Given an \(\infty \)-operad \(\Oo \) with finite operadic limits, the reduced excisive functors \(\An _*^{\fin }\to U(\Oo )\) form the underlying \(\infty \)-category of its stabilization, and its multimorphisms are supplied by Day convolution. To prove its stability, we embed \(\Oo \) into a presheaf category, using the multiplicative Yoneda embedding. The universal property will then follow by showing that evaluation at \(S^0\) recovers \(\Oo \) when it is already stable.
Construction 20.2.1. Let \(\Oo \) be an \(\infty \)-operad that admits finite operadic limits. Consider the \(\infty \)-category \(\An _*^{\fin }\) equipped with the smash product monoidal structure from Definition 16.3.5. We define \[ \oSp (\Oo ) \quad \subseteq \quad \oDay (\Mm _{\An _*^{\fin }}, \Oo ) \] to be the full suboperad whose colors are those functors \(F\colon \An _*^{\fin } \to U(\Oo )\) that are reduced and excisive. Note that by construction we have \[ U(\oSp (\Oo )) \simeq \Sp ^{\exc }(U(\Oo )), \] and so by Proposition 16.5.10 we may identify this underlying \(\infty \)-category with the abstract stabilization \(\Sp (U(\Oo ))\). The monoidal unit \(S^0\) defines an operad map \(\Comm \to \Mm _{\An _*^{\fin }}\), and restricting the evaluation map along it defines a morphism of \(\infty \)-operads \[ \Omega ^\infty \colon \oSp (\Oo ) \to \Oo . \]
Lemma 20.2.2. Let \(D\) be a small symmetric monoidal \(\infty \)-category and equip \(\PSh (D)\) with the Day convolution symmetric monoidal structure. Then \(\oSp (\Mm _{\PSh (D)})\) is a stable \(\infty \)-operad.
Proof. By construction of the Day convolution structure on \(\PSh (D)\), its multimorphism operad \(\Mm _{\PSh (D)}\) is identified with \(\oDay (\Mm _{D\catop },\Mm _{\An })\). Thus, using the universal property of Day convolution twice, \(\oDay (\Mm _{\An _*^{\fin }},\Mm _{\PSh (D)})\) identifies with \[ \oDay (\Mm _{D\catop },\oDay (\Mm _{\An _*^{\fin }},\Mm _{\An })). \] Under this identification, \(\oSp (\Mm _{\PSh (D)})\) is obtained by restricting the inner \(\oDay (\Mm _{\An _*^{\fin }},\Mm _{\An })\) to reduced excisive functors, using Lemma 19.6.1. By Theorem 16.6.1 and the construction of the symmetric monoidal localization there, this restriction is the multimorphism operad \(\Mm _{\Sp }\). Hence \[ \oSp (\Mm _{\PSh (D)}) \simeq \oDay (\Mm _{D\catop },\Mm _{\Sp }). \] The latter is the multimorphism operad of \(\Fun (D\catop ,\Sp )\) with Day convolution. This functor category is stable, with finite limits and colimits computed pointwise. Its Day tensor product preserves colimits separately in each variable, since it is obtained from the pointwise tensor product of spectra by a colimit-preserving left Kan extension. Thus its finite limits and colimits are operadic, proving the claim. □
Proposition 20.2.3. Let \(\Oo \) be an \(\infty \)-operad that admits finite operadic limits. Then \(\oSp (\Oo )\) is stable.
Proof. Put \(E:=\Env (\Oo )\). By Corollary 20.1.6, the multiplicative Yoneda embedding realizes \(\Oo \) as a full suboperad of \(\Mm _{\PSh (E)}\) and preserves finite operadic limits. It therefore induces a fully faithful functor \[ \Sp ^{\exc }(U(\Oo ))\longrightarrow \Sp ^{\exc }(\PSh (E)) \] preserving finite limits, since these are computed pointwise. Both \(\infty \)-categories are stable by Proposition 16.5.7, so this functor is exact and its essential image is closed under finite limits and colimits. By Lemma 19.6.1, the corresponding full suboperad of \(\oSp (\Mm _{\PSh (E)})\) is \(\oSp (\Oo )\). The ambient \(\infty \)-operad is stable by Lemma 20.2.2, so Lemma 20.1.9 proves the claim. □
For a stable \(\infty \)-operad \(\Oo \), evaluation at \(S^0\) already gives an equivalence on underlying categories (Lemma 16.5.8). We must show that it also induces equivalences on multimorphism animae. By Theorem 19.6.3, these consist of coherent families of operations \[ F_1(K_1),\dots ,F_n(K_n)\longrightarrow G(K_1\wedge \cdots \wedge K_n). \] We will show that such a family is determined by its value at \(K_1=\cdots =K_n=S^0\). The following lemma allows us to evaluate one variable at a time.
