adjoint functor

Adjunction is arguably the central concept of category theory: it makes precise the ubiquitous pattern of a “best approximation from one side.” Free constructions, forgetful maps, abelianization, Stone-Čech compactification, tensor-hom, and limits-and-colimits are all instances. The slogan is that a left adjoint is the most efficient way to add structure freely, and its right adjoint the most efficient way to forget structure — and the two are locked together by a single natural bijection of arrows.

Functors F : C -> D and G : D -> C are an adjoint pair, written F ⊣ G (F left adjoint, G right adjoint), if there is a bijection Hom_D(F(A), B) ≅ Hom_C(A, G(B)) natural in both A and B. Equivalently the data can be packaged as a unit natural transformation η : id_C => G ∘ F and a counit ε : F ∘ G => id_D satisfying the triangle identities. The hom-set bijection is the most memorable form: a map “out of the free thing” is the same as a map “into the forgotten thing.”

Two structural theorems make adjoints powerful. Left adjoints preserve all colimits and right adjoints preserve all limits — so a functor failing to preserve, say, products simply cannot be a right adjoint. And the adjoint functor theorem gives conditions (a solution set condition plus limit preservation) under which a limit-preserving functor actually has a left adjoint. Adjoints are unique up to natural isomorphism whenever they exist.