An algebraic variety is a geometric object defined, locally or globally, by polynomial equations over a field. Lines, conics, many familiar surfaces, and higher-dimensional spaces of solutions all provide examples. Varieties are the central objects of algebraic geometry, which translates geometric questions into algebraic ones and uses geometric structures to study equations. Their importance extends to number theory, complex geometry, mathematical physics, and computational mathematics.
The precise definition depends on context. In classical algebraic geometry, a variety is usually an irreducible algebraic set over an algebraically closed field, equipped with its regular functions. In modern treatments, it is commonly an integral, separated scheme of finite type over a field, although some authors allow reducible varieties by requiring only that they be reduced. These conventions distinguish varieties from arbitrary solution sets and from more general schemes, which may retain multiplicities and infinitesimal information invisible in their points.
The subject connects several complementary descriptions: equations in affine or projective coordinates, rings of functions, local structures near individual points, and global invariants such as dimension, degree, and genus. Changing the underlying field can change the geometry substantially; a variety may have abundant complex points but no real or rational points.
Affine varieties and polynomial equations
Let be a field. The -dimensional affine space over , denoted , has -valued points identified with tuples . Given polynomials , their common zero set is
Over an algebraically closed field, such a set is an affine algebraic set. Under the convention requiring irreducibility, it is an affine variety only if it cannot be expressed as the union of two proper algebraic subsets. Algebraic closure means that every nonconstant polynomial in one variable has a root in the field; the complex numbers are the principal characteristic-zero example.
Linear equations define affine linear subspaces, while a single nonconstant equation defines a hypersurface. The equation defines a parabola, and defines a hyperbola. The set , by contrast, is the union of the two coordinate axes and is reducible. The hypersurface in three-dimensional affine space is irreducible over the complex numbers, although it has a singular point at the origin. Irreducibility and smoothness are therefore independent properties.
The equations presenting an affine variety are not unique, and the number of coordinates is not an intrinsic invariant. The parabola , for example, is algebraically equivalent to the affine line: the maps and are mutually inverse polynomial maps. Its curved appearance in a particular embedding does not distinguish it intrinsically from a line. Algebraic geometry consequently studies both embedded varieties, where an ambient space is specified, and abstract varieties, where only the intrinsic structure matters.
The base field is essential even at this elementary stage. The equation has no solutions over the real numbers but defines a nonempty complex curve. Modern terminology therefore distinguishes a variety defined over from its set of -rational points. The variety includes algebraic information that cannot, over an arbitrary field, be recovered from that set alone. Likewise, equations such as and have the same ordinary zero set, but the first can define a different scheme because it records a nonzero infinitesimal direction.
Coordinate rings and the Zariski topology
For a subset , its vanishing ideal is
This is an ideal in the polynomial ring. The coordinate ring of an affine algebraic set is
Its elements are polynomial functions on , with two polynomials identified when their restrictions agree. Thus the coordinate ring records functions rather than a preferred list of equations. It is the basic algebraic counterpart of the geometry.
Hilbert’s Nullstellensatz makes this correspondence precise over an algebraically closed field:
where is the radical of . Consequently, radical ideals correspond to algebraic sets, and prime ideals correspond to irreducible algebraic sets. An affine algebraic set is irreducible exactly when its coordinate ring is an integral domain. For example, has nonzero elements whose product is zero, reflecting the decomposition into two axes. The ring , in contrast, is isomorphic to .
The algebraic subsets define the closed sets of the Zariski topology. This topology is much coarser than the usual topology on real or complex points. On the affine line over an algebraically closed field, every proper closed subset is finite. On any irreducible variety, two nonempty open subsets must intersect, so a positive-dimensional irreducible variety is not a Hausdorff space in its Zariski topology. Despite this unfamiliar behavior, the topology is adapted to algebraic functions: the complement of a polynomial zero set is open, and algebraic maps are continuous.
Hilbert’s basis theorem implies that polynomial rings over fields are Noetherian: every ideal has finitely many generators. Geometrically, descending chains of algebraic closed subsets stabilize, and every algebraic set has finitely many irreducible components. These finiteness properties make global decomposition manageable even when infinitely many equations initially appear.
