Topos Theory

Topos theory is the branch of category theory that studies toposes: categories that behave like the category of sets, or like the category of sheaves on a space, but which are defined by abstract structural properties rather than by any particular construction.

5 frameworks, from 1958 to today · Category Theory · Mathematics

still a working view ended · each leads to its entry below

Overview

Topos theory is the branch of category theory that studies toposes: categories that behave like the category of sets, or like the category of sheaves on a space, but which are defined by abstract structural properties rather than by any particular construction1. The subject began as a tool for algebraic geometry, where Alexander Grothendieck and his school needed a way to make "a space" and "the sheaves on it" interchangeable1. It then split into two traditions that still define the field: the Grothendieck topos, a geometric object built from a site (a category equipped with a notion of covering), and the elementary topos, an axiomatic notion of "a universe of sets" that turned out to be a foundation for mathematics and a semantics for higher-order logic12. The stakes are large: topos theory is where geometry, logic, and set theory meet, and it is the technical heart of much of modern categorical logic and of the "spaces as categories" viewpoint in geometry.

Sheaves and the problem of local data

The prehistory of topos theory is the theory of sheaves1. A sheaf on a topological space assigns to each open set a set of "sections" (think of continuous functions, or functions satisfying some local condition), such that sections agree on overlaps and are determined by their restrictions to a cover3. The sheaf condition says: a family of local sections that agree on all intersections glues to a unique global section3. This is the precise form of the idea that a sheaf is "data that can be checked locally."

By the 1950s sheaf theory was central to several complex variables and algebraic topology. Jean Leray had introduced sheaves in the 1940s while a prisoner of war, and Henri Cartan's seminar in Paris developed sheaf cohomology. The category of sheaves on a space, with its natural transformations as arrows, was known to be a well-behaved category—an abelian category with enough injectives—but it was treated as a category associated to a space, not as a space in its own right.

Grothendieck and the étale topos

The decisive turn came from Grothendieck's algebraic geometry in the late 1950s and early 1960s. To prove the Weil conjectures, Grothendieck needed cohomology theories for schemes over a field that behaved like the classical cohomology of complex varieties. The Zariski topology on a scheme is too coarse: open sets are too large, and there are too few of them. Grothendieck's insight was to replace open subsets by étale maps—maps that are locally isomorphisms in a suitable sense—and to define a "topology" in which covers are families of étale maps that are jointly surjective.

This required a generalization. A site is a category equipped with a collection of covering families, satisfying axioms that make "gluing" meaningful4. A sheaf on a site is a functor to sets that sends covering families to compatible families of sections. The category of sheaves on a site is a Grothendieck topos1. Grothendieck's school, notably in the Séminaire de Géométrie Algébrique du Bois Marie (SGA), developed this machinery in the early 1960s; SGA 4, written with Jean-Louis Verdier and others, is the foundational text. The étale topos of a scheme, and the étale cohomology built from it, gave Grothendieck the tools that led to the proof of the Weil conjectures by Pierre Deligne in 1974.

The conceptual payoff was that a site and its topos of sheaves carry the same information for cohomological purposes: many different sites give equivalent toposes1. A topos became a generalized space, and geometric morphisms between toposes—adjoint pairs of functors with the left adjoint preserving finite limits—became the generalized continuous maps. This is the origin of the slogan that topos theory is "topology without points."5

Giraud's theorem and the geometric characterization

A natural question was: which categories are Grothendieck toposes? The answer, due to Jean Giraud in the early 1960s, is a list of exactness conditions: a Grothendieck topos is a category with finite limits, arbitrary small coproducts that are disjoint and stable under pullback, and a small generating set, among other properties. Giraud's theorem made "Grothendieck topos" a purely categorical notion, independent of any site, and it is the version used in most modern treatments1. It also made clear that toposes form a class of categories closed under many constructions, and it set the stage for the comparison with the elementary notion.

Lawvere, Tierney, and the elementary topos

The second tradition began independently, from logic and foundations. In the early 1960s F. William Lawvere, working on categorical algebra and the foundations of geometry, became convinced that category theory could serve as a foundation for mathematics and that the category of sets should be characterized by universal properties rather than by a membership relation. With Myles Tierney, in the late 1960s, Lawvere isolated the axioms of an elementary topos: a category with finite limits, a subobject classifier (an object Ω such that subobjects of any object correspond to arrows into Ω), and power objects (exponentials). The category of sets is the basic example; the subobject classifier is the two-element set of truth values, and a subobject corresponds to its characteristic function6.

The elementary axioms are first-order and do not require a site or a generating set. Lawvere and Tierney showed that every Grothendieck topos is an elementary topos, so the two notions are compatible, but the elementary notion is much broader: it includes categories of sets varying over a parameter, categories of presheaves on small categories, and categories built from realizability. The subobject classifier Ω is the key innovation: it internalizes the logic of the topos6. The arrows Ω $\rightarrow $ Ω form a Heyting algebra, so the internal logic of an elementary topos is intuitionistic—excluded middle need not hold7. This was a striking result: a purely geometric-looking axiom system carries a non-classical logic.