Lemma 20.2.4. Let \(T\colon (\An _*^{\fin })\catop \times \An _*^{\fin }\to \An \) be a functor such that \(T(-,L)\) sends finite colimits to limits for every \(L\), and \(T(K,-)\) is reduced and excisive for every \(K\). Then the projection \[ \int _{K \in \An _*^{\fin }} T(K,K) \to T(S^0,S^0), \] is an equivalence.
Proof. By Lemma 23.3.9, the condition on the first variable gives a functor \(F\colon \An _*^{\fin }\to \An _*\) and natural identifications \[ T(K,L)\simeq \Hom _{\An _*}(K,F(L)). \] Since the restricted Yoneda embedding preserves and reflects limits, the condition on the second variable says that \(F\) is reduced and excisive. Let \(j\colon \An _*^{\fin }\hookrightarrow \An _*\) be the inclusion. The end formula for natural transformations and Lemma 16.5.14 give \begin {align*} \int _{K\in \An _*^{\fin }}T(K,K) &\simeq \Nat (j,F)\simeq \Nat (P_1j,F)\\ &\simeq \Hom _{\An _*}(S^0,F(S^0)) \simeq T(S^0,S^0). \end {align*}
The composite sends a transformation \(j\to F\) to its component at \(S^0\), so it is the stated projection. □
Proposition 20.2.5. If \(\Oo \) is stable, then the evaluation map \[ \Omega ^\infty \colon \oSp (\Oo ) \to \Oo \] is an equivalence of \(\infty \)-operads.
Proof. On underlying \(\infty \)-categories this is Lemma 16.5.8. By Lemma 14.1.11, it remains to prove that \(\Omega ^\infty \) is fully faithful.
Let \(I\) be a finite set, and let \(\{F_i\colon \An _*^{\fin } \to U(\Oo )\}_{i \in I}\) and \(G\colon \An _*^{\fin } \to U(\Oo )\) be reduced excisive functors. By Theorem 19.6.3, the source of the induced map on multimorphism animae is \[ \int _{\{K_i\}_{i \in I}\in (\An _*^{\fin })^I} \Oo (\{F_i(K_i)\}_{i \in I};G(\bigwedge \nolimits _{i \in I}K_i)). \] By naturality of Theorem 19.6.3 in the evaluation operad map and the identification \(\bigwedge _{i\in I}S^0 \simeq S^0\), the map induced by \(\Omega ^\infty \) is the projection from this end to the component where all \(K_i\) are equal to \(S^0\).
We evaluate the variables at \(S^0\) one at a time. In order to iterate the argument, we allow additional fixed colors \(z_1,\dots ,z_m\) among the inputs and prove the assertion by induction on \(|I|\). The case \(I=\emptyset \) is immediate. For \(I\neq \emptyset \), choose \(i_0\in I\). By Fubini, the end is \(\int _K T(K,K)\) for the bifunctor \[ T(K,L):= \int _{\{K_i\}_{i\neq i_0}\in (\An _*^{\fin })^{I\setminus \{i_0\}}} \Oo (z_1,\dots ,z_m,F_{i_0}(K),\{F_i(K_i)\}_{i\neq i_0}; G(L\wedge \bigwedge _{i\neq i_0}K_i)). \] We check the two hypotheses of Lemma 20.2.4. Since \(U(\Oo )\) is stable, the reduced excisive functor \(F_{i_0}\) preserves finite colimits. These colimits are operadic, so \(T(-,L)\) sends them to limits.
For the other variable, smashing with the fixed \(\bigwedge _{i\neq i_0}K_i\) preserves finite colimits. Applying \(G\) gives a reduced excisive functor, and its finite limits are operadic. Thus \(T(K,-)\) is reduced and excisive. In both verifications, taking the end preserves the relevant limits. These arguments also apply with the fixed inputs \(z_1,\dots ,z_m\), since operadic limits and colimits are tested on multimorphism animae with arbitrary other inputs.
