A compact object behaves like a finite object with respect to filtered colimits: mapping out of it commutes with taking such colimits. This notion suffices for the applications in this book; we will not need the more general notion of \(\kappa \)-compactness.

Definition 23.3.1 (Compact object). Let \(C\) be an \(\infty \)-category with filtered colimits. An object \(X\in C\) is called compact if the functor \(\Hom _C(X,-)\colon C\to \An \) preserves filtered colimits. We write \(C^\omega \subseteq C\) for the full subcategory of compact objects.

Corollary 23.3.2. If \(C\) is a stable \(\infty \)-category with filtered colimits, then \(C^\omega \) is a thick subcategory of \(C\).

Definition 23.3.3. An \(\infty \)-category \(C\) is called compactly generated if it admits small colimits, the subcategory \(C^\omega \) is small, and every object of \(C\) is a filtered colimit of compact objects.

If \(C\) is stable, a small collection \(\Gg \) of compact objects is called a collection of compact generators if the smallest localizing subcategory of \(C\) containing \(\Gg \) is \(C\) itself. Here a localizing subcategory is a full stable subcategory closed under small colimits.

Proposition 23.3.4 (Compact generation criterion, cf.ย Hovey et al. (1997), Corollary 2.3.12). Let \(C\) be a stable \(\infty \)-category with small colimits and let \(\Gg \) be a small collection of compact objects. The following conditions are equivalent:

(1)

The collection \(\Gg \) generates \(C\).

(2)

The functors \[ \Hom _C(G[n],-)\colon C\to \An , \qquad G\in \Gg ,\ n\in \Z , \] are jointly conservative.

(3)

Every object of \(C\) is a filtered colimit of objects in the smallest thick subcategory of \(C\) containing \(\Gg \).

If these conditions hold, then \(C\) is compactly generated and its compact objects form the smallest thick subcategory of \(C\) containing \(\Gg \).

Proof. Apply the argument of the proof of [Lurie (2017), Corollary 1.4.4.2] to the full subcategory generated by \(\Gg \) under shifts and finite colimits. The description of compact objects then follows from [Lurie (2009b), Lemma 5.4.2.4]. โ–ก

The preceding statement is often the most convenient way to prove compact generation. Notice the distinction between the two closure operations: the generators produce all objects under small colimits, but the compact objects are obtained from the generators using only finite colimits, shifts, and retracts.

Definition 23.3.5 (Ind-completion). Let \(C\) be a small \(\infty \)-category. Its Ind-completion is the full subcategory \[ \Ind (C)\subseteq \PSh (C) \] spanned by the filtered colimits of representable presheaves. The Yoneda embedding factors through a fully faithful functor \(C\hookrightarrow \Ind (C)\).

Thus \(\Ind (C)\) is the partial free cocompletion from Proposition 23.1.2 obtained by freely adjoining filtered colimits. In particular, if \(D\) admits filtered colimits, restriction along the Yoneda embedding induces an equivalence \[ \Fun _\omega (\Ind (C),D)\iso \Fun (C,D), \] where the left-hand side denotes the full subcategory of functors preserving filtered colimits.

Proposition 23.3.6 ([Lurie (2009b), Corollary 4.2.3.11, Propositions 5.3.2.9, 5.3.5.10 and 5.3.5.14, and Theorem 5.5.1.1]). If \(C\) is a small \(\infty \)-category with finite colimits, then \(\Ind (C)\) admits small colimits and is presentable. Moreover, the inclusion \(C\hookrightarrow \Ind (C)\) preserves finite colimits, and the inclusion \[ \Ind (C)\hookrightarrow \PSh (C) \] identifies \(\Ind (C)\) with the full subcategory spanned by the presheaves \(C\catop \to \An \) which send finite colimits in \(C\) to limits in \(\An \).

Proposition 23.3.7 (Compactly generated \(\infty \)-categories as Ind-completions, [Lurie (2009b), Propositions 5.5.7.8 and 5.5.7.10]). Let \(D\) be an \(\infty \)-category with small colimits and assume that \(D^\omega \) is small. The following are equivalent:

(1)

Every object of \(D\) is a filtered colimit of compact objects.

