Loading field map
Riemannian geometry is the branch of mathematics that studies smooth curved spaces using the tools of calculus and linear algebra. Its central object is the Riemannian manifold: a space that locally resembles ordinary Euclidean space but may be curved globally, equipped with a smoothly varying choice of inner product—the metric—on the tangent space at each point. This metric allows one to measure lengths of curves, angles between vectors, volumes of regions, and, crucially, to define the notion of curvature that quantifies how the space deviates from flatness.
The field’s foundational question is deceptively simple: What does it mean for a space to be curved, and how can that curvature be measured intrinsically—without reference to an ambient space in which the manifold sits? From this question flows a rich theory connecting analysis, topology, and geometry, with consequences ranging from the mathematical structure of general relativity to the shape of data in modern machine learning.
Before Riemann, the geometry of curved surfaces was studied primarily as the geometry of surfaces embedded in three-dimensional Euclidean space. A sphere, a saddle, or a torus was understood through its position in the surrounding flat space. The German mathematician Bernhard Riemann, in his 1854 habilitation lecture, proposed a radical shift: the curved space itself should carry all the information needed for its geometry, without any reference to an external embedding. This is the intrinsic viewpoint.
On a Riemannian manifold, the metric is a smoothly varying inner product \( gp \) on the tangent space \( TpM \) at each point \( p \). Given a smooth curve \( \gamma(t) \), its length is the integral of \( \sqrt{g_{\gamma(t)}(\gamma'(t), \gamma'(t))} \), the speed as measured by the metric. The distance between two points is the infimum of lengths of curves joining them, making the manifold a metric space. This distance function, in turn, determines the topology of the manifold in a natural way.
The intrinsic viewpoint has a profound consequence: curvature becomes a property of the manifold itself, not of how it sits in space. The Riemann curvature tensor is built from the metric and its first and second derivatives. It measures the failure of parallel transport to commute around infinitesimal loops, or equivalently, the extent to which the manifold deviates from being locally Euclidean. A manifold is flat—locally isometric to Euclidean space—if and only if this tensor vanishes identically.
Curvature is not a single number but a family of related objects, each capturing a different aspect of the geometry. The sectional curvature assigns to each two-dimensional plane in the tangent space a number: the Gaussian curvature of the surface swept out by geodesics tangent to that plane. When the sectional curvature is constant across all planes and points, the manifold is a space form: spherical, Euclidean, or hyperbolic, depending on the sign.
Two other contractions of the curvature tensor play central roles. The Ricci curvature averages sectional curvatures over planes containing a given direction; it controls how volumes of small geodesic balls deviate from their Euclidean values. The scalar curvature is a further contraction, a single number at each point that averages all sectional curvatures. In general relativity, the Ricci curvature appears in the Einstein field equations, relating the geometry of spacetime to its matter content.
The study of curvature is organized around a central tension: curvature constrains topology, and topology constrains curvature. The Gauss–Bonnet theorem for surfaces is the simplest example: the integral of Gaussian curvature over a closed surface equals \( 2\pi \) times its Euler characteristic, a purely topological invariant. In higher dimensions, the relationship is far more subtle, but the guiding principle remains. The Bonnet–Myers theorem states that a complete manifold with Ricci curvature bounded below by a positive constant has finite diameter and finite fundamental group. The sphere theorem asserts that a simply connected complete manifold with sectional curvature pinched between 1 and 4 is homeomorphic to a sphere. These results exemplify how local curvature conditions force global topological conclusions.
Riemannian geometry has developed through two deeply intertwined but methodologically distinct traditions.
The analytic tradition treats the metric as a solution to partial differential equations. The geodesic equation, a second-order ODE, defines the paths of shortest length locally. The Laplace–Beltrami operator, the natural generalization of the Laplacian to a Riemannian manifold, governs diffusion and harmonic functions. The heat equation on a manifold, whose fundamental solution describes how heat spreads, encodes geometric information in its short-time asymptotics: the heat kernel expansion recovers the scalar curvature, and more generally, the full curvature tensor.
This tradition reached maturity in the mid-twentieth century with the development of geometric analysis. The Hodge theory of harmonic forms relates the de Rham cohomology of a manifold to solutions of elliptic equations, linking topology to analysis. The Atiyah–Singer index theorem connects the analytic index of an elliptic operator to a topological expression involving characteristic classes. The Yau solution of the Calabi conjecture—the existence of Ricci-flat Kähler metrics under a topological condition—demonstrated that solving nonlinear PDEs can construct metrics with prescribed curvature properties.
