Loading field map
Arithmetic geometry is the study of solutions to polynomial equations using the combined tools of number theory and algebraic geometry. Its central objects are algebraic varieties—geometric shapes defined by polynomial equations—but the equations are considered over number-theoretic domains such as the rational numbers, finite fields, or rings of integers. The field asks how the purely geometric properties of these shapes (their curves, surfaces, and higher-dimensional analogues) constrain and explain the arithmetic behavior of their solutions.
To understand arithmetic geometry, one must first see its two constituent traditions. The older is Diophantine analysis, the study of integer and rational solutions to polynomial equations, named after the ancient Greek mathematician Diophantus. For centuries, this was a collection of clever but ad hoc methods for individual equations: finding Pythagorean triples, proving Fermat's Last Theorem for particular exponents, or determining which integers are sums of two squares. The subject had no unifying geometric picture; each equation was treated on its own terms.
The younger parent is algebraic geometry, which matured in the nineteenth and twentieth centuries as the study of solution sets to polynomial equations over the complex numbers or other algebraically closed fields. Here the emphasis is on structure: a curve has a genus (a measure of how many "holes" it has), a surface has intersection numbers, and varieties have cohomology groups that encode their topology. These invariants are continuous or topological in flavor, and they are computed over fields where every polynomial has roots.
Arithmetic geometry arises from the observation that these two subjects are not merely parallel but deeply intertwined. A polynomial equation with integer coefficients defines a geometric object over the complex numbers, but the same equation also has solutions modulo prime numbers, over finite fields, and over the rational numbers. The central insight of the field is that these different "worlds" of solutions are governed by the same underlying object, and that geometric invariants computed in one world control arithmetic facts in another.
The field is organized around a small number of enduring problems, each of which can be stated simply but has driven decades of research.
The structure of rational points. Given a polynomial equation in several variables, when does it have a rational solution? If it has one, how many are there, and how are they distributed? For a single equation in two variables (a curve), the answer depends almost entirely on the genus. Genus zero curves with one rational point have infinitely many, parameterized by rational functions. Genus one curves (elliptic curves) either have no rational points, or finitely many torsion points plus a finitely generated group of "free" points—this is the Mordell–Weil theorem. For genus at least two, Faltings' theorem (formerly the Mordell conjecture) states that there are only finitely many rational points. For higher-dimensional varieties, the situation is far less settled, and the question of when rational points are "dense" in the topological sense is a major open area.
Integral points and Diophantine approximation. A related but distinct question asks for solutions in integers rather than rationals. The finiteness of integer points on many curves follows from deep results in Diophantine approximation, such as Siegel's theorem, which in turn connects to the geometry of the curve at infinity. The modern formulation uses the language of heights—a measure of the "size" of a rational point—and relates the growth of heights to geometric invariants.
The Hasse principle and obstructions to it. A natural strategy for finding rational solutions is to check for solutions modulo every prime, and also over the real numbers. If a variety has solutions everywhere locally, does it have a global rational solution? This "local-to-global" principle holds for some simple classes (quadratic forms in many variables) but fails in general. The systematic study of why it fails—through the Brauer group, the Tate–Shafarevich group, and more refined obstructions—is a central theme. The question of whether the Brauer–Manin obstruction is the only obstruction for rationally connected varieties is a major open problem.
Modularity and reciprocity. The most celebrated achievement in the field is the proof that elliptic curves over the rational numbers are modular: they arise from modular forms, which are highly symmetric analytic functions on the upper half-plane. This statement, proved by Andrew Wiles and Richard Taylor for semistable curves and later completed by others, implies Fermat's Last Theorem. More broadly, the Langlands program conjectures a vast web of connections between Galois representations (arithmetic objects) and automorphic forms (analytic objects). Arithmetic geometry provides the geometric bridge: the cohomology of algebraic varieties carries Galois actions, and understanding these actions is the key to the correspondence.