By the lemma, the projection \(\int _K T(K,K)\to T(S^0,S^0)\) is an equivalence. The latter is the end over the remaining variables, with \(F_{i_0}(S^0)\) now a fixed input. The induction hypothesis identifies this end with the multimorphism anima obtained by setting all remaining \(K_i=S^0\). Taking \(m=0\), we obtain \[ \Oo (\{F_i(S^0)\}_{i\in I};G(S^0)). \] Each step is the projection to the indicated component, so the resulting equivalence is the map induced by \(\Omega ^\infty \). This proves full faithfulness. □
Theorem 20.2.6 (Stabilization of \(\infty \)-operads). Let \(\Oo \) be an \(\infty \)-operad that admits finite operadic limits. Then the \(\infty \)-operad \(\oSp (\Oo )\) is stable, and the resulting functor \[ \oSp \colon \Op _{\infty }^{\lex } \to \Op _{\infty }^{\st } \] is right adjoint to the inclusion functor.
Proof. Stability is Proposition 20.2.3. Postcomposition with a map in \(\Op _{\infty }^{\lex }\) preserves reduced excisive functors. Finite limits of these functors are computed pointwise, so postcomposition and evaluation at \(S^0\) preserve finite operadic limits. Thus \(\oSp \) is a functor on \(\Op _{\infty }^{\lex }\), with a natural transformation \(\Omega ^\infty \colon \oSp \to \id \).
By Proposition 20.2.5, this transformation is an equivalence on stable \(\infty \)-operads. For the right adjoint criterion dual to Proposition 22.7.9, it remains to show that \(\oSp (\Omega ^\infty _{\Oo })\) is an equivalence for any \(\Oo \in \Op _{\infty }^{\lex }\). The universal property of Day convolution identifies \(\oSp (\oSp (\Oo ))\) with the full suboperad of \(\oDay (\Mm _{\An _*^{\fin }\times \An _*^{\fin }},\Oo )\) on bifunctors that are reduced and excisive separately. The maps \(\oSp (\Omega ^\infty _{\Oo })\) and \(\Omega ^\infty _{\oSp (\Oo )}\) evaluate at \(S^0\) in the two respective variables, so interchanging those variables exchanges the maps. The latter is an equivalence because \(\oSp (\Oo )\) is stable; hence so is the former. □
For a presentably symmetric monoidal \(\infty \)-category \(C\), Theorem 18.2.5 equips \(\Sp (C)\simeq \Sp \otimes C\) with a symmetric monoidal structure. Its multimorphism operad is the operadic stabilization constructed above.
Corollary 20.2.7 (Presentable operadic stabilization, [Nikolaus (2016), Proposition 4.9]). Let \(C\) be a presentably symmetric monoidal \(\infty \)-category. For the presentably symmetric monoidal structure on \(\Sp (C)\simeq \Sp \otimes C\), there is a natural equivalence \[ \oSp (\Mm _C) \quad \simeq \quad \Mm _{\Sp (C)}. \]
Proof. By Proposition 23.2.7(3), the reduced excisive functors form a presentable reflective subcategory of \(\Fun (\An _*^{\fin },C)\). Equip the latter with its presentably symmetric monoidal Day convolution structure. For an arbitrary functor \(F\) and a reduced excisive functor \(G\), Proposition 16.2.7 gives \[ \iHom _{\Day }(F,G)(K)\simeq \int _{L\in \An _*^{\fin }}\iHom _C(F(L),G(K\wedge L)). \] This is reduced excisive: smashing with \(L\) preserves finite colimits, while internal homs and ends preserve limits. Thus Lemma 14.7.3, Proposition 14.7.2 make the reflection symmetric monoidal, with a presentably symmetric monoidal target whose multimorphism operad is \(\oSp (\Mm _C)\).
Left Kan extension along \(S^0\colon *\to \An _*^{\fin }\) is symmetric monoidal by Lemma 16.2.5. Its composite with the reflection is left adjoint to evaluation at \(S^0\) on reduced excisive functors, hence is the stabilization functor \(C\to \Sp (C)\). This functor is therefore symmetric monoidal.
Since \(\Sp (C)\) is stable, the localization associated to the mode \(\Sp \) factors the stabilization functor through a colimit-preserving symmetric monoidal functor from \(\Sp \otimes C\) (Proposition 18.2.8). On underlying categories this is the stabilization equivalence \(\Sp \otimes C\simeq \Sp (C)\) of Theorem 18.2.5. Hence it is an equivalence of symmetric monoidal \(\infty \)-categories, as claimed. □
Remark 20.2.8. For a general symmetric monoidal \(\infty \)-category \(C\) with finite limits, \(\oSp (\Mm _C)\) need not be represented by a symmetric monoidal structure on \(\Sp (C)\) (see Chapterexercise 20.2).
