Category Theory

An abelian category is a category in which you can do linear algebra and homological algebra: you can add morphisms, take kernels and cokernels, form exact sequences, and every short exact sequence behaves like 0 → A → B → C → 0. The category of abelian groups Ab, the category of R-modules R-Mod, and the category of sheaves of abelian groups on a space are all abelian. Grothendieck introduced the axioms in 1957 to unify these examples and enable cohomology for sheaves.

The derived category D(A) of an abelian category A is constructed by formally inverting quasi-isomorphisms (chain maps inducing isomorphisms on all cohomology groups). Derived categories provide the natural setting for homological algebra: derived functors (Ext, Tor, RHom, L\otimes) become honest functors between derived categories rather than sequences of cohomology groups. They are indispensable in modern algebraic geometry, representation theory, and mathematical physics.

An adjunction is one of the most fundamental and pervasive concepts in mathematics. A functor \(F: \mathcal{C} \to \mathcal{D}\) is left adjoint to \(G: \mathcal{D} \to \mathcal{C}\) (written \(F \dashv G\)) when there is a natural bijection \(\text{Hom}_{\mathcal{D}}(FC, D) \cong \text{Hom}_{\mathcal{C}}(C, GD)\). Adjunctions capture the notion of 'optimal solution to a problem' — the left adjoint builds the freest or most economical object, while the right adjoint extracts underlying structure. Free groups, tensor products, and the compactification of a space are all left adjoints; forgetful functors are typically right adjoints.

A category is a mathematical structure consisting of objects and morphisms (arrows) between them, satisfying identity and associativity laws. Category theory — sometimes called 'abstract nonsense' — provides a universal language that unifies disparate branches of mathematics by focusing on structure-preserving maps rather than the elements of individual structures. Morphisms generalize functions, group homomorphisms, continuous maps, and linear maps under one roof.

In an ordinary category, the hom-sets Hom(A, B) are just sets. In an enriched category, each hom-set is replaced by an object in some monoidal category V — called the enriching category. When V = Ab, you get additive categories (hom-sets are abelian groups). When V = [0,∞], you get generalised metric spaces. When V = Cat, you get strict 2-categories. Enriched category theory provides a unified framework for all these structures.

A functor is a structure-preserving map between categories. Just as a group homomorphism sends elements to elements while preserving the group operation, a functor sends objects to objects and morphisms to morphisms while preserving composition and identities. Functors are the morphisms of the category \(\text{Cat}\) of all small categories, making category theory itself categorical. Concrete examples appear everywhere: the fundamental group \(\pi_1\), homology groups \(H_n\), and forgetful functors from algebraic to set-theoretic categories are all functors.

Limits and colimits are universal constructions that unify many familiar mathematical objects: products, coproducts, pullbacks, pushouts, equalizers, coequalizers, kernels, and direct/inverse limits all arise as special cases. A limit is the 'most economical' way to map into a diagram; a colimit is the 'most economical' way to map out of one. The duality between them — obtained by reversing all arrows — is one of the deepest symmetries in mathematics.

A monad is a triple (T, η, μ) consisting of an endofunctor T : C → C and two natural transformations — the unit η : Id ⇒ T and the multiplication μ : T² ⇒ T — satisfying associativity and unit laws. Monads unify constructions as diverse as free algebras, powerset, probability distributions, and computational effects (in Haskell). Every adjunction gives rise to a monad, and conversely every monad arises from an adjunction (via the Kleisli or Eilenberg–Moore construction).

A natural transformation is a morphism between functors. If functors are 'translations' between mathematical universes, natural transformations are 'consistent' ways of converting one translation into another — with the consistency condition expressed by a commutative square for every morphism in the source category. Natural transformations were historically the motivation for inventing categories and functors in the first place: Eilenberg and Mac Lane introduced the entire categorical apparatus just to make precise what it means for two constructions to be 'naturally' (rather than arbitrarily) isomorphic.

Given any category C, its opposite category C^{op} is formed by reversing all the arrows: the objects are the same but every morphism f: A → B becomes a morphism f^{op}: B → A in C^{op}. This simple construction is the formal basis of the duality principle: any theorem proved in the language of category theory has a dual theorem obtained by replacing every concept by its categorical dual (limits ↔ colimits, monomorphisms ↔ epimorphisms, initial ↔ terminal objects).

A topos is a category that behaves like a universe of sets — but generalised far beyond the category Set. Every topos has finite limits, exponential objects (internal function spaces), and crucially a subobject classifier Ω that plays the role of the two-element set {true, false}. Toposes arise as sheaf categories on a site (Grothendieck toposes, central to algebraic geometry) and as categories satisfying pure categorical axioms (elementary toposes, supporting an internal intuitionistic logic).

A universal property is a characterisation of a mathematical object solely by the maps into or out of it, up to unique isomorphism. Rather than constructing an object and then proving properties, one specifies what the object must do — a 'best solution' to a mapping problem — and deduces the construction. Products, coproducts, free groups, tensor products, and completions all arise as solutions to universal problems. The Yoneda lemma provides the rigorous foundation: an object is determined (up to unique isomorphism) by the functor it represents.

The Yoneda lemma is arguably the single most important result in category theory. It states that a functor \(F: \mathcal{C} \to \text{Set}\) is completely determined by its behavior on a single representable functor, and that natural transformations from \(\text{Hom}(A, -)\) to \(F\) are in natural bijection with elements of \(F(A)\). As a corollary, the Yoneda embedding \(\mathcal{C} \hookrightarrow [\mathcal{C}^{\text{op}}, \text{Set}]\) is fully faithful, meaning every category embeds into a presheaf category — one of the most powerful tools in modern mathematics.

Categorical logic establishes a deep correspondence between logical systems and categorical structures. Under this correspondence, a theory (in a suitable logic) corresponds to a category, and models of the theory correspond to functors out of that category. Toposes, for instance, are categories that model intuitionistic higher-order logic. This framework unifies proof theory, model theory, and category theory, and underpins the Curry-Howard-Lambek correspondence between proofs, programs, and morphisms.

Higher category theory extends ordinary category theory by allowing morphisms between morphisms (2-morphisms), morphisms between those (3-morphisms), and so on. An n-category has morphisms at every level up to n (or all levels, in the case of \infty-categories). These structures arise naturally in homotopy theory, mathematical physics (topological field theories), and the study of algebraic structures with coherence conditions. The central challenge is correctly formulating the coherence conditions that govern the associativity and unit laws at each level.

An \infty-category (or (\infty,1)-category) is a higher-categorical structure where there are morphisms at all levels, but morphisms above level 1 are all invertible (i.e., they encode homotopies, homotopies between homotopies, etc.). The most tractable model is that of quasi-categories (Joyal, Lurie): simplicial sets satisfying a weak inner horn-filling condition. \infty-categories provide the correct framework for homotopy-coherent mathematics: derived algebraic geometry, topological field theories, and stable homotopy theory.

Sheaf theory provides a systematic framework for tracking local data on a topological space and assembling it into global information. A sheaf assigns algebraic objects (sets, groups, rings, modules) to open sets of a space, subject to a locality condition (sections determined locally) and a gluing condition (compatible local sections glue uniquely). Sheaves are fundamental in algebraic geometry, topology, and mathematical physics.