The geometry of numbers over finite fields. The Weil conjectures, proved by Pierre Deligne, describe the number of solutions to polynomial equations over finite fields. They state that these counts are governed by the same topological invariants (Betti numbers) that describe the complex solutions, and that the "error terms" satisfy analogues of the Riemann hypothesis. This result is foundational: it shows that the arithmetic of equations over finite fields is a shadow of their complex geometry, and it provides the technical machinery (étale cohomology) on which much of modern arithmetic geometry rests.
The modern subject took shape in the mid-twentieth century through the work of André Weil, who systematically imported the language of algebraic geometry into number theory. Weil's key insight was that to count solutions over finite fields, one needed a cohomology theory that worked in positive characteristic—the usual topological methods failed there. His conjectures, formulated in 1949, set the agenda for the next three decades.
The technical breakthrough came from Alexander Grothendieck, who rebuilt algebraic geometry on a foundation of schemes and developed étale cohomology, a cohomology theory that behaves like singular cohomology but works over arbitrary fields. This provided the tools to prove the Weil conjectures and, more importantly, created a framework in which geometric ideas could be applied uniformly across all characteristics. Grothendieck's school also introduced the notion of a motive, a hypothetical universal cohomology theory that would unify all the different cohomology theories (étale, de Rham, crystalline) attached to a variety. Motives remain partly conjectural but organize much of the field's thinking.
A second major strand developed around elliptic curves. The Mordell–Weil theorem (proved by Weil in 1928, generalizing Mordell's earlier work) showed that rational points on elliptic curves form finitely generated abelian groups. The Birch and Swinnerton-Dyer conjecture, formulated in the 1960s, relates the rank of this group (the number of independent infinite-order points) to the behavior of the associated L-function at a special point. This conjecture, one of the Clay Millennium Problems, remains open and is a central driver of research.
The 1980s and 1990s saw the proof of two conjectures that had seemed unreachable. Gerd Faltings proved the Mordell conjecture in 1983, showing finiteness of rational points on high-genus curves. Then Wiles proved modularity for a large class of elliptic curves in 1995, with the full modularity theorem completed by Christophe Breuil, Brian Conrad, Fred Diamond, and Richard Taylor in 2001. These results demonstrated that the geometric and analytic perspectives could be combined to resolve problems that had resisted purely number-theoretic or purely geometric approaches for centuries.
Arithmetic geometry is not divided into rival schools in the way that, say, foundations of mathematics once were. Instead, it is organized around complementary methodological traditions that often overlap and borrow from each other. The distinctions below are real but porous; most researchers work in several of these modes.
The cohomological approach. This is the Grothendieck–Deligne tradition. Its organizing assumption is that the arithmetic of a variety is encoded in its cohomology groups, which carry actions of the Galois group of the base field. The method is to compute these cohomology groups, extract Galois representations from them, and then use representation theory to draw arithmetic conclusions. This approach proved the Weil conjectures and provides the language for the Langlands program. Its strength is its power and generality; its weakness is that the objects are highly abstract and often difficult to compute explicitly. The theory of motives is the attempt to push this approach to its logical conclusion by finding a universal cohomology theory.
The analytic and modular approach. This tradition, associated with the theory of modular forms and L-functions, studies arithmetic objects through their analytic avatars. Elliptic curves are paired with modular forms; the L-functions attached to varieties are studied as complex analytic functions. The Birch and Swinnerton-Dyer conjecture is the paradigm: it asserts that an analytic object (the order of vanishing of an L-function) determines an arithmetic object (the rank of a Mordell–Weil group). The strength of this approach is that it connects to deep analytic techniques and to the Langlands program's automorphic side. Its limitation is that the analytic objects are often only conjecturally related to the geometry, and proving these relations (as in modularity) is extremely hard.