The dictionary also reverses the direction of maps. A morphism between affine varieties pulls functions back to a -algebra homomorphism . Conversely, every such homomorphism determines a morphism. This contravariant equivalence allows questions about geometric embeddings, decompositions, and maps to be reformulated as questions about finitely generated reduced algebras.
Projective varieties and global geometry
Affine coordinates omit points at infinity and do not provide a natural setting for many intersection problems. Projective space addresses this by identifying nonzero vectors that differ by a nonzero scalar:
at the level of -valued points. A point is written in homogeneous coordinates. The equation of a projective algebraic set must be homogeneous, so its vanishing is unaffected by rescaling the coordinates. The open subset where is an affine -space, giving a cover by affine charts.
An affine equation can be homogenized to construct a projective closure. For the parabola , homogeneous coordinates give the projective equation . The chart recovers the affine parabola, while contributes the point . This additional point changes the global geometry: the affine parabola is an affine line, whereas the projective conic is isomorphic to the projective line over an algebraically closed field. Projective closure also explains why parallel affine lines meet at infinity.
A projective variety is a closed subvariety of some projective space. A quasi-projective variety is a locally closed subvariety of projective space, meaning an open subset of a closed one. This class includes all affine varieties and most classical examples. In sufficiently high dimension, however, abstract varieties need not be quasi-projective; embedding into projective space is an additional geometric property, not part of every modern definition.
Projectivity supplies a form of algebraic compactness. A projective variety is proper over its base field, which implies, among other properties, that projection from its product with another variety sends closed subsets to closed subsets. Over the complex numbers, a projective variety is compact in the usual analytic topology. These properties allow families of intersections and limiting configurations to be studied without losing points at infinity.
Degree is a fundamental invariant of a projective embedding. An -dimensional projective variety has degree if a general linear subspace of codimension meets it in points, counted with appropriate multiplicities. Bézout’s theorem states that plane projective curves of degrees and , with no common component, have total intersection multiplicity over an algebraically closed field. A conic and a cubic therefore have total intersection number six, even when some intersections coincide or lie outside an affine chart.
Regular maps, rational maps, and birational equivalence
Maps between varieties must respect their algebraic structure. A function on an open subset is regular if, near every point, it can be written as a quotient of polynomials whose denominator does not vanish there. On an affine variety, every globally regular function is represented by an element of its coordinate ring. A morphism is a map locally described by regular functions. For affine varieties, morphisms can therefore be specified by polynomial coordinate functions, subject to the equations of the target.
An isomorphism is a morphism possessing a morphism as its inverse. It preserves all intrinsic algebraic structure, not merely the points or their topology. A useful contrast is the parametrization
Over an algebraically closed field of characteristic zero, this is bijective on points, but it is not an isomorphism. Away from the origin, its inverse is ; that quotient does not extend to a regular function on the cusp. A bijective algebraic map can therefore fail to preserve the local structure.
A rational function on an irreducible affine variety is an element of the fraction field , called its function field. More generally, a rational map is a morphism defined on a nonempty open subset of , with two descriptions identified if they agree on a common dense open subset. Projective coordinates may make a rational map meaningful where an individual denominator vanishes, but they can also leave an indeterminacy locus where all defining coordinates vanish simultaneously.
Two irreducible varieties are birationally equivalent when they contain isomorphic nonempty open subsets. Equivalently, their function fields are isomorphic as extensions of the base field. The affine line, the projective line, and a nonsingular conic with a rational point are birational, although they need not be isomorphic. A variety of dimension is rational if it is birational to ; its function field is then a purely transcendental extension .
Birational equivalence deliberately ignores lower-dimensional subsets while retaining generic algebraic behavior. It is therefore coarser than isomorphism but still restrictive: not every curve admits a rational parametrization, and higher-dimensional rationality is often difficult to decide. Even for familiar varieties given by low-degree equations, distinguishing rationality from weaker properties, such as being dominated by projective space through a rational map, can require sophisticated invariants.
