The theory of \(\infty \)-categories admits most of the basic constructions familiar from classical category theory, like terminal and initial categories, products and coproducts of categories, pullbacks, functor categories, and pushouts. In classical category theory, one may give explicit constructions of these categories, and prove their universal properties β€˜by hand’ by writing down what the relevant functors should do on objects and morphisms. When working with \(\infty \)-categories, we need to reverse this logic: since functors can no longer be prescribed on objects and morphisms, we must always construct them by invoking universal properties. We will now introduce these universal properties for the basic constructions of \(\infty \)-categories. They always come in two parts: first we prescribe how to construct functors from suitable data, and then we pin down the anima of identifications between two such functors.

1.3.1 Products and coproducts

We start with products and coproducts of \(\infty \)-categories. For animae, the products, projections, and pairings below are those already introduced in Axiom A.1. Likewise, the empty product is the one-point anima \(*\), with its functors \(p_T\colon T\to *\). The next axiom specifies their universal properties and extends products to arbitrary families of \(\infty \)-categories. It also introduces coproducts.

Axiom C.1 (Products and coproducts of \(\infty \)-categories). Let \(I\) be a set and let \(C_i\) be an \(\infty \)-category for every \(i \in I\).

(1)

There is a product \(\prod _{i \in I} C_i\), equipped with projection functors \(\pr _i\colon \prod _{i \in I} C_i \to C_i\) for every \(i\). Given another \(\infty \)-category \(T\) and functors \(f_i\colon T \to C_i\) for every \(i\), we obtain a functor \((f_i)\colon T \to \prod _{i \in I} C_i\) together with identifications \(\pr _i \circ (f_i) = f_i\). For two functors \(g,h\colon T\to \prod _{i\in I}C_i\), the projection-induced functor \[ (g=h)\longrightarrow \prod _{i\in I}(\pr _i g=\pr _i h) \] is an equivalence. If every \(C_i\) is an anima, their product is an anima.

(2)

There is a coproduct or disjoint union \(\bigsqcup _{i \in I} C_i\), equipped with inclusion functors \(\incl _i\colon C_i \hookrightarrow \bigsqcup _{i \in I} C_i\) for every \(i\). Given another \(\infty \)-category \(T\) and functors \(f_i\colon C_i \to T\) for every \(i\), we obtain a functor \(\lra {f_i}\colon \bigsqcup _{i \in I} C_i \to T\) together with identifications \(\lra {f_i} \circ \incl _i = f_i\). For two functors \(g,h\colon \bigsqcup _{i\in I}C_i\to T\), restriction induces an equivalence \[ (g=h)\longrightarrow \prod _{i\in I}(g\incl _i=h\incl _i). \]

Notation 1.3.1. The empty product is the one-point anima \(*\), which we also call the terminal \(\infty \)-category. We denote the empty coproduct by \(\emptyset \) and call it the empty \(\infty \)-category. Thus every \(\infty \)-category \(C\) has canonical functors \[ C \to *, \qquad \qquad \emptyset \to C. \]

Remark 1.3.2. In the empty-index cases the displayed equivalences say that \((f=g)\) is contractible for any two functors \(f,g\colon T\to *\), and likewise for functors \(\emptyset \to T\). Thus any two such identifications are themselves identified.

Notation 1.3.3. When \(I = \{1,2\}\), we write \(C_1 \times C_2\) and \(C_1 \sqcup C_2\) for the resulting product and coproduct. Given functors \(f_1\colon T \to C_1\) and \(f_2\colon T \to C_2\), we write \((f_1,f_2)\colon T \to C_1 \times C_2\) for the induced functor. Given functors \(g_1\colon C_1 \to T\) and \(g_2\colon C_2 \to T\), we write \(\lra {g_1,g_2}\colon C_1 \sqcup C_2 \to T\) for the induced functor.

