We start by introducing the basic vocabulary of synthetic category theory.
Axiom A.1 (Primitives). The following are the primitives of synthetic category theory:
(1) There are categories, which we denote by symbols like \(C\), \(D\), and \(E\). Among categories, there are animae.
(2) For categories \(C\) and \(D\), there is a primitive notion of a functor \(f\colon C\to D\). Every category has an identity functor \(\id_C\colon C\to C\). Given \(f\colon C\to D\) and \(g\colon D\to E\), there is a composite functor \(g\circ f\colon C\to E\), sometimes also denoted \(gf\).
(3) There exists an anima \(*\), called the terminal category.
Categories and functors play the foundational role in synthetic category theory that sets and functions play in naive set theory. The role of the animae among the categories is to encode the idea of a homotopy-coherent collection: where objects in classical mathematics assemble into sets, in synthetic category theory they assemble into animae. A category additionally allows objects to be related by morphisms that need not be invertible.
For now, animae form a primitive distinguished class of categories. We regard every category equivalent to an anima as an anima. Once we have introduced morphisms and their invertibility, we can define groupoids as those categories in which every morphism is invertible. In the next chapter, we will impose the axiom that the animae are precisely the groupoids.
Definition 1.1.1 (Terms). A functor \(f\colon C\to D\) may be displayed in term notation as \[ x:C\vdash f(x):D. \] The functor and the expression \(f(x)\) are two notations for the same datum. We call \(x:C\) its context. More generally, we may write \(C\vdash t:D\) when the name of the variable is left implicit.
Throughout the chapter, we will indicate how the various rules we introduce may be written in their corresponding term notation. Conversely, any term expression obtained by combining these rules corresponds to a well-defined functor. The notation is inspired by type-theoretic approaches to synthetic category theory, such as [RS17, GWB24].
For example, composition of functors is displayed by substitution: for a second functor \(g\colon D\to E\), the term notation for the composite \(g\circ f\) is the expression \(x :C \vdash g(f(x)):E\). The term notation for the identity functor on \(C\) is \(x : C \vdash x : C\). We also refer to the latter as the universal term of \(C\).
Definition 1.1.2 (Objects). Given a term \(x:C\) in context \(\Gamma\), we say that \(x\) is an object (in context \(\Gamma)\) if \(\Gamma\) is an anima. If we speak about an object of \(C\), we mean an object in an arbitrary anima context.
An absolute object is a functor \(x\colon * \to C\), i.e. an object in context \(*\).
As already mentioned in the introduction of this book, one major difference between classical mathematics and homotopy theory is the idea that equality is structure. In classical mathematics, equality is a yes-no-question: two elements of a set are either equal or not. In homotopy theory, we change our perspective and allow there to be potentially several different identifications. The collection of identifications thus needs to assemble into an anima.
We thus introduce a primitive anima of identifications between every pair of functors. All our definitions and constructions should always respect identifications; we choose to implement this via explicit rules rather than via some deeper foundational principle as is for example done in homotopy type theory.
Axiom A.2 (Identifications). Given parallel functors \(f,g\colon C\to D\), there is an anima \((f= g)\), called the anima of identifications of \(f\) and \(g\). An identification \(\alpha: f = g\) is an absolute object of this anima.
For \(h\colon D\to E\) and \(k\colon B\to C\), there are functors \[ h\star-\colon(f= g)\to(hf= hg), \qquad -\star k\colon(f= g)\to(fk= gk), \] called postwhiskering and prewhiskering, respectively.
Notation 1.1.3. We also denote identifications \(\alpha : f = g\) in term notation asThere is a slight clash between this notation and that of Definition 1.1.1: this does not mean a functor \(C \to f(x) = g(x)\), which does not even make sense. \[ x : C \vdash \alpha(x) : f(x) = g(x). \]
Given an anima \(A\) and two absolute objects \(a,b:A\), we obtain an anima \(a = b\) of identifications between them. In particular, given two identifications \(\alpha,\beta: f = g\), we may form the anima \(\alpha = \beta\). This construction may be iterated.
Definition 1.1.4. A commutative square of functors, often displayed by means of a diagram
consists of four functors \(f\), \(g\), \(h\) and \(f'\) with sources and targets as displayed in the diagram, together with a specified identification \(\alpha\colon h \circ f = f' \circ g\).
We will often refer to the data of just the four functors \(f\), \(g\), \(h\) and \(f'\) as a square, and say that the square commutes whenever an identification \(\alpha\colon h \circ f = f' \circ g\) can be provided.
A commutative triangle of functors, displayed by means of a diagram
consists of three functors \(f\), \(g\) and \(h\) with sources and targets as displayed in the diagram, together with a specified identification \(\alpha\colon h = g \circ f\).
