Basic Data. We assume that is a presentable -category equipped with a fully faithful functor .
7. The Unicity Theorem[0MLV]
We now introduce our axioms for the theory of -categories.
The first axiom states that every object of can be written as a colimit of gaunt -categories in a canonical manner; this is called strong generation.
Axiom (C.1: Strong generation). The functor is dense, or, equivalently in the language of [29, 4.4.2], it strongly generates . That is, the left Kan extension of along itself is the identity functor on .
Equivalently, any object is the canonical colimit of the gaunt -categories mapping to it:
This is equivalent to the condition that the functor induces an localization ; that is, can be written as for some class of maps of small generation ([27, Remark 20.4.1.5], [28, Proposition 5.5.4.16]).
The next axiom states that every object of can be written as a colimit of objects of , but not necessarily in this canonical manner.
Axiom (C.2: Weak generation). If is a full subcategory that contains the image and is closed under colimits, then .
Remark 7.1. In any -category which satisfies Axiom C.2 the cells detect equivalences. That is is an equivalence in if and only if it induces equivalences for all . This is clear, since for such a map the full subcategory of those such that is an equivalence is stable under colimits and contains the cells, and is thus all of .
We also demand that admit internal Homs for correspondences.
Axiom (C.3: Internal Homs for correspondences). For any morphism of , the fiber product functor
preserves colimits.
The Adjoint Functor Theorem [28, Corollary 5.5.2.9] implies that this is equivalent to the existence of internal homs for the categories of correspondences .
Equivalently, Axiom C.3 states that every morphism to is, in the language of AyalaβFrancis [3], an exponentiable fibration. (We are grateful to the referee for suggesting this observation.)
We introduce a special collection of maps of . Denote consist of the union of the following four finite sets of maps of :
(when , we interpret this as the initial object of mapping to the image under of the empty -category),
and, lastly,
Axiom (C.4: Fundamental pushouts). Each of the finite number of maps comprising is an equivalence.
Finally, we will require that be versal with Axioms C.1β4, so that any other presentable -category satisfying Axioms C.1β4 admits some left adjoint from :
Axiom (C.5: Versality). For any -category and any fully faithful functor satisfying Axioms C.1β4, there exist a left adjoint and a natural transformation such that is an equivalence.
Definition 7.2. A pair satisfying these axioms (C.1β5) will be said to be a theory of -categories. We define a subcategory of the -category of -categories: the objects are -categories that underlie a theory of -categories, and the morphisms are equivalences of these -categories.
The -category is an -groupoid and thus a homotopy type; it is the moduli space of theories of -categories. Our main theorem is a computation of this homotopy type.
Theorem 7.3 (Unicity). One has
The proof will occupy the next few sections, and is organized as follows.
- (a)
In SectionΒ 8 we introduce a particular -category and verify that it satisfies Axioms C.1-5. This shows that is non-empty, and we denote our chosen model by . For Axioms C.1-4 this means simply varifying the corresponding properties of our model. For Axiom C.5, Versality, we suppose another -category satisfying Axioms C.1-4 and must then show how those axioms guerentee the existence of the desired functor and transformation out of . In fact a stronger versality property holds: need only satisfy C.1, C.3, and C.4.
- (b)
In SectionΒ 9 we show that the space is connected, i.e., any two -categories satisfying C.1-5 are equivalent. The proof relies essentially on our -categories satisfying Axioms C.1, C.2, and C.5. We note, however, that Axiom C.5 is quantified over the other four axioms, and so this step in fact relies on all five axioms.
- (c)
Having shown is non-empty and connected, it remains to compute its loopspace. This is done in SectionΒ 10 by directly computing the automorphism space of our prefered model.
Original source: arXiv:1112.0040v6
Original source Β· 1112.0040v6