The internal language and categorical logic

The discovery that a topos has an internal logic was developed through the 1970s into the internal language of a topos, also called the Mitchell–Bénabou language. The idea is that one can reason about a topos as if its objects were sets and its arrows were functions, using a typed higher-order intuitionistic logic, and any statement provable in that logic holds in the topos6. Michael Makkai and Gonzalo Reyes, in their 1977 book First Order Categorical Logic, gave a systematic treatment of the correspondence between theories and toposes. The internal language made topos theory a branch of categorical logic: a topos is a model of a theory, and geometric morphisms are the appropriate maps of models.

This line of work connected topos theory with the broader program of categorical logic: the Curry–Howard–Lambek correspondence between proofs, programs, and arrows; the semantics of intuitionistic type theory; and the study of realizability toposes, which model computability. The effective topos of Martin Hyland (1982) is a landmark: a topos whose internal logic captures computable mathematics, showing that topos theory could give a foundation for constructive and computable mathematics, not just classical geometry.

Synthetic differential geometry

A parallel development used toposes to give a rigorous foundation for infinitesimal calculus8. In the 1960s and 1970s, Lawvere and others explored the idea of a topos in which the real line carries nilpotent infinitesimals—elements ε with ε² = 0 but ε ≠ 0—so that derivatives can be defined algebraically. This is synthetic differential geometry (SDG)8. The problem was that no ordinary topos of sets can have such infinitesimals without collapsing, by a classical argument. The solution, developed by Lawvere, Anders Kock, and others in the 1970s and 1980s, is to work in a topos with intuitionistic internal logic, where the classical argument fails. Kock's book Synthetic Differential Geometry (1981, later editions) became the standard reference. SDG showed that topos theory could host a coherent alternative to the ε-δ foundations of analysis, and it remains an active area connecting topos theory with differential geometry and physics.

Higher toposes and the homotopical turn

The most recent major development is the theory of higher toposes, or ∞-toposes. In the 2000s and 2010s, building on André Joyal's and Jacob Lurie's work on quasicategories and ∞-categories, Charles Rezk, Lurie, and others developed a theory in which the category of sets is replaced by the ∞-category of spaces (homotopy types), and sheaves take values in spaces. An ∞-topos is an ∞-category that behaves like a category of sheaves of spaces on a site9. Lurie's Higher Topos Theory (2009) is the foundational text10. This theory unifies classical topos theory with homotopy theory and has become central to derived algebraic geometry and to the foundations of homotopy type theory, where the connection between type theory and ∞-groupoids is a major theme.

The present

Topos theory today is a meeting point of several traditions. Grothendieck toposes remain a tool in algebraic geometry and in the theory of motives. Elementary toposes and their internal languages are central to categorical logic and to the semantics of type theory, including the ongoing program of homotopy type theory. Synthetic differential geometry continues as a niche but coherent approach to analysis. Higher topos theory has become a standard language for derived geometry and for the homotopical foundations of mathematics. Open questions include the precise relationship between elementary and Grothendieck toposes in the higher setting, the classification of toposes with given properties, and the extent to which topos-theoretic methods can serve as a foundation for constructive and computable mathematics. The field's history is one of a geometric tool becoming a general theory of "spaces" and "universes," and the questions it leaves open are about how far that generality can be pushed.

References

  1. Topos (introduction, Introduction, Equivalent definitions, Examples), Wikipedia
  2. Elementary topos, Wikipedia
  3. Sheaf (mathematics), Wikipedia
  4. Grothendieck topology, Wikipedia
  5. Pointless topology, Wikipedia
  6. Subobject classifier, Wikipedia
  7. Heyting algebra, Wikipedia
  8. Synthetic differential geometry, Wikipedia
  9. ∞-topos, Wikipedia
  10. Higher Topos Theory (introduction, History), Wikipedia

The frameworks of Topos Theory

Each framework is a way of working the field has used: when it held, and the question it answers.

  1. 1958-Present

    Grothendieck Toposes and Sites

    When do local data glue into global data?

    Fly to its tower in Nousfarer

  2. 1969-Present

    Elementary topos

    What axioms make a category behave like sets?

    Fly to its tower in Nousfarer

  3. 1970-Present

    Internal Language and Categorical Logic

    How can one reason inside an abstract category?

    Fly to its tower in Nousfarer

  4. 1970-Present
  5. 2002-Present

    ∞-Topos

    What are sheaves when sets become spaces?

    Fly to its tower in Nousfarer

How the ideas connect

Each arrow runs from an earlier idea to the later one it shaped. The list below says what passed between them.

Grothendieck Toposes and SitesElementary toposInternal Language and Categorical Log…Synthetic Differential Geometry∞-Topos

Beyond the field

Where Topos Theory's ideas came from in other fields, and what they led to there: the strongest 12 of 24 links.

A city in Nousfarer: a red-carpeted stair climbing terraces to a crowned tower on an acropolis, more towers round it, forest and snowy mountains beyond.
A city in Nousfarer, the game played in the browser.

Topos Theory is a planet in Nousfarer

Its city has a tower for each of these 5 frameworks. Land, walk in, and climb one concept at a time.

Land on Topos Theory

Related fields

Fields its ideas are linked to

More in Category Theory

See where Topos Theory sits on the Map of Human Knowledge