Taking \(T = *\), we see that objects of \(C_1 \times C_2\) have the form \((x,y)\), where \(x\) is an object of \(C_1\) and \(y\) is an object of \(C_2\).

Definition 1.3.4. Let \(C\) and \(D\) be \(\infty \)-categories and let \(x\) be an object of \(D\), regarded as a functor \(x\colon * \to D\). We define the constant functor \(\const _x\colon C \to D\) as the following composite: \[ \const _x\colon C \xrightarrow {p_C} * \xrightarrow {x} D. \]

Definition 1.3.5 (Contractible \(\infty \)-category). An \(\infty \)-category \(C\) is called contractible if the functor \(p_{C}\colon C \to *\) is an equivalence.

Observe that an \(\infty \)-category \(C\) is contractible if and only if there exists an object \(x\colon * \to C\) such that the composite \(\const _x\colon C \to C\) is identified with the identity \(\id _C\colon C \to C\). An identification \(H\colon \const _x = \id _C\) is called a contraction of \(C\) onto \(x\).

The familiar symmetries of products and coproducts follow from these universal properties just as in ordinary category theory, and we will use them freely.

Exercise 1.3.6. Show that (co)products of \(\infty \)-categories are associative, commutative and unital: \[ C \times D \iso D \times C, \qquad (C \times D) \times E \iso C \times (D \times E), \qquad C \times * \iso C \iso * \times C, \] and \[ C \sqcup D \iso D \sqcup C, \qquad (C \sqcup D) \sqcup E \iso C \sqcup (D \sqcup E), \qquad \emptyset \sqcup C \iso C \iso C \sqcup \emptyset . \]

Construction 1.3.7. The formation of products and coproducts of \(\infty \)-categories is functorial: given two functors \(f\colon C \to C'\) and \(g\colon D \to D'\), we obtain new functors \begin {align*} f \times g \; &:= \; (f \circ \pr _{C}, g \circ \pr _{D}) \colon \; C \times D \to C' \times D' \\ f \sqcup g \; &:= \; \lra {\incl _{C'} \circ f, \incl _{D'} \circ g} \; \colon C \sqcup D \to C' \sqcup D'. \end {align*}

We leave it to the reader to verify that there are identifications \begin {align*} \id _{C} \times \id _{D} = \id _{C \times D} &\quadtext { and } (f' \times g') \circ (f \times g) = (f' \circ f) \times (g' \circ g), \\ \id _{C} \sqcup \id _{D} = \id _{C \sqcup D} &\quadtext { and } (f' \sqcup g') \circ (f \sqcup g) = (f' \circ f) \sqcup (g' \circ g) \end {align*}

for functors \(f\colon C \to C'\), \(f'\colon C' \to C''\), \(g\colon D \to D'\) and \(g'\colon D' \to D''\).

1.3.2 Pullbacks of \(\infty \)-categories

We will now axiomatize the pullback of two functors \(f\colon C \to E\) and \(g\colon D \to E\) of \(\infty \)-categories. This will be similar to the axiom for the product \(C \times D\), but it is more involved due to the fact that we need to record compatibilities with the structure maps to \(E\).

Axiom C.2 (Pullbacks of \(\infty \)-categories). For functors \(f\colon C\to E\) and \(g\colon D\to E\), their pullback, or fiber product over \(E\), is an \(\infty \)-category \(C\times _E D\) with projections \(\pr _1\colon C\times _E D\to C\) and \(\pr _2\colon C\times _E D\to D\), and an identification \(\tau \colon f\circ \pr _1=g\circ \pr _2\); in other words, a commutative square of the form

Commutative square with maps C\times E D to C labeled \pr 1, D to E labeled g, C\times E D to D labeled \pr 2, C to E labeled f.