(2)

The restricted Yoneda functor induces an equivalence \[ D\iso \Ind (D^\omega ). \]

Under these conditions, a colimit-preserving functor out of \(D\) is uniquely determined by its restriction to \(D^\omega \).

Remark 23.3.8 (Compact presentations). A presentable \(\infty \)-category \(D\) is compactly generated if and only if it admits a presentation \[ D\simeq \PSh (C)^{S\text {-}\mathrm {loc}} \] in which \(C\) is small and \(S\) is a small collection of morphisms between compact objects of \(\PSh (C)\). We will not use this characterization; see [Lurie (2009b), Corollary 5.5.7.3 and Proposition 5.5.7.8].

Lemma 23.3.9 (Animae as Ind-completions). The restricted Yoneda functors induce equivalences \[ \An \iso \Ind (\An ^{\fin }) \qquad \text {and}\qquad \An _*\iso \Ind (\An _*^{\fin }). \]

Proof. The first equivalence follows immediately from Proposition 4.3.11, Proposition 23.3.6. For the second, Lemma 4.1.13 gives \[ \Fun ^{\lex }((\An _*^{\fin })\catop ,\An ) \iso \Fun ^{\lex }((\An _*^{\fin })\catop ,\An _*), \] and the right-hand side is equivalent to \(\An _*\) by Lemma 4.3.12. In both cases the inverse to evaluation is the restricted Yoneda functor, so the result follows from Proposition 23.3.6. โ–ก

Example 23.3.10. The \(\infty \)-categories \(\An \), \(\An _*\), and \(\Sp \) are compactly generated. For animae and pointed animae this follows from Lemma 23.3.9. The sphere spectrum is a compact generator of \(\Sp \): compactness follows because \(\Hom _{\Sp }(\S ,-)\simeq \Omega ^\infty \) preserves filtered colimits by Lemma 4.3.21, while its shifts jointly detect isomorphisms by Corollary 4.4.27; now apply Proposition 23.3.4.

For every associative ring spectrum \(A\), the free module \(A\) is a compact generator of \(\LMod _A\) by Lemma 8.1.11; for every commutative ring \(R\), the complex \(R[0]\) is a compact generator of \(\D (R)\) by Corollary 21.2.6. Consequently, \[ \Sp \simeq \Ind (\Sp ^\omega ), \qquad \LMod _A\simeq \Ind (\Perf (A)), \qquad \D (R)\simeq \Ind (\Perf (R)). \] The equivalence \(\D (R)\simeq \Mod _R(\Sp )\) of Corollary 8.2.4 identifies the last presentation with the corresponding presentation of \(\Mod _R(\Sp )\).

Exercises

Exercise 23.1 (Representables as compact generators). Let \(C\) be a small \(\infty \)-category.

(1)

Show that every representable presheaf \(Y(c)\in \PSh (C)\) is compact.

(2)

Show that the representables jointly detect isomorphisms.

(3)

Explain how the density presentation of a presheaf exhibits \(\PSh (C)\) as generated under colimits by the representables.

Exercise 23.2 (A presheaf localization). Let \(C\) be small and let \(S\) be a small collection of morphisms in \(C\). Regard each \(s\colon c\to d\) as a morphism \(Y(c)\to Y(d)\) of representable presheaves.

(1)

Show that a presheaf \(F\) is local with respect to these maps if and only if \(F(s)\colon F(d)\to F(c)\) is an equivalence for every \(s\in S\).

(2)

Identify the local presheaves with presheaves on the localization \(C[S^{-1}]\).

(3)

Conclude that this full subcategory is presentable.

Exercise 23.3 (An Ind-completion). Show that every set is a filtered colimit of its finite subsets and that finite sets are compact in \(\Set \). Prove that \[ \Ind (\Fin )\simeq \Set . \] Explain why the analogous statement for animae requires finite animae rather than finite sets.