Definition 5.1. A correspondence (or -correspondence, for clarity) of gaunt -categories is an object of the overcategory . In other words, a -correspondence of gaunt -categories is a gaunt -category along with a functor .
5. Correspondences and the category [0MLT]
The category is cartesian closed, which means that for each strict -category , the endo-functor admits a right adjoint, . The strict -categories make enriched in itself, and thus it can be regarded as a large strict -category.
The category , as a full subobject of , is likewise a large strict -category. Moreover the -category is gaunt whenever is gaunt, making itself cartesian closed. In fact, if we replace by a skeleton, then it is a large gaunt -category, though we will not need this observation.
If is a strict -category (possibly large) then we may consider the slice category of gaunt -categories equipped with a functor to . This is naturally enriched in as well and hence also a (large) -category. If and are in , then the -enriched hom from to is given by the following pull-back:
Remark 5.2. By Lemma 3.5 is locally finitely presentable. It follows (see [1, Corollary 2.44, 2.47]) that each of the categories of -correspondence is also locally finitely presentable.
Given two -correspondences and , we may form the product correspondence . This is the product in the category of -correspondences.
The terminology and connection to the classical theory of correspondences is made more clear by introducing the following category.
Definition 5.3. Define , and for , define recursively as follows. The objects are triples consisting of two gaunt -categories and and a functor
(Here is the opposite obtained by reversing just the 1-morphisms of .) A morphism is a triple consisting of functors and and a natural transformation
Lemma 5.4. There is a natural equivalence of categories
Proof. The functor is simply the identity. We now define recursively. Suppose that , and assume that the equivalence has been defined.
Let us define the functor . For any object of , let be the triple , where and are the fibers of over and , respectively, and the functor is the composite , where the functor
is defined as follows:
- •
For any objects , let be the -category , equipped with the functor induced by .
- •
For any objects and of , the functor
is simply composition.
For any commutative triangle
of gaunt -categories, we define
where and are the restrictions of to the fibers, and is the composite , in which the natural transformation is the one whose components are given by the functor induced by .
We now construct a quasi-inverse to . Again, when , we let be the identity, and we proceed recursively. We assume and that the quasi-inverse to has been defined.
For any object , define a gaunt -category with object set and
The composition in is the obvious one, and it is clear that this defines a functor . We now apply this functor to the terminal object of , namely the triple . Since , it follows that factors through a functor .
It is now a simple matter to observe that is indeed quasi-inverse to . ∎
Remark 5.5. Unwinding the functor in the argument above, one finds that the product in the category may be written recursively in the following manner:
where is defined as the composite:
Lemma 5.6. The category of -correspondences is cartesian closed.
Proof. The claim is that for any -correspondence , the functor
admits a right adjoint .
Since is locally finitely presentable, it is cocomplete and admits a strong generator [1, Th. 1.20] and is co-wellpowered [1, Th. 1.58]. Thus by the special adjoint functor theorem [30, Sect. V.8] the functor
admits a right adjoint precisely if it commutes with colimits.
When , this follows from the fact that itself is cartesian closed.
For , suppose is cartesian closed. To prove that the category is cartesian closed, we require a description of colimits in terms of the equivalent category . For any small category and any diagram with
it is easy to see that the colimit is given by the triple where
and is the enriched left Kan extension of
along the diagonal
Now for any object , we wish to compare and . In light of our descriptions of products in , we see that the former is , and the latter is the colimit of the diagram that carries to
Note that, since is cartesian closed, one has
hence our description of colimits in exhibits the colimit of as , where is the enriched left Kan extension of
along the diagonal
Our induction hypothesis is that is cartesian closed; so this enriched left Kan extension can be identified with the composition of the enriched left Kan extension of
along the diagonal
But now this left Kan extension is simply the product of with the left Kan extension that defines . In other words, we have an isomorphism , whence the proof is complete. ∎
Original source: arXiv:1112.0040v6
Original source · 1112.0040v6