A Riemann surface is a connected manifold that locally resembles the plane of complex numbers, with coordinate changes given by holomorphic functions. Although its complex dimension is one, its real dimension is two. The complex structure makes it possible to define differentiation, integration, and analytic continuation intrinsically, without choosing a single global coordinate.
Riemann surfaces originated in the study of multivalued functions such as square roots and logarithms. Replacing the complex plane by a suitable surface turns these expressions into single-valued functions and reveals how their branches fit together. The subject subsequently became a meeting point of complex analysis, topology, differential geometry, and algebraic geometry. Its central results explain how local analytic coordinates constrain global shape, how meromorphic functions encode compact surfaces as algebraic curves, and how all simply connected surfaces arise from three basic models.
Complex structures and local coordinates
Formally, a Riemann surface is a connected, Hausdorff, second-countable topological space equipped with a complex atlas. Each chart is a homeomorphism , where is an open subset of and is an open subset of . Whenever two chart domains overlap, the transition function
must be holomorphic. Atlases define the same complex structure when their union still has holomorphic transition functions. The Hausdorff and countability conditions exclude pathological spaces and place the theory within the usual framework of manifold topology.
Holomorphic transition functions are much more restrictive than arbitrary smooth coordinate changes. A holomorphic bijection between planar domains has a nonzero derivative, and its real differential is multiplication by a nonzero complex number. It therefore preserves angles and orientation. Every Riemann surface consequently has a natural orientation and a conformal structure: angles are meaningful even though a preferred unit of length is not. In the smooth category, an oriented surface equipped with a conformal class of Riemannian metrics determines a Riemann surface through local isothermal coordinates.
A function is holomorphic if its expression in every local coordinate is holomorphic in the ordinary sense. Equivalently, its local real and imaginary parts satisfy the Cauchy–Riemann equations. A map between two Riemann surfaces is holomorphic when its expressions in source and target charts are holomorphic. A holomorphic map with a holomorphic inverse is called biholomorphic, and biholomorphism is the natural notion of equivalence in the subject. It is stronger than homeomorphism and usually stronger than smooth equivalence: topologically identical surfaces can carry inequivalent complex structures.
A Riemann surface should not be confused with a surface supplied with a particular Riemannian metric. A compatible metric has the local form
Multiplying this metric by any positive smooth function changes lengths but not the complex structure. Nor does a Riemann surface require an embedding in three-dimensional space. Its essential data are intrinsic coordinate charts and their analytic compatibility. Unless explicitly stated otherwise, the term refers to a surface without boundary; bordered Riemann surfaces require a modified local definition near boundary points.
Basic examples and the resolution of multivalued functions
Every connected open subset of is a Riemann surface with its usual coordinate. More globally, the Riemann sphere is
with coordinate away from infinity and coordinate near infinity. Their transition function is holomorphic wherever both are defined. Topologically this is a sphere, and analytically it is the complex projective line . A meromorphic function on any Riemann surface can be viewed as a holomorphic map to the Riemann sphere, with poles mapped to infinity.
The square root illustrates why surfaces entered complex analysis. The equation does not define a single-valued holomorphic function of on : continuation around the origin exchanges the two values. On the surface
however, is an ordinary global holomorphic function. Indeed, itself is a global coordinate on , while the projection to the -plane is the map . The point above zero is not a singularity of the surface; it is a ramification point of the projection. Describing this construction as two slit planes glued along their edges is useful for visualization, but the cuts are auxiliary choices rather than intrinsic features.
The logarithm leads to an infinite-sheeted example. The exponential map
covers the punctured plane, and translation by , for an integer , leaves its value unchanged. The logarithm becomes the single-valued coordinate on this covering surface. Unlike the square-root example, no finite number of sheets suffices. More generally, analytic continuation follows local function elements along paths; different closed paths may return different elements. This dependence is called monodromy and explains why the topology of the underlying domain governs the behavior of analytic branches.
Another fundamental construction is the complex torus
Translations by lattice vectors identify points, and sufficiently small disks descend to complex coordinate charts. The quotient is compact and topologically a torus. It also carries addition inherited from . Different lattice shapes can yield non-biholomorphic complex tori even though their underlying topological surfaces are identical. This example already displays a central distinction: the number of handles determines a compact surface topologically, but generally does not determine its complex structure.