If \(C,D,E\) are animae, then \(C\times _E D\) is an anima. Given \(t_C\colon T\to C\), \(t_D\colon T\to D\), and \(\sigma \colon f\circ t_C=g\circ t_D\), there is a functor \(t=(t_C,t_D)_\sigma \colon T\to C\times _E D\) with identifications \(\alpha \colon \pr _1\circ t=t_C\), \(\beta \colon \pr _2\circ t=t_D\), and \[ \sigma \circ (f\star \alpha )=(g\star \beta )\circ (\tau \star t). \] For any \(h,k\colon T\to C\times _E D\), the functor \[ (h=k)\longrightarrow (\pr _1\circ h=\pr _1\circ k)\times _{(f\circ \pr _1\circ h=g\circ \pr _2\circ k)}(\pr _2\circ h=\pr _2\circ k) \] induced by whiskering with the two projections is an equivalence. The two functors defining the pullback on the right are \[ \alpha \longmapsto (\tau \star k)\circ (f\star \alpha ), \qquad \beta \longmapsto (g\star \beta )\circ (\tau \star h), \] and the identification needed to define the displayed functor is supplied by the interchange law.

Remark 1.3.8. More informally, the axiom says that functors into a pullback \(C \times _E D\) may be constructed by giving functors into \(C\) and \(D\) which agree after mapping to \(E\), and that an identification of two such functors \(h,k\colon T \to C \times _E D\) consists of identifications \(\alpha \colon \pr _1\circ h=\pr _1\circ k\) and \(\beta \colon \pr _2\circ h=\pr _2\circ k\), together with an identification \[ H\colon (\tau \star k)\circ (f\star \alpha )=(g\star \beta )\circ (\tau \star h). \]

Exercise 1.3.9. Show that the fiber product is commutative, associative and unital: for functors \(C\to E\), \(D \to E\) and \(B \to E\), there are equivalences \[ C \times _{E} D \iso D \times _{E} C, \qquad B \times _{E} (C \times _{E} D) \iso (B \times _{E} C) \times _{E} D, \qquad C \times _{E} E \iso C. \]

Exercise 1.3.10. Given \(\infty \)-categories \(C\) and \(D\), produce an equivalence \(C \times D \iso C \times _* D\).

Exercise 1.3.11. The construction of pullbacks of \(\infty \)-categories is functorial: given a commutative diagram

Commutative diagram involving C, E, D, C', E', D', with arrows labeled f, phi, chi, g, psi, f', g'.

there is an induced functor \[ \phi \times _{\chi } \psi := (\phi \circ \pr _{C}, \psi \circ \pr _{D})\colon C \times _{E} D \to C' \times _{E'} D'. \] Show that if each of the functors \(\phi \), \(\psi \) and \(\chi \) is an equivalence, then so is \(\phi \times _{\chi } \psi \). Deduce that equivalences are stable under pullback: if \(g\colon D \to E\) is an equivalence and \(f\colon C \to E\) is an arbitrary functor, then the projection \(C \times _{E} D \to C\) is an equivalence.

Pullbacks of \(\infty \)-categories lead to the notion of a fiber of a functor:

Definition 1.3.12. Let \(f\colon C \to D\) be a functor. For every object \(x\) of \(D\), we define the fiber of \(f\) over \(x\) as the pullback

Commutative square with maps C x to C, * to D labeled x, C x to *, C to D labeled f.

The definition of pullbacks of \(\infty \)-categories naturally leads to the notion of a pullback square.

Definition 1.3.13 (Pullback square). A commutative square

Commutative square with maps T to C labeled s, D to E labeled g, T to D labeled t, C to E labeled f.

is called a pullback square if the induced functor \((s,t)\colon T \to C \times _{E} D\) is an equivalence.

Given objects \(x,y\colon * \to C\) and an \(\infty \)-category \(T\), we expect identifications \(\const _x = \const _y\) between constant functors \(T \to C\) to correspond to functors from \(T\) to the identification anima \((x = y)\). The following axiom makes this precise.

Axiom C.3. For objects \(x,y\colon *\to C\), there is an identification \(\varepsilon \colon \const _x=\const _y\) of functors \((x=y)\to C\) exhibiting a pullback square

Commutative square with maps (x=y) to *, * to C labeled x, (x=y) to *, * to C labeled y.