The Diophantine approximation approach. This is the oldest tradition, descending from classical number theory. It studies rational points by approximating them with rational numbers of small denominator and using inequalities to show that too many good approximations cannot exist. The Thue–Siegel method and its modern descendants (the Bombieri–Pila method, the determinant method) prove finiteness results by showing that rational points of bounded height lie on few algebraic curves. This approach proved early versions of finiteness theorems and remains important for explicit bounds. Its limitation is that it works best for curves and hypersurfaces; it has not yet been extended to the full generality of the cohomological approach.
The p-adic approach. This tradition studies varieties over the p-adic numbers, the completions of the rationals with respect to a prime p. The p-adic world has its own geometry (rigid analytic geometry, perfectoid spaces) and its own cohomology theories (crystalline, de Rham). The key insight is that p-adic methods can interpolate between the finite field world and the complex world, providing information that neither extreme alone can give. This approach has produced the theory of p-adic L-functions, p-adic Hodge theory, and recent breakthroughs in the Langlands program. Its strength is that it is often more explicit and computable than the complex analytic approach; its weakness is that it is technically demanding and only sees one prime at a time.
The anabelian and homotopy-theoretic approach. A more recent development, associated with Alexander Grothendieck's later ideas and pursued by a smaller community, seeks to recover the arithmetic structure of a variety from its fundamental group or from higher homotopical invariants. The idea is that the Galois action on the fundamental group of a variety might be so rich that it determines the variety itself. This approach has produced striking results in special cases (the section conjecture for curves remains open) and has connections to homotopy theory through the theory of motives and derived algebraic geometry. It is the least developed of the major approaches but has attracted growing attention.
Contemporary arithmetic geometry is characterized by the deep integration of these approaches. The proof of Fermat's Last Theorem required the cohomological machinery (to construct Galois representations from elliptic curves), the analytic theory (to match these with modular forms), and the p-adic methods (to prove the necessary congruences). The Langlands program, which is the field's largest open project, explicitly aims to unify the cohomological and analytic perspectives across all dimensions.
Several active research fronts define the present shape of the field. The Langlands program seeks to establish correspondences between Galois representations and automorphic forms for all reductive groups, not just the special case of elliptic curves. Major progress has been made in recent years, particularly through the work of Laurent Lafforgue on function fields and the proof of local Langlands correspondences, but the global correspondence for number fields remains largely open. The Birch and Swinnerton-Dyer conjecture continues to drive research on elliptic curves, with substantial progress on the direction that bounds the rank by the order of vanishing. The section conjecture and related anabelian questions remain open but have stimulated new connections to homotopy theory. The theory of perfectoid spaces, developed by Peter Scholze, has revitalized p-adic geometry and provided new tools for attacking problems in mixed characteristic.
A notable feature of the current landscape is the increasing role of computation. While arithmetic geometry is a theoretical subject, computational experiments with elliptic curves, modular forms, and Galois representations have guided conjectures and provided evidence for decades. The databases of elliptic curves and modular forms maintained by various research groups are important resources, and computational methods have been essential in testing the Langlands correspondence in small cases.
The field also maintains a distinctive relationship with its own history. Unlike some areas of mathematics where older results are absorbed and forgotten, the classical problems of Diophantine analysis remain alive: Fermat's Last Theorem, the congruent number problem (which elliptic curves are related to right triangles with rational area), and the classification of rational points on curves are all still active research topics, now approached with vastly more powerful tools. The subject's continuity with its pre-modern past is part of its identity: it is not uncommon for a modern research paper to cite a result from the nineteenth century as a key ingredient.
The deepest open problems in arithmetic geometry share a common feature: they assert that two seemingly different descriptions of the same object—geometric and arithmetic, analytic and algebraic, local and global—must agree. The Birch and Swinnerton-Dyer conjecture equates an analytic invariant with an arithmetic one. The Langlands program equates Galois-theoretic objects with automorphic ones. The Hodge conjecture and the Tate conjecture equate topological or étale cohomology classes with algebraic cycles. This pattern of conjectured equivalences between different mathematical worlds is the field's signature, and the gradual proof of such equivalences is its main mode of progress.