Dimension, tangent spaces, and singularities
The dimension of an irreducible variety can be defined as the transcendence degree of its function field over the base field:
It also equals the Krull dimension of an affine coordinate ring, measured by chains of prime ideals. Affine -space has dimension , an irreducible plane curve has dimension one, and a nonempty hypersurface in affine -space has dimension . For a reducible algebraic set, dimension is the maximum dimension of its irreducible components.
Dimension is not generally the number of coordinates minus the number of equations in a chosen presentation. Equations may be redundant or impose dependent conditions. The three equations in affine three-space define the union of the three coordinate axes, a one-dimensional set rather than a zero-dimensional one. Noether normalization gives a structural explanation of dimension: every irreducible affine variety of dimension admits a finite surjective morphism to . Its coordinate ring is integral and finite over a polynomial subring in algebraically independent elements.
Local geometry is measured by the Zariski tangent space. If has ideal generated by , the tangent space at a rational point consists of vectors satisfying
It is therefore the kernel of the Jacobian matrix evaluated at . Intrinsically, it is the dual of , where is the maximal ideal in the local ring. At a nonsingular point over an algebraically closed field, its dimension equals the local dimension of the variety; at a singular point, it is larger.
For the cusp , both first partial derivatives vanish at the origin, giving a two-dimensional tangent space to a one-dimensional curve. The nodal curve also has an oversized tangent space at the origin, but its local branches differ from those of the cusp. First-order tangent information detects singularity without completely describing it. Tangent cones, local rings, and normalization distinguish finer features.
Over a perfect field, regularity and smoothness agree for varieties, but they can diverge over imperfect fields. This distinction matters in positive characteristic. In characteristic zero, resolution of singularities replaces a singular variety by a smooth one through a proper birational morphism. Heisuke Hironaka proved the general characteristic-zero theorem in 1964; resolution in arbitrary dimension and positive characteristic remains an important unresolved problem.
Algebraic curves and their invariants
Algebraic curves are varieties of dimension one. They are the first setting in which algebraic, analytic, and arithmetic classifications interact in substantial depth. A nonsingular projective curve over the complex numbers determines a compact complex manifold of complex dimension one, called a compact Riemann surface. Conversely, every compact Riemann surface arises from a nonsingular projective algebraic curve. Thus complex algebraic curves can be studied through complex analysis and topology as well as equations.
The central invariant is the genus. For a smooth projective curve , it can be defined as
the dimension of its space of regular differential one-forms. Over the complex numbers, it equals the number of handles of the associated compact oriented surface, whose Euler characteristic is . A smooth plane curve of degree has genus
Consequently, smooth conics have genus zero, smooth plane cubics genus one, and smooth plane quartics genus three. For a singular plane curve, this degree formula gives its arithmetic genus; the genus of its smooth normalization can be smaller.
A genus-zero smooth projective curve over an algebraically closed field is isomorphic to the projective line. Over a general field, a rational point is additionally required. A smooth projective genus-one curve equipped with a specified rational point is an elliptic curve. In characteristic different from two and three, it can be expressed in Weierstrass form
The chosen point at infinity serves as the identity for an algebraic group law. A genus-one curve without a chosen rational point need not itself be an elliptic curve over its field of definition.
Divisors encode zeros and poles as finite formal sums of points with integer coefficients. The Riemann–Roch theorem relates the number of rational functions with prescribed poles to the degree of a divisor and the genus:
Here is a canonical divisor and is the dimension of the permitted function space. The theorem produces maps to projective space and controls many classification questions.
Curves also illustrate the distinction between geometric and arithmetic complexity. A smooth projective curve of genus at least two over a number field has only finitely many rational points, by the theorem proved by Gerd Faltings in 1983. Genus-one curves have a different arithmetic structure, while genus-zero curves with a rational point admit rational parametrizations. These statements concern rational points, not the abundance of points after extending to an algebraically closed field.
