growth as virtually nilpotent, meaning they contain a nilpotent subgroup of finite index. This result directly connects a geometric counting property with an algebraic structural condition.
Polynomial equations, arithmetic, and physical applications
In Galois theory, groups encode symmetries of field extensions. For a field extension , the automorphisms of fixing every element of form . For a finite Galois extension, subgroups of this group correspond, in reverse order, to intermediate fields. A larger subgroup fixes fewer elements; a smaller subgroup fixes more. Normal subgroups correspond to intermediate fields that are themselves Galois over the base field.
If is the splitting field of a separable polynomial, each field automorphism permutes its roots while preserving their algebraic relations. Its Galois group therefore embeds in a symmetric group on the roots. Over a field of characteristic zero, a polynomial is solvable by radicals exactly when the Galois group of its splitting field is solvable. This explains why no formula by radicals solves the general quintic: the general degree-five equation has Galois group , which is not solvable. Particular quintics can nevertheless have solvable Galois groups.
Groups also organize number theory through residue classes, ideal classes, and points on algebraic curves. The multiplicative group of a finite field is cyclic. On a nonsingular elliptic curve with a chosen identity point, a geometric addition law turns its points into an abelian group. Over a finite field this group is finite, and computing repeated addition is efficient. These structures support cryptographic systems based on the difficulty of recovering a scalar from a known point and its multiple. The difficulty depends on the group and parameters, not merely on the existence of a group law.
In crystallography, the symmetries of a periodic arrangement form a space group containing a lattice of translations. In three-dimensional Euclidean space there are crystallographic space-group types under the conventional classification. The crystallographic restriction limits ordinary lattice-preserving rotational orders to and . Continuous symmetry groups serve a different role in physics, where rotations, translations, and spacetime transformations constrain equations and classify states.
For suitable variational systems, Emmy Noether's theorem relates differentiable continuous symmetries of an action to conservation laws. Time-translation symmetry corresponds to energy conservation under the theorem's hypotheses, and rotational symmetry to angular momentum conservation. In quantum theory, representations of symmetry groups determine how state spaces transform; the distinction between and its double cover is central to describing spin. These uses depend on both the abstract group and the particular representation or geometric action involved.
Historical development
The origins of group theory lie in several initially separate subjects: solving equations, studying permutations, and analyzing geometric transformations. Eighteenth-century work on polynomial equations increasingly used permutations of roots to investigate why solution formulas exist. Joseph-Louis Lagrange's memoir of 1770–1771 examined expressions in roots and how their values change under permutations. His methods did not yet constitute modern abstract group theory, but they helped identify permutation structure as a central issue.
Paolo Ruffini and Niels Henrik Abel established limitations on solving general polynomial equations by radicals; Abel's 1824 work gave the decisive impossibility result for the general quintic. Évariste Galois then related solvability to a group of permutations preserving algebraic relations among roots. His work of 1830–1832 introduced ideas that became the foundation of Galois theory. Joseph Liouville published Galois's mathematical manuscripts in 1846, after Galois's death in 1832.
Permutation groups developed through the work of Augustin-Louis Cauchy, Camille Jordan, and others. Arthur Cayley's 1854 paper “On the Theory of Groups, as Depending on the Symbolic Equation ” helped separate the group concept from any single type of object. Cayley treated multiplication abstractly and established the representation of groups by permutations. The transition was gradual: nineteenth-century definitions and terminology varied before the modern axiomatic formulation became standard.
Other developments greatly broadened the subject. Sylow's 1872 theorems gave powerful control over finite groups through prime-power subgroups. Klein's Erlangen program in the same year placed transformation groups at the center of geometry. Sophus Lie developed the theory of continuous transformation groups, linking symmetry with differential equations. Henri Poincaré's work in the 1890s helped establish the fundamental group as a tool of topology.
At the turn of the twentieth century, Ferdinand Georg Frobenius, William Burnside, and Issai Schur developed character and representation theory. Otto Hölder and other algebraists clarified structural ideas involving simple groups and composition series. Later, Noether's approach to abstract algebra emphasized homomorphisms, invariant structures, and connections among groups, rings, and modules, helping consolidate the structural language used today.
Twentieth-century group theory expanded in several directions rather than following one linear path. Finite-group classification, Lie theory, algebraic groups, topology, and combinatorial group theory each developed specialized methods. Their interactions remained important: geometry supplied examples of finite simple groups, representations supplied tools for their analysis, and computational methods made large concrete groups accessible. By the late twentieth century, geometric methods had also transformed the study of infinite finitely generated groups.
Computation, decidability, and continuing structural questions
Computational group theory studies both algorithms for concrete groups and the limits of algorithmic methods. A group may be supplied by a multiplication table, permutation generators, matrices, or an abstract presentation. The input model matters greatly. A multiplication table makes every product immediately accessible but requires storage quadratic in the group order. Generators can describe a very large group compactly, but extracting basic information from them may require substantial computation.
For finite permutation groups, stabilizer chains and the Schreier–Sims method support algorithms for computing group order, testing membership, and finding stabilizers. These methods repeatedly reduce a problem to the stabilizer of a point, using orbit–stabilizer relations. For finitely presented groups, Todd–Coxeter coset enumeration attempts to enumerate cosets of a specified subgroup. It can establish an index when successful, but a run need not terminate in general. Systems such as GAP and Magma implement extensive collections of group-theoretic algorithms and databases.
The word problem asks whether a given word in generators represents the identity. For a finite group given by explicit permutations or matrices, direct evaluation decides this question. For arbitrary finitely presented groups, however, there is no universal algorithm solving it in every case. Results of Pyotr Novikov and William Boone in the 1950s established the existence of finitely presented groups with undecidable word problem. This is a limitation on possible algorithms, not merely a statement that some computations are slow.
Other problems ask whether two words represent conjugate elements or whether two presentations define isomorphic groups. These too are undecidable in full generality, though many important classes admit effective solutions. Free groups have straightforward word reduction, and hyperbolic groups have strong algorithmic properties. A useful group-theoretic theorem can therefore have an algorithmic consequence, while an undecidability result identifies where no completely general computational procedure can exist.
Infinite groups exhibit behavior that cannot be inferred from familiar finite examples. The Burnside problems ask, in several formulations, whether finitely generated groups whose elements have finite order must be finite. Infinite finitely generated torsion groups exist, and Novikov and Sergei Adian constructed infinite groups with a fixed finite exponent for suitable exponents. In contrast, Efim Zelmanov's solution of the restricted Burnside problem shows that for fixed numbers of generators and a fixed exponent, the orders of finite groups satisfying those bounds are bounded.
Another important property is residual finiteness: every nonidentity element remains nonidentity in some finite quotient. This lets finite images detect individual elements, but not every finitely generated or finitely presented group has the property. Research on growth, rigidity, subgroup structure, finite approximations, and effective classification continues to distinguish classes where strong structural or algorithmic theorems are possible from classes whose behavior is fundamentally less constrained.