The geometric tradition emphasizes the global behavior of geodesics and the structure of the manifold as a metric space. The Hopf–Rinow theorem characterizes completeness: a Riemannian manifold is geodesically complete if and only if it is complete as a metric space, in which case any two points are joined by a minimizing geodesic. The study of geodesic flows—the dynamics of particles moving freely on the manifold—connects geometry to ergodic theory and dynamical systems. On negatively curved manifolds, geodesic flows exhibit chaotic behavior, while on positively curved manifolds they tend to be more regular.
The comparison geometry program, initiated by Rauch and developed by Gromov and others, compares a manifold with curvature bounds to a model space of constant curvature. The Toponogov comparison theorem and the Bishop–Gromov volume comparison allow one to transfer metric inequalities from model spaces to general manifolds with bounded curvature. This approach has produced striking results, including Gromov’s compactness theorem: families of manifolds with uniform curvature and volume bounds have convergent subsequences in a suitable sense, allowing the study of geometric limits and degenerations.
A central theme of modern Riemannian geometry is the search for canonical metrics—metrics that are distinguished by curvature conditions and whose existence reflects the topology of the underlying manifold. The most famous example is the uniformization theorem for surfaces: every closed surface admits a metric of constant curvature, uniquely determined up to scaling. In higher dimensions, no such simple statement holds, but the geometrization conjecture, proved by Perelman, asserts that every closed three-manifold can be decomposed into pieces, each admitting one of eight homogeneous geometries. The proof, via Ricci flow—a PDE that evolves the metric in the direction of its Ricci curvature—represents a triumph of the analytic tradition.
In higher dimensions, the Kähler geometry of complex manifolds provides a rich setting for canonical metric problems. A Kähler manifold is a complex manifold with a Riemannian metric compatible with the complex structure. The Calabi–Yau theorem guarantees the existence of Ricci-flat Kähler metrics under a topological condition, and the Kähler–Einstein problem asks when a Kähler manifold admits a metric with Ricci curvature proportional to the metric itself. This problem, resolved through the Yau–Tian–Donaldson conjecture, connects algebraic geometry, PDE theory, and geometric invariant theory.
The positive mass theorem in general relativity, proved by Schoen and Yau using minimal surface techniques, exemplifies the interplay between geometry and physics. It states that the total mass of an asymptotically flat spacetime is nonnegative, and zero only for flat space. The proof uses the geometry of minimal hypersurfaces, illustrating how Riemannian techniques resolve questions in mathematical physics.
Contemporary Riemannian geometry is characterized by the proliferation of tools and the blurring of traditional boundaries. Geometric measure theory provides the machinery to study minimal surfaces and singular geometric objects, essential for understanding the variational problems that arise in the field. Scalar curvature has emerged as a particularly subtle invariant, with the Gromov–Lawson conjecture and its resolution connecting it to the topology of spin manifolds and the index theory of Dirac operators.
The synthetic approach, pioneered by Gromov and developed through the theory of Alexandrov spaces and Ricci limit spaces, extends Riemannian geometry to singular spaces. These tools allow the study of limits of Riemannian manifolds under curvature bounds, where smoothness may be lost but metric and curvature information survive. This perspective has proven essential in understanding degenerations and in applying geometric ideas to other fields.
The interaction with data science has opened new frontiers. Manifold learning assumes that high-dimensional data lies on a low-dimensional Riemannian manifold, and the geometry of that manifold—its geodesics, curvature, and metric—becomes a tool for dimensionality reduction and clustering. Optimal transport on Riemannian manifolds, the Wasserstein geometry of probability measures, and the Ricci flow on graphs and discrete spaces represent active areas where Riemannian ideas are adapted to non-smooth settings.
Throughout its history, Riemannian geometry has maintained a distinctive character: a field where local computations with tensors and PDEs yield global topological conclusions, where the intrinsic geometry of a space is studied for its own sake and for its applications to physics, topology, and increasingly, to the analysis of complex data. Its central questions—what curvature determines, what metrics exist, how geometry constrains topology—remain open in their most general forms, and the field continues to develop through the productive tension between its analytic and geometric traditions.