Holomorphic maps, branching, and topological degree
The local behavior of a nonconstant holomorphic map is exceptionally rigid. If is such a map and , suitable local coordinates centered at and put it in the form
where is a positive integer. When , the map is locally biholomorphic. When , is a ramification point, is its ramification index, and is a branch value. This normal form follows by factoring the first nonzero term in the local power series and absorbing a nonvanishing holomorphic factor into the coordinate.
If and are compact and connected, every nonconstant holomorphic map is surjective and has a finite degree . Away from its finitely many branch values, it is a -sheeted covering map. Over every point , multiplicities satisfy
The degree can thus be computed by counting the inverse images of any regular value. The map on the Riemann sphere has degree , with ramification index at both zero and infinity. Infinity is essential to the global count, even when a formula is initially written in an affine coordinate.
The relation between branching and topology is expressed by the Riemann–Hurwitz formula:
Here and are the genera of the compact surfaces. The formula can be understood by triangulating the target with branch values among the vertices and lifting the triangulation: ramification reduces the number of lifted vertices relative to an unbranched covering. Since a compact oriented genus- surface has Euler characteristic , the resulting counting identity gives the formula.
For example, a double cover of the sphere with distinct simple branch values has genus . Smooth compactifications of equations
with distinct , provide such covers. For , these are hyperelliptic surfaces. Riemann–Hurwitz also explains why a holomorphic self-map of degree greater than one cannot exist on a compact surface of genus at least two: the right-hand side would exceed the left-hand side. Questions about continuation and sheets can also be expressed through the fundamental group: loops in the complement of branch values permute the sheets, producing a monodromy representation into a permutation group.
Uniformization and canonical geometry
The uniformization theorem states that every simply connected Riemann surface is biholomorphic to exactly one of three models: the Riemann sphere, the complex plane, or the unit disk
This is a classification by complex structure, not merely by topology. The plane and disk are homeomorphic but not biholomorphic. For instance, a biholomorphism from the plane to the disk would be a bounded nonconstant entire function, contradicting Liouville’s theorem. Every connected Riemann surface has a simply connected universal covering surface, so uniformization presents it as a quotient of one of these models by a suitably acting group of holomorphic automorphisms.
For compact surfaces, the universal covering model is determined by genus. Genus zero gives the sphere itself. Genus one gives , with the surface represented as a lattice quotient. Genus at least two gives the disk, or equivalently the upper half-plane. In the latter case, the surface is a quotient by a torsion-free discrete group of orientation-preserving hyperbolic isometries, commonly called a Fuchsian group. These quotient descriptions translate between complex analysis and group theory: the abstract fundamental group records topology, while its geometric action records the complex structure.
The three models support natural constant-curvature metrics. On the unit disk, the Poincaré metric is
with Gaussian curvature . Every biholomorphic automorphism of the disk preserves this metric, so it descends to any surface uniformized by the disk. On a compact genus- surface with , the resulting metric is the unique compatible metric of curvature . Its area is
by the Gauss–Bonnet theorem. A torus instead admits compatible flat metrics, unique after fixing total area, whereas the sphere admits compatible positive constant-curvature metrics without the same uniqueness.
Uniformization also exposes limits of purely topological classification for noncompact surfaces. The plane and disk are both simply connected and noncompact, yet have different analytic behavior. The plane admits no nonconstant bounded holomorphic functions, while the disk admits many. More generally, the geometry of an open surface’s ends and the action of its covering group affect potential theory, harmonic functions, and the existence of Green’s functions. The terms elliptic, parabolic, and hyperbolic are often used for the three universal-cover types, although terminology involving potential-theoretic parabolicity requires care because it is not identical in every setting.
Meromorphic functions, divisors, and Riemann–Roch
Compactness creates a basic obstruction: every holomorphic function on a compact connected Riemann surface is constant, by the maximum modulus principle. Meromorphic functions are therefore the appropriate global substitutes for coordinates. Locally a nonzero meromorphic function has the form
where is holomorphic and nonvanishing and . Positive records a zero and negative a pole. These local orders are independent of the chosen coordinate. Globally, they are organized into a divisor
A divisor is any finite formal integer combination of points, , and its degree is . It is effective when every coefficient is nonnegative. Divisors of meromorphic functions are called principal; on a compact surface every principal divisor has degree zero. Thus a meromorphic function has equally many zeros and poles when counted with multiplicity. Two divisors are linearly equivalent when their difference is principal. This equivalence retains information about whether prescribed zeros and poles can be realized by a single meromorphic function, rather than merely whether their total degrees match.
