nLab continuous functor



Category theory

Limits and colimits



In ordinary (Set-based) category theory, a functor F:CDF \colon C \to D is continuous if it preserves all small limits that exist in its domain CC, i.e. given any small category diagram A:ECA \colon E \to C in CC for which limA\lim A exists, with universal cone limAA\lim A \to A, application of FF to that cone results in a universal cone F(limA)FAF(\lim A) \to F \circ A, thus making F(limA)F(\lim A) the limit of FAF \circ A.

Put slightly differently, the limit lim(FA)\lim (F \circ A) exists and the natural map induced by the universal property of the limit

F(limA)lim(FA) F(\lim A) \longrightarrow \lim (F\circ A)

is an isomorphism. In many and perhaps most cases, one refers to continuous functors F:CDF \colon C \to D in cases where one already knows CC and DD are complete categories, i.e., where limA\lim A and limFA\lim F \circ A exist as a matter of course.

In such cases, all limits can be obtained from (small) products and binary equalizers (see here), and so it follows that a functor from a complete category is continuous if and only if it preserves all products and all binary equalizers.



The archetypical example is the Set-valued hom-functor: its continuity in both arguments (cf. hom-functors preserve limits) is indeed equivalent to the very definition of limit: for F:D opCF \colon D^{op} \to C a diagram and cCc \in C, the covariant hom-functor Hom C(c,):CSetHom_C(c,-) : C \to Set satisfies by definition of limit

Hom C(c,limF)limHom(c,F()). Hom_C(c, \lim F) \simeq \lim Hom(c,F(-)) \,.

Likewise for contravariant representable functors on CC, i.e., functors of the form C(,c):C opSetC(-, c): C^{op} \to Set, which take limits in C opC^{op} to limits in SetSet, or colimits in CC to limits in SetSet.


Every right adjoint functor is continuous, since right adjoints preserve limits.

In enriched category theory

One can also formulate limit preservation in terms of representable functors. This is especially appropriate in enriched category theory, where ordinary limits in terms of cones may no longer suffice and one should use weighted limits instead. Thus, working over a base of enrichment VV, assumed to be a Bénabou cosmos for convenience, suppose given a ( V V -enriched) functor F:CDF \colon C \to D, a small weight W:JVW\colon J \to V, and a diagram A:JCA \colon J \to C over that weight. Assume the weighted limit {W,A}\{W, A\} exists in CC, i.e., suppose we have a representing object for the VV-functor cV J(W,C(c,A()))c \mapsto V^{J}(W, C(c, A(-))), giving an isomorphism

C(c,{W,A})V J(W,C(c,A()))C(c, \{W, A\}) \cong V^{J}(W, C(c, A(-)))

natural in cc. In this situation, playing the role analogous to a universal cone is a universal VV-natural transformation

WC({W,A},A())W \to C(\{W, A\}, A(-))

and we say F:CDF \colon C \to D preserves the weighted limit if the composite

WC({W,A},A())D(F{W,A},(FA)())W \to C(\{W, A\}, A(-)) \to D(F\{W, A\}, (F \circ A)(-))

induces, à la Yoneda, a VV-natural isomorphism

D(d,F{W,A})V J(W,D(d,(FA)())).D(d, F\{W, A\}) \cong V^J(W, D(d, (F \circ A)(-))).

Then we say that F:CDF \colon C \to D is VV-continuous if it preserves all weighted limits that exist in CC.

Relation to other concepts


  1. It is not enough to demand that there exists an abstract isomorphism F(limA)lim(FA)F(\lim A) \cong \lim (F\circ A).

  2. Topologists sometimes use “continuous functor” to mean a Top-enriched functor, since a functor between topologically enriched categories is enriched iff its actions on hom-spaces are continuous functions.

  3. Sheaf-theorists sometimes say “continuous functor” for a cover-preserving functor between sites, with the intuition being that it generalizes the inverse image induced by a continuous function of topological spaces.

  4. H. Bass in his treatment of K-theory uses the older term ‘right continuous functor’ for the dual notion of cocontinuous functor in a version which is additive. If the domain of an additive functor which commutes with direct sums is a cocomplete category, then the functor automatically has right adjoint. Following this fact, some people in ring theory and noncommutative geometry use the simple term ‘continuous functor’ for a functor with a right adjoint (even if the domain abelian category is not cocomplete). In general, of course, this is just a bit more than cocontinuous in the standard sense.

basic properties of…


Last revised on April 11, 2024 at 15:58:14. See the history of this page for a list of all contributions to it.