Higher-dimensional varieties and birational classification
For surfaces and higher-dimensional varieties, dimension and degree do not determine the geometry. Products, hypersurfaces, Grassmannians, and algebraic groups provide families with widely different structures. A Grassmannian parametrizes linear subspaces of a fixed vector space and embeds into projective space through Plücker coordinates satisfying quadratic equations. It is a variety whose points are themselves geometric objects, illustrating the principle that classification problems often lead to new varieties.
Birational geometry studies higher-dimensional objects by modifying lower-dimensional subsets. A basic operation is the blowup. Blowing up a smooth point of a surface replaces that point by a projective line parametrizing tangent directions through it. Away from the point, the operation is an isomorphism. Although a blowup does not change the function field, it changes divisors and intersection numbers. It can separate branches, resolve indeterminacies of maps, and alter the form of singularities.
For smooth projective surfaces over an algebraically closed field, a curve isomorphic to with self-intersection can be contracted to a smooth point. Repeated contractions lead to minimal models. Minimal models of smooth projective surfaces are unique within a birational class when the Kodaira dimension is nonnegative, but rational and ruled surfaces require separate treatment. This exception shows why a single canonical representative is not available in every birational class.
The canonical bundle, formed from top-degree differential forms, organizes much of the classification. The Kodaira dimension measures the growth of spaces of sections of its tensor powers and takes values
Varieties of general type have maximal Kodaira dimension. Fano varieties instead have ample anticanonical bundle, a positivity condition associated with very different geometry. Calabi–Yau varieties, under commonly used definitions involving a trivial canonical bundle and additional hypotheses, occupy another important part of the classification. These classes cannot be distinguished merely by the complexity of their defining equations.
In dimension three and above, birational modifications require contractions and flips, and the natural models may have controlled singularities. The minimal model program seeks to organize these operations into a systematic classification. The existence of minimal models for varieties of general type in characteristic zero was established in work published in 2010 by Caucher Birkar, Paolo Cascini, Christopher Hacon, and James McKernan. Broader existence and termination questions remain subjects of research, particularly beyond characteristic zero.
Schemes, sheaves, and algebraic families
A point set with its Zariski topology does not retain enough information for all geometric questions. The modern framework attaches to each open set a ring of regular functions, organized as a sheaf. This structure sheaf records how local functions restrict and glue. For an affine variety, the topology together with the structure sheaf recovers the coordinate ring and its localizations. Abstract varieties are built by gluing affine pieces along compatible open subsets, much as manifolds are assembled from coordinate charts.
A scheme generalizes this construction by starting from arbitrary commutative rings. The affine scheme has prime ideals of as its points and a structure sheaf determined by localizations of . Maximal ideals recover classical closed points over an algebraically closed field, while nonmaximal prime ideals supply generic points of irreducible closed subsets. The prime ideal in , for example, is the generic point of the affine line; its closure is the entire line.
Schemes can distinguish objects with identical ordinary point sets. The rings and each have a single prime ideal, but the second contains a nonzero element whose square is zero. Its spectrum is a doubled point, retaining infinitesimal information. Such nonreduced structures are useful rather than defective: intersections and degenerations naturally produce multiplicities that reduced varieties alone would discard.
Families are expressed as morphisms , with the fiber over representing an object varying with the parameter . The equation defines a family of plane curves over the affine -line. For nonzero , the fiber is a smooth hyperbola; at , it becomes the union of the coordinate axes. Flatness is the algebraic condition that controls continuity of this variation in an appropriate sense. In projective flat families, the Hilbert polynomial of fibers is locally constant, preserving numerical information even when smooth fibers become singular.
Sheaf cohomology measures obstructions to passing from local data to global data. A line bundle may have local sections that cannot be assembled into a global section, or transition functions that cannot be globally trivialized. Cohomology groups record these phenomena and define invariants such as arithmetic genus. This framework makes it possible to treat embeddings, differential forms, deformation theory, and intersection theory within a common language rather than through separate coordinate calculations.