For a divisor , define
Positive coefficients in permit poles; negative coefficients require zeros. The Riemann–Roch theorem states
where is a canonical divisor, obtained from a nonzero meromorphic differential. Its degree is . The correction term measures the obstruction to the naive count of allowed principal parts. In particular, if , then
Riemann–Roch ensures that every compact Riemann surface has nonconstant meromorphic functions and hence admits a finite branched map to the sphere. It also makes geometric constructions quantitative. On the sphere, for , corresponding to polynomials of degree at most . On a genus-one surface, , so no nonconstant function can have only one simple pole, whereas . In higher genus, sufficiently positive divisors supply enough functions to distinguish points and tangent directions. The language of holomorphic line bundles packages these spaces as spaces of sections; Riemann–Roch then becomes a relation between their dimensions and topological degree.
Holomorphic differentials, periods, and the Jacobian
A holomorphic differential is a globally defined one-form locally expressible as
with holomorphic. Under a coordinate change, the coefficient transforms by the derivative, so these forms are sections of the canonical line bundle rather than ordinary functions. On a compact genus- Riemann surface, their complex vector space has dimension exactly . This supplies an analytic interpretation of genus. The sphere has no nonzero holomorphic differentials, a complex torus has the nowhere-vanishing differential , and for every nonzero holomorphic differential has zeros counted with multiplicity.
Integrating a holomorphic differential along a path gives a locally holomorphic primitive, but different paths can produce different values. Their differences are periods, the integrals around closed cycles. Choose a symplectic basis
of the first integral homology group, so that intersects with intersection number , while intersections within either family vanish. A basis of differentials can be normalized by
The remaining periods form a matrix . The Riemann bilinear relations imply that is symmetric and its imaginary part is positive definite.
These periods define the Jacobian
It is a complex torus of dimension , equipped with a canonical principal polarization coming from the intersection pairing. Choosing a base point gives the Abel–Jacobi map
Changing paths changes the vector by a period, so the map is well defined. For genus one it identifies the surface with its Jacobian after choosing an origin; for higher genus it places a one-dimensional curve inside a higher-dimensional torus.
Abel’s theorem relates the Jacobian to divisors: a degree-zero divisor is principal precisely when its Abel–Jacobi class vanishes. Jacobi inversion, in a standard formulation, states that every point of the Jacobian is represented by a degree-zero divisor of the form . These statements convert questions about meromorphic functions into additive relations in a complex torus. The theta function
provides an analytic means of studying this torus and its distinguished divisors. Positive definiteness of ensures convergence, and the resulting functions generalize the elliptic functions associated with genus-one surfaces.
Equivalence with complex algebraic curves
Every smooth complex algebraic curve has an underlying Riemann surface. Locally, an equation defines a complex coordinate whenever at least one partial derivative of is nonzero, by the holomorphic implicit function theorem. Projective curves are compact in their analytic topology. The converse is a central theorem: every compact Riemann surface is biholomorphic to a smooth projective algebraic curve over . Moreover, holomorphic maps between such surfaces correspond to algebraic morphisms. The analytic and algebraic descriptions therefore agree not merely on individual objects but also on the natural maps between them.
One route to the converse uses Riemann–Roch. A divisor of sufficiently large degree yields enough meromorphic functions, interpreted as sections of a line bundle, to embed the surface holomorphically into complex projective space. A line bundle of degree at least is very ample, so its complete linear system gives such an embedding. The resulting compact complex submanifold is algebraic. From the function-field viewpoint, a compact surface has a field of meromorphic functions of transcendence degree one over ; a nonconstant meromorphic function makes this field a finite extension of .
The correspondence must distinguish smooth surfaces from singular plane models. A polynomial equation may have a node, where two branches cross, or a cusp, where the local parameterization degenerates. Such a singular curve is not itself a Riemann surface at the singular point. Its normalization replaces the local singularity by smooth branch parameters. For example, the cusp is parametrized by , ; is a regular coordinate on its normalization. At a node, normalization separates the two local branches into distinct points.
For a smooth plane curve of degree , the genus is
Thus smooth conics have genus zero, smooth cubics genus one, and smooth quartics genus three. Singularities reduce the genus of the normalization relative to this formula. A complex elliptic curve can be represented by a smooth projective cubic with a chosen origin, or analytically by a complex torus. Its Weierstrass functions satisfy
connecting lattice periods with a cubic equation. This equivalence is specific to compact surfaces: a general noncompact Riemann surface need not be the analytification of an algebraic curve of finite type.