For each identification \(\alpha \colon x=y\), its restriction along \(\alpha \colon *\to (x=y)\) is identified with \(\alpha \), using the identifications supplied by the terminal universal property. We write this as \(\varepsilon (\alpha )=\alpha \).

Exercise 1.3.14 (Iterated pullbacks). Consider a commutative diagram

Commutative diagram involving C 1, C 2, C 3, D 1, D 2, D 3, with arrows labeled g 1, f 1, g 2, f 2, f 3, h 1, h 2.

Construct an equivalence \[ D_1 \times _{D_3} C_3 \iso D_1 \times _{D_2} (D_2 \times _{D_3} C_3). \] Hint: define one direction by \((\pr _{D_1}, (h_1 \circ \pr _{D_1}, \pr _{C_3}))\) and construct an inverse by projecting to \(D_1\) and \(C_3\).

Lemma 1.3.15 (Pasting lemma for pullback squares). Consider a commutative diagram

Commutative diagram involving C 1, C 2, C 3, D 1, D 2, D 3, with arrows labeled g 1, f 1, g 2, f 2, f 3, h 1, h 2.

If the right-hand square is a pullback square, then the left-hand square is a pullback square if and only if the outer rectangle is a pullback square.

Proof. By the previous exercise there is a preferred equivalence \[ D_1 \times _{D_3} C_3 \iso D_1 \times _{D_2} (D_2 \times _{D_3} C_3). \] If the right-hand square is a pullback square, the functor \(C_2 \to D_2 \times _{D_3} C_3\) is an equivalence. Hence, by Exercise 1.3.11, the induced functor \(D_1 \times _{D_2} C_2 \to D_1 \times _{D_2} (D_2 \times _{D_3} C_3)\) is an equivalence. The claim now follows from 2-out-of-3 applied to the square

Commutative square with maps C 1 to D 1 \times D 2 C 2 labeled (f 1,g 1), D 1 \times D 3 C 3 to D 1 \times D 2 (D 2 \times D 3 C 3) labeled simeq, C 1 to D 1 \times D 3 C 3 labeled (f 1,g 2 circ g 1), D 1 \times D 2 C 2 to D 1 \times D 2 (D 2 \times D 3 C 3) labeled simeq.

β–‘

Exercise 1.3.16. Consider a commutative square

Commutative square with maps C to D labeled f, E to F labeled k, C to E labeled g, D to F labeled h.

and assume that \(h\) is an equivalence. Show that \(g\) is an equivalence if and only if the square is a pullback square.

We record one further property of coproducts.

Axiom C.4 (Universality of coproducts). Let \(I\) be a set, and let \(D_i\) be an \(\infty \)-category for every \(i \in I\).

(1)

Given functors \(f_i\colon C_i \to D_i\) for all \(i \in I\), the commutative square

Commutative square with maps C i to \bigsqcup i in I C i \bigsqcup i in I f i labeled \incl i, D i to \bigsqcup i in I D i labeled \incl i, C i to D i labeled f i, \bigsqcup i in I C i \bigsqcup i in I f i to \bigsqcup i in I D i.

is a pullback square.

(2)

For a functor \(h\colon E \to \bigsqcup _{i \in I} D_i\), the functor \[ \lra {\pr _E}_i \colon \bigsqcup _{i \in I}(E \times _{\bigsqcup _{i \in I} D_i} D_i) \to E \] is an equivalence.

(3)

The inclusions \(\{i\} \hookrightarrow I\) for \(i \in I\) induce an equivalence \(\bigsqcup _{i \in I} \{i\} \iso I\).

Remark 1.3.17. This is the categorical analogue of three basic facts about disjoint unions of sets: base change along an inclusion picks out the corresponding summand, every map into a disjoint union decomposes the source into the corresponding fibers, and every set is the disjoint union of its elements.