Recall from Definition 11.1.16 that a symmetric monoidal \(\infty \)-category \(D\) is stably symmetric monoidal if it is stable and its tensor product is exact in each variable. Equivalently, its multimorphism operad \(\Mm _D\) is stable. The operadic stabilization adjunction therefore gives the following universal property whenever \(\oSp (\Mm _C)\) is represented by a symmetric monoidal structure.
Corollary 20.2.9. Let \(C\) be a symmetric monoidal \(\infty \)-category with finite limits whose operadic stabilization \(\oSp (\Mm _C)\) is represented by a symmetric monoidal structure on \(\Sp (C)\). For every stably symmetric monoidal \(\infty \)-category \(D\), evaluation induces an equivalence \[ \Omega ^{\infty }\circ -\colon \Fun ^{\otimes \dlax ,\lex }(D,\Sp (C))\iso \Fun ^{\otimes \dlax ,\lex }(D,C), \] where \(\lex \) means that the underlying functor preserves finite limits.
Proof. Since \(\Mm _D\) is stable, Theorem 20.2.6 shows that postcomposition with the counit \(\Omega ^\infty \) induces an equivalence on groupoid cores. To include the monoidal natural transformations, we apply the same argument with \(\Fun (E,C)\) in place of \(C\), for any small \(\infty \)-category \(E\).
A spectrum object in \(\Fun (E,C)\) is an \(E\)-diagram of spectrum objects in \(C\), since reducedness and excisiveness are checked pointwise in \(E\). Under this identification, the pointwise tensor product on \(\Fun (E,\Sp (C))\) represents \(\oSp (\Mm _{\Fun (E,C)})\). Indeed, the Day formula of Theorem 19.6.3 computes the \(I\)-ary multimorphisms on the two sides by taking the ends over \(E\) and \((\An _*^{\fin })^I\) in opposite orders. Interchanging these ends respects substitution, since the operations compose pointwise in \(E\). Evaluation at \(S^0\) likewise applies \(\Omega ^\infty \) to each object of the diagram.
By Lemma 14.4.13, a lax symmetric monoidal functor \(D\to \Fun (E,C)\) is an \(E\)-diagram of lax symmetric monoidal functors \(D\to C\), and preservation of finite limits is again checked pointwise. Thus postcomposition with \(\Omega ^\infty \) induces an equivalence between the groupoid cores of the two categories of \(E\)-diagrams. As this holds for every \(E\), the result follows from the Yoneda lemma. For the case \(C=\An \), see also [Nikolaus (2016), Corollary 4.13]. □
The other right adjoints from Section 20.1 are obtained by adapting the construction of stabilization.
Theorem 20.2.10 (Additivization of \(\infty \)-operads, [Nikolaus (2016), Theorem 5.7]). Let \(\Oo \) be an \(\infty \)-operad with an operadic terminal object, and assume finite operadic products for the monoid and group constructions. The \(\infty \)-categories of pointed objects, commutative monoids, and commutative groups in \(U(\Oo )\) admit natural operad structures \(\Oo _*\), \(\oCMon (\Oo )\), and \(\oCGrp (\Oo )\), respectively. The resulting functors \[ (-)_*\colon \Op _{\infty }^* \to \Op _{\infty }^{\pt }, \qquad \oCMon \colon \Op _{\infty }^{\mathrm {prod}} \to \Op _{\infty }^{\sadd } \qquadtext { and } \oCGrp \colon \Op _{\infty }^{\mathrm {prod}} \to \Op _{\infty }^{\add } \] are right adjoint to the respective inclusion functors.
For pointed objects, replace \(\An _*^{\fin }\) by \([1]\) with its minimum tensor product and use diagrams whose value at \(0\) is terminal. For commutative monoids and groups, use finite-product-preserving diagrams on \(\Span (\Fin )\), with its tensor product induced by cartesian products of finite sets, imposing the grouplike condition for groups. Restricting Day convolution to these diagrams gives the three \(\infty \)-operads; evaluation at the monoidal unit gives the counits. For a presentable \(\infty \)-category \(C\) whose cartesian product preserves colimits separately in each variable, these constructions on \(\Mm _{(C,\times )}\) recover the smash product on \(C_*\) and the monoidal structures on \(\CMon (C)\) and \(\CGrp (C)\) from Chapter 16. The constructions and proof of Theorem 20.2.10 are given in Section 20.4.