Moduli, deformation, and automorphisms
The moduli space records compact Riemann surfaces of genus up to biholomorphism. Genus alone is enough in genus zero: every compact genus-zero surface is biholomorphic to the sphere. In genus one, a torus can be written
Two parameters describe biholomorphic tori precisely when they are related by a transformation
This action reflects a change of oriented lattice basis, rather than a change in the underlying complex torus.
For , moduli has complex dimension . A more manageable intermediate object is Teichmüller space, whose points are complex structures together with a marking by a fixed oriented topological surface, modulo an appropriate equivalence. Retaining the marking removes the identifications caused by changing the topological description. Teichmüller space has real dimension and is contractible. The mapping class group—orientation-preserving self-homeomorphisms modulo isotopy—acts on it, and the quotient gives moduli. Finite automorphism groups produce orbifold features, so moduli is not everywhere an ordinary manifold.
The dimension count has complementary analytic and geometric explanations. Holomorphic quadratic differentials, locally written , form the cotangent space to Teichmüller space and have dimension . In hyperbolic geometry, a closed genus- surface can be cut along disjoint simple closed curves into pairs of pants. The lengths of the cutting geodesics and corresponding twist parameters give real Fenchel–Nielsen coordinates. These descriptions connect deformation of complex charts with changes in the lengths and gluing of hyperbolic pieces.
Automorphisms exhibit a sharp distinction between genera. The sphere has the Möbius transformation group ; a torus has all translations, together with a finite group of automorphisms fixing its origin. For , the holomorphic automorphism group is finite and satisfies Hurwitz’s bound
The genus-three Klein quartic attains the bound with 168 automorphisms. Moduli also includes limiting behavior: shrinking suitable closed geodesics leads to nodal curves rather than smooth surfaces. The Deligne–Mumford compactification adds stable nodal curves, organizing degenerations into a compact algebraic parameter space and distinguishing genuine limits from surfaces that merely escape every compact subset of ordinary moduli.
Historical development and broader uses
Bernhard Riemann introduced the geometric viewpoint in his 1851 doctoral dissertation on complex function theory and developed it further in his 1857 work on Abelian functions. His innovation was to treat the domain of a multivalued analytic expression as a surface assembled from local analytic elements. This made topology part of function theory rather than an external complication. Earlier work on elliptic integrals by Niels Henrik Abel and Carl Gustav Jacobi had already revealed the importance of periods, inversion, and functions with two independent periods; Riemann extended the framework to arbitrary genus.
The nineteenth-century theory combined geometric insight with methods whose foundations were still being clarified. Riemann used the Dirichlet principle to establish existence results through energy minimization. Karl Weierstrass identified gaps in the general justification of this principle, and later developments in potential theory and the calculus of variations, including work by David Hilbert, supplied rigorous existence methods under suitable hypotheses. Gustav Roch published the theorem now bearing his and Riemann’s names in 1865. Uniformization emerged through the work of Henri Poincaré and Paul Koebe, who established its general form in 1907.
During the twentieth century, Riemann surfaces became a testing ground for interactions among analysis, topology, and geometry. Hermann Weyl’s 1913 book Die Idee der Riemannschen Fläche helped establish an intrinsic formulation independent of pictures of overlapping sheets. Teichmüller theory then made deformation of complex structures a subject in its own right, with quasiconformal maps providing a quantitative measure of distortion. Hodge theory interpreted holomorphic differentials through harmonic representatives of cohomology classes, while sheaf cohomology and line bundles reorganized Riemann–Roch into a form that extends to higher-dimensional complex and algebraic geometry.
Applications retain the same local-to-global structure. In boundary-value problems, conformal maps transport Laplace’s equation and harmonic functions between planar domains or surfaces. In integrable systems, algebraic curves and theta functions encode special solutions of nonlinear differential equations. In two-dimensional conformal field theory and string perturbation theory, Riemann surfaces serve as spaces on which fields are defined, with genus and moduli specifying global geometry. Numerical conformal mapping and discrete surface parameterization use related ideas to construct coordinates on meshes. In each setting, the relevant data include more than the underlying topological surface: periods, marked points, complex structure, and sometimes a chosen metric or differential determine the analytic problem.