We will similarly display more general diagrams of functors by drawing directed planar graphs with vertices labeled by categories and edges labeled by functors. We consider diagrams whose faces are bounded by two directed paths with the same source and target, and fix a parenthesization for each composite. We say the diagram commutes if every face comes equipped with an identification between these two composites.
Definition 1.1.5 (Equivalence of categories). A functor \(f\colon C \to D\) is said to be an equivalence if there exists another functor \(g\colon D \to C\) equipped with identifications \(\id_{C} = gf\) and \(\id_{D} = fg\). We refer to \(g\) as the inverse of \(f\). Equivalences will often be introduced by writing \(f\colon C \iso D\). We may write \(C \simeq D\) to indicate that \(C\) and \(D\) are equivalent via an unnamed equivalence.
The next two axioms belong to the universal constructions of group B. We introduce them here because they are needed to formulate the coherence axioms of group A. We begin with the universal property of the terminal category.
Axiom B.1 (Terminal category). For every category \(T\), there is a functor \(p_T\colon T\to *\). For any functors \(f,g\colon T\to *\), the functor \[ p_{(f= g)}\colon(f= g)\longrightarrow * \] is an equivalence.
In particular, any two functors to \(*\) are identified.
Definition 1.1.6 (Contractible category). A category \(C\) is called contractible if the functor \(p_{C}\colon C \to *\) is an equivalence.
Definition 1.1.7 (Constant functor). Let \(C\) and \(D\) be categories and let \(x\) be an absolute object of \(D\). We define the constant functor \(\const_x\colon C \to D\) as the composite \(x \circ p_C\colon C \to D\): \[ \const_x: C \xrightarrow{p_C} * \xrightarrow{x} D. \]
Remark 1.1.8. Observe that a category \(C\) is contractible if and only if there exists an absolute object \(x\colon * \to C\) equipped with an identification \(\id_C = \const_x\). Such an identification \(H\colon \id_C = \const_x\) is called a contraction of \(C\) onto \(x\).
We will postpone most category constructors until later in the chapter, but we need one crucial one right now already in order to state various coherence relations.
Axiom B.2 (Product of categories). The product of categories \(C\) and \(D\) is a category \(C\times D\) equipped with projection functors \(\pr_1\colon C\times D\to C\) and \(\pr_2\colon C\times D\to D\).
Given \(f\colon T\to C\) and \(g\colon T\to D\), there is a functor \((f,g)\colon T\to C\times D\) equipped with identifications \(\beta^1_{f,g}:\pr_1(f,g)=f\) and \(\beta^2_{f,g}:\pr_2(f,g)=g\). For any two functors \(h,k\colon T\to C\times D\), the functor \[ (\pr_1\star-,\pr_2\star-)\colon (h= k)\longrightarrow (\pr_1 h=\pr_1 k)\times(\pr_2 h=\pr_2 k) \] is an equivalence. If \(C\) and \(D\) are animae, then \(C\times D\) is an anima.
We refer to the functor \((f,g)\) as the pair construction. In term notation, it is written \[ x:T \vdash (f(x),g(x)): C \times D, \] with identifications \(\pr_1(f(x),g(x))= f(x)\) and \(\pr_2(f(x),g(x))= g(x)\). For arbitrary expressions \(h(x),k(x):C\times D\), two identifications \(\alpha(x):\pr_1 h(x)=\pr_1 k(x)\) and \(\beta(x):\pr_2 h(x)=\pr_2 k(x)\) give an identification \((\alpha(x),\beta(x)):h(x)= k(x)\) by applying an inverse of the displayed equivalence. Its two projections are identified with \(\alpha\) and \(\beta\).
Remark 1.1.9. Given categories \(C\) and \(D\), we define a span from \(C\) to \(D\) to be a triple \((T,f,g)\) consisting of a category \(T\) together with functors \(f\colon T \to C\) and \(g\colon T \to D\). Axiom B.2 may be rephrased as saying that the triple \((C \times D, \pr_1, \pr_2)\) is the terminal span from \(C\) to \(D\): every span \(C \leftarrow T \to D\) admits a factorization through \(C \times D\), and an identification of two functors \(T \to C \times D\) amounts to identifications between their respective components.
Convention 1.1.10. A term of the form \(z : C \times D \vdash t : E\) may also be written as \[ x:C, y:D \vdash t(x,y) : E \] whenever the term \(t\) only depends on the variable \(z\) through the two projections \(\pr_1(z)\) and \(\pr_2(z)\). The expression \(t(x,y)\) is then obtained from \(t\) by replacing \(\pr_1(z)\) by \(x\) and \(\pr_2(z)\) by \(y\). For more than two variables we use a fixed parenthesization of their product.