Base fields, rational points, and arithmetic geometry
A variety defined over can be extended to a larger field , producing a base-changed variety . This separates its equations from the field in which solutions are sought. An irreducible variety over need not remain irreducible after extension to an algebraic closure; if it does, it is geometrically irreducible. Similarly, being reduced over is weaker than remaining reduced after every field extension. These distinctions explain why modern definitions often specify whether a variety is integral, geometrically integral, or smooth over its base.
The real affine variety provides a simple example of field dependence. Its defining polynomial is irreducible over the reals, and its real point set consists only of the origin. Over the complex numbers, the equation factors as , revealing two lines. Thus the dimension and components of a variety cannot generally be read from the appearance of its real points. Real algebraic geometry additionally studies inequalities, producing semialgebraic sets rather than only varieties.
Over a finite field , the sets are finite, and their cardinalities encode geometric information. For a smooth projective geometrically connected curve of genus ,
This is the Hasse–Weil bound. The counts over all finite extensions are packaged into the zeta function
The Weil conjectures connected its properties with topological analogues; their proof, completed by Pierre Deligne in 1974, showed the depth of the relationship between counting solutions and geometry.
Over the rational numbers or other number fields, the main objects are rational points and their distribution. A variety may have points over the real numbers and every -adic completion while having no rational point. Thus the local solvability of equations does not guarantee global solvability. On elliptic curves over number fields, rational points form a finitely generated abelian group, a statement of the Mordell–Weil theorem.
Arithmetic applications frequently use reduction modulo primes, comparing a variety over a number field with related varieties over finite fields. Elliptic-curve cryptography uses finite groups of points, while algebraic-geometric error-correcting codes evaluate functions on curves at selected rational points. These applications depend on the same structural ingredients—function fields, divisors, and point counts—that arise in the abstract theory.
Historical foundations and computational methods
The origins of algebraic varieties lie in the study of polynomial equations and analytic geometry. Seventeenth-century coordinate methods associated with René Descartes and Pierre de Fermat made curves accessible through equations. Nineteenth-century projective geometry clarified points at infinity and intersection behavior, while Bernhard Riemann’s work connected algebraic curves with complex analysis. By the late nineteenth century, algebraic surfaces had become a major subject, and the Italian school developed influential geometric classifications, sometimes using arguments whose rigorous foundations were established only later.
The algebraic foundations grew from ideal theory and finiteness theorems. David Hilbert’s basis theorem and Nullstellensatz established essential links between equations, ideals, and geometric sets. Emmy Noether and other algebraists developed the ring-theoretic language needed to study these structures systematically. In the twentieth century, Oscar Zariski recast local and birational questions using commutative algebra, and André Weil developed foundations suited to arbitrary fields. Alexander Grothendieck’s scheme theory, developed from the late 1950s, unified varieties with arithmetic spaces and made families and cohomology central tools.
Explicit calculation remains indispensable. Gröbner bases, introduced algorithmically through Bruno Buchberger’s work in 1965, replace a generating set of a polynomial ideal by one adapted to a chosen monomial order. They support ideal-membership testing, elimination of variables, and the computation of dimensions and algebraic relations. With an elimination order, a Gröbner basis for an ideal in can produce generators for its intersection with a smaller polynomial ring. Geometrically, elimination describes algebraic relations satisfied by projections of solution sets.
Other computational tasks include decomposing algebraic sets into irreducible components, computing singular loci through Jacobian ideals, and determining degrees or Hilbert polynomials. Primary decomposition retains information beyond irreducible components, including embedded structures. Numerical algebraic geometry approaches complex solution sets through approximate points and polynomial homotopy continuation, while symbolic methods work with exact algebraic expressions. The two approaches address different aspects of the same equations and are often combined.
Applications also produce varieties whose geometry constrains possible models. Rank conditions on a matrix are given by vanishing minors, defining determinantal varieties. These occur in low-rank approximation and tensor problems. In algebraic statistics, polynomial parametrizations of some probabilistic models yield algebraic relations among observable probabilities. The corresponding complex variety records polynomial constraints, while the statistical model itself also requires real inequalities, nonnegativity, and normalization. Keeping these distinctions explicit is essential when translating algebraic calculations back into an applied problem.