In this chapter, we introduce synthetic analogues of the notions of cartesian and cocartesian fibrations, which play an important foundational role in this book.
In order to appreciate the relevance of these notions within our theory, let us start by recalling their role in classical category theory. Cartesian and cocartesian fibrations come up whenever one is interested in functorial families of categories. To make precise what this means, assume we are given some category \(\Cat\) whose objects are small categories and whose morphisms are functors between them; we will introduce such a category in Chapter 8. For a category \(C\), we may then think of functors \(F\colon C \to \Cat\) as ‘\(C\)-indexed families of (small) categories’. Informally, they consist of a category \(F(c)\) for each object \(c\) of \(C\), a ‘transport functor’ \(F(u)\colon F(c) \to F(d)\) for each morphism \(u\colon c \to d\), and various compatibilities of these transport functors with composition in \(C\). In practice, these compatibilities are often not strict and need to be tracked; in classical language such functors are referred to as pseudofunctors. Constructing such functors by hand is often hard, and it becomes practically impossible in the setting of higher category theory.
Fortunately, there is an alternative way of packaging the data in such functors \(F\) via their Grothendieck construction \(\int F\). The objects of this category are given by pairs \((c,x)\), where \(c\) is an object of \(C\) and \(x\) is an object of \(F(c)\). A morphism from \((c,x)\) to \((d,y)\) consists of a morphism \(u\colon c \to d\) together with a morphism \[ F(u)(x) \to y \] in \(F(d)\). It comes equipped with a projection functor \(p_F\colon \int F \to C\) which remembers the object or morphism in \(C\). The individual categories \(F(c)\) for \(c \in C\) may be recovered from \(p_F\) by forming the fiber over \(c\). The transport functors \(F(u)\) can also be recovered from \(p_F\) as follows. Given a morphism \(u\colon c \to d\) and an object \(x \in F(c)\), there is a distinguished morphism over \(u\) of the form \[ \widetilde{u}\colon (c,x) \to (d,F(u)(x)), \] determined by the identity of \(F(u)(x)\). Moreover, every other morphism \((c,x) \to (d,y)\) over \(u\) factors uniquely through \(\widetilde{u}\), followed by a morphism living in the fiber over \(d\). More generally, a morphism from \((c,x)\) over a composite morphism \(vu\) in \(C\) factors uniquely through \(\widetilde{u}\), followed by a morphism over \(v\). This universal property characterizes all these distinguished morphisms \(\widetilde{u}\) purely in terms of \(p_F\); they are called the cocartesian morphisms. We may thus recover \(F(u)(x)\) from \(p_F\) by considering the unique cocartesian morphism over \(u\) in \(\int F\) starting at \((c,x)\) and passing to its second component. The compatibilities of the functor \(F\) with composition are recorded by \(p_F\) in an analogous manner.
The Grothendieck construction thus allows functors \(F\colon C \to \Cat\) to be packaged in terms of functors \(p\colon E \to C\). The functors arising from this construction are precisely the cocartesian fibrations, characterized by the property that every morphism \(u\colon c \to d\) in \(C\) admits cocartesian lifts in \(E\) starting at an arbitrary object \(x\) in the fiber over \(c\). In practice, functors into \(\Cat\) are often obtained by constructing a category \(E\) over \(C\) and checking it is a cocartesian fibration; since the latter is merely a property, this does not require any higher coherences and is therefore usually an easier task. One then obtains a functor \(C \to \Cat\) by applying the Grothendieck correspondence. The construction can be dualized, leading to a notion of cartesian fibrations encoding contravariantly functorial families of categories.
Cocartesian fibrations already play an important role in the classical treatments of higher category theory, but they take on an even more foundational role within our synthetic setup:
• We need them to introduce the category \(\Cat\) of small categories. Since our theory does not allow us to define categories by giving their objects and morphisms, we need to find a way to characterize \(\Cat\), and we will do so via the correspondence between functors \(C \to \Cat\) and cocartesian fibrations over \(C\).
• Moreover, they are essential for formulating the functoriality of universals property in Chapter 6. This crucial property is what makes precise the idea that various universal properties may be checked objectwise, and then automatically assemble into coherent constructions.
Since we do not yet have these tools at our disposal at this point, we need to be careful in how to precisely formulate the definition of a cocartesian fibration in our synthetic setup to guarantee that it really contains all the required higher coherences. We do so by considering a certain directed evaluation functor \[ \Ar(E) \to E \vectimes{p} C, \] which sends a morphism in \(E\) to its source object and its image morphism in \(C\). The target category records precisely the data for which we seek a cocartesian lift. We may then define \(p\) to be a cocartesian fibration by requiring this functor to admit a so-called left adjoint section: the section provides the lifts functorially, while the adjunction expresses their universal property.
We begin with directed pullbacks and left and right fibrations, followed by slice categories. We then develop adjunctions, including adjoint sections and relative adjunctions, before introducing cartesian and cocartesian fibrations. The remaining sections study the source and target fibrations, cocartesian functors, and locally cocartesian fibrations.
Sections
Directed pullbacks
7 numbered results
Left and right fibrations
16 numbered results
Slice categories
9 numbered results
Adjunctions
30 numbered results
Cartesian and cocartesian fibrations
19 numbered results
The source and target fibrations
4 numbered results
Cocartesian functors
11 numbered results
Locally cocartesian fibrations
9 numbered results
Exercises Chapter 4
2 numbered results