1.3.3 Functor categories

Functor categories let us package families of functors internally, and they will also allow us to define pushouts by duality in the next subsection.

Axiom C.5 (Functor category). The functor category between two \(\infty \)-categories \(C\) and \(D\) is an \(\infty \)-category denoted by \(\Fun (C,D)\). It comes equipped with a functor \(\ev \colon \Fun (C,D)\times C\to D\), called the evaluation functor. For another \(\infty \)-category \(E\) and a functor \(g\colon E\to \Fun (C,D)\), we may then define its uncurrying \(g^u\colon E\times C\to D\) as the composite \[ E\times C\xrightarrow {g\times \id _C}\Fun (C,D)\times C\xrightarrow {\ev }D. \] We demand that every functor \(f\colon E\times C\to D\) gives rise to a functor \(f_c\colon E\to \Fun (C,D)\), called the currying of \(f\), equipped with an identification \(f=(f_c)^u\). Moreover, for two functors \(g,h\colon E\to \Fun (C,D)\), we demand that uncurrying induces an equivalence \[ (g=h)\longrightarrow (g^u=h^u). \] Thus a specified identification between the uncurried functors lifts to one between \(g\) and \(h\). By Lemma 1.2.9, identifications between prescribed identifications lift as well. This allows us to translate back and forth between curried and uncurried functors.

In the case \(E=*\), objects \(*\to \Fun (C,D)\) of the functor category correspond to functors \(C\simeq *\times C\to D\) from \(C\) to \(D\), justifying the terminology for \(\Fun (C,D)\).

Exercise 1.3.18. Show that, for \(k\colon E'\to E\) and \(g\colon E\to \Fun (C,D)\), there is a preferred identification \((gk)^u=g^u(k\times \id _C)\), compatible with identifications of \(g\). Deduce the corresponding identification for currying. Hint: Expand uncurrying using evaluation and the identities expressing functoriality of products.

The functoriality of the product naturally provides functoriality for the functor categories:

Construction 1.3.19 (Functoriality of functor categories in \(C\) and \(D\)). Given a functor \(g\colon D \to E\), we define the functor \(g \circ -\colon \Fun (C,D) \to \Fun (C,E)\) as the currying of the composite \[ \Fun (C, D) \times C \xrightarrow {\ev } D \xrightarrow {g} E. \] Similarly, given a functor \(f\colon C \to D\), we define the functor \(- \circ f\colon \Fun (D,E) \to \Fun (C,E)\) as the currying of the composite \[ \Fun (D, E) \times C \xrightarrow {\id \times f} \Fun (D,E) \times D \xrightarrow {\ev } E. \] In a completely analogous way, every identification \(\beta \colon g = g'\) of functors \(D \to E\) induces an identification \((\beta \circ -) \colon (g \circ -) = (g' \circ -)\) of functors \(\Fun (C,D) \to \Fun (C, E)\), and similarly for the construction \(- \circ f\).

Alternative notations for these functors that we will frequently use are \(g_*\) and \(f^*\): \[ g_* := g \circ -\colon \Fun (C,D) \to \Fun (C,E), \qquad f^*:= - \circ f\colon \Fun (D,E) \to \Fun (C,D). \]

Exercise 1.3.20. Formulate and prove that the assignments \(g \mapsto (g \circ -)\) and \(f \mapsto (- \circ f)\) are functorial, in the sense that they respect identity functors and composition of functors.

Exercise 1.3.21. Given \(\infty \)-categories \(C\), \(D\) and \(E\), construct a composition functor \[ - \circ -\colon \Fun (D,E) \times \Fun (C,D) \to \Fun (C,E). \] Show that it reduces to the functors \(g \circ -\) and \(- \circ f\) when fixing one of the two variables.

Exercise 1.3.22. Let \(f,g\colon C\to D\) be functors, also regarded as objects of \(\Fun (C,D)\) by currying. Show that their primitive identification anima \((f=g)\) fits into a pullback square

Commutative square with maps (f=g) to Fun(C,D), * to Fun(C,D)timesFun(C,D) labeled (f,g), (f=g) to *, Fun(C,D) to Fun(C,D)timesFun(C,D) labeled Delta.

Hint: First use currying at \(*\) and Lemma 1.2.9 to compare the identification animae of the functors and their corresponding objects. Then apply Axiom C.3, and express the pullback of two objects as a fiber of the diagonal.

Proposition 1.3.23 (Internal universal properties). Let \(T\) be an \(\infty \)-category and let \(\{C_i\}_{i \in I}\) be a set-indexed collection of \(\infty \)-categories. The canonical functors below are equivalences: \begin {align*} \Fun (T,*) &\iso *, & \Fun (\emptyset ,T) &\iso *, & \Fun (*,T) &\iso T, \\ \Fun \left (T,\prod _{i \in I}C_i\right ) &\iso \prod _{i \in I}\Fun (T,C_i), &&& \Fun \left (\bigsqcup _{i \in I}C_i,T\right ) &\iso \prod _{i \in I}\Fun (C_i,T). \end {align*}

Moreover, for functors \(f\colon C \to E\) and \(g\colon D \to E\) there is an equivalence \[ \Fun (T,C \times _E D) \iso \Fun (T,C) \times _{\Fun (T,E)} \Fun (T,D), \] and for \(\infty \)-categories \(C\) and \(D\) there is an equivalence \[ \Fun (T,\Fun (C,D)) \iso \Fun (T \times C,D). \]

Proof sketch.

These properties are developed in Section 1.7.1 of [Cisinski et al. (2026)]; the proof for binary (co)products treated there adapts to one for arbitrary (co)products.

The functor \(\Fun (*,T) \to T\) is the evaluation map. In each other case, the displayed functor is constructed using the universal property of its target. For the inverse, we may map into the relevant functor category by first uncurrying and then using the universal property of the target. In the coproduct case, we use universality of coproducts to distribute the product with the parameter \(\infty \)-category over \(\bigsqcup _iC_i\). In the pullback case, we must show that the projection identifications agree over the base; this may be done after uncurrying, where the equivalence of identification animae in Axiom C.5 allows us to use the pullback constructor. β–‘

1.3.4 Pushouts of \(\infty \)-categories

Using functor categories, we may now dualize the notion of pullback square and define pushouts:

Definition 1.3.24 (Pushout square). A commutative square of \(\infty \)-categories

Commutative square with maps C' to C labeled u, D' to D labeled v, C' to D' labeled f', C to D labeled f.

is called a pushout square (or cocartesian) if, for every \(\infty \)-category \(E\), the induced square

Commutative square with maps Fun(D,E) to Fun(D',E) labeled v^*, Fun(C,E) to Fun(C',E) labeled u^*, Fun(D,E) to Fun(C,E) labeled f^*, Fun(D',E) to Fun(C',E) labeled (f')^*.

is a pullback square.

Axiom C.6 (Pushout of \(\infty \)-categories). For functors \(f\colon E \to C\) and \(g\colon E \to D\), there exists a pushout square

Commutative square with maps E to C labeled f, D to C \sqcup E D labeled \incl D, E to D labeled g, C to C \sqcup E D labeled \incl C.

We refer to the \(\infty \)-category \(C \sqcup _E D\) as the pushout of \(C\) and \(D\) along \(E\).

Lemma 1.3.25 (Pasting lemma for pushout squares). Consider a commutative diagram

Commutative diagram involving C 1, C 2, C 3, D 1, D 2, D 3, with arrows labeled g 1, f 1, g 2, f 2, f 3, h 1, h 2.

If the left-hand square is a pushout square, then the right-hand square is a pushout square if and only if the outer rectangle is a pushout square.

Proof. This follows immediately from Lemma 1.3.15. β–‘