定义 0.1. We denote by Γop the category whose objects are finite sets ans whose morphisms are partially defined maps(only defined for some elements, not all). Write ⟨n⟩={1⋯,n}. There is a functor Cut:Δop→Γop defined by Cut([n])=⟨n⟩ and for a morphism α:[m]→[n] in Δ, we define Cut(αop) via Cut(αop)=⎩⎨⎧undefinedjundefinedifi⩽α(0)ifα(j−1)<i⩽α(j)ifα(n)<i
定义 0.2. If C is an ∞-category with finite products, a cartesian commutative monoid in C is a functor X:Γop→C such that X∘Cut:Δop→C is a cartesian monoid. If it is even a cartesian group, X is called a cartesian commutative group in C. We let CMon(C),CGrp(C) denote the full category spanned by them.
定义 0.3. An ∞-category is called semi-additive if it admits finite products and finite coproducts, its initial and final objects agree, and for any x,y∈C the natural map (idx00idy):x⊔y→x×yis an equivalence, and we denote their product and coporduct as x⊕y. We say C is additive iff it is semi-additive and the map (idx0idxidx):x⊕x→x⊕xis an equivalence, for example K(Z) and D(Z).
命题 0.4. If C is semi-additive, then the natural forgetful functors CMon(C)→Mon(C)→Care equivalences. If C is additive, so are CGrp(C)→Grp(C)→C
推论 0.5. If C is additive, then there is a canonical lift HomC:Cop×C→CGrp(Ani)If C is only semi-additive, we get a lift to CMon(Ani).
证明. In general, if F:C→D is a finite-preserving functor, then it induces a functor F∗:CGrp(C)→CGrp(D) and F∗:CMon(C)→CMon(D) because it perserves the Segal conditions. Notice that HomC:Cop×C→Ani does not preserve finite products, but the Yoneda embedding HomC(c,−):C→Ani does. Thus we can lift it to the functor HomC(c,−):C≃CGrp(C)→CGrp(Ani). To show then functorality in c, consider the Yoneda embedding HomC:Cop→Fun(C,Ani). Define the full subcategory Funlex(C,Ani)⊆Fun(C,Ani) spanned by the finite product preserving functors. We have the map Fun×(C,Ani)→Fun(CGrp(C),CGrp(Ani))≃Fun(C,CGrp(Ani))by construction. And the image of the Yoneda embedding lies in Fun×(C,Ani). Thus we have HomC:Cop→Fun(C,CGrp(Ani)). Similarly for the semi-additive case.
□
Recall that we have the adjunction B:Mon(Ani)↔∗/Ani:ΩNow B=∣∣ and Ω both preseve finite products(B by explicit computations and Ω by right adjointness). Thus we can lift this adjunction:B:CMon(Mon(Ani))↔CMon(∗/Ani):ΩIt is easy to see the equivalence CMon(∗/C)≃CMon(C). We want to show the diagram
commutes and all arrows are equivalences. To prove this statement, we let Cat∞×⊆Cat∞ be the subcategory spanned by ∞-categories admitting finite products and functors preserving finite products. Also let Cat∞add⊆Cat∞semi−add⊆Cat∞× denote the full subcategory spanned by additive and semi-additive ∞-categories.
定理 0.6. If an ∞-category C has products, then CMon(C) and CGrp(C) are semi-additive and additive respectively. Furthermore, the functors CMonCGrp:Cat∞×→Cat∞semi−add:Cat∞×→Cat∞addare right adjoints to the inclusion functors.
Using this theorem, we can get a proof of the claim. First we prove that the maps forget:CMon(CMon(C))→CMon(C) and forget∗:CMon(CMon(C))→CMon(C) agree and are equivalences. In fact forget:CMon(−)→− is the counit of the adjunction in theroem 0.6(see the proof). Hence forget∗ is an equivalence. Because the inclusion map is full fatihful, we know that the other is an equivalence. Moreover, by writing down the triangle identies we find that the maps are inverses to the unit CMon(C)→CMon(CMon(C))(one regards CMon(C)∈Cat∞× and the other regard CMon(C)∈Cat∞semi−add then using the identities respectively). We want to replace one CMon by Mon. To do this, note that if we have CMon(C) is semi-additive, we have CMon(CMon(C))≃Mon(CMon(C)). Also CMon(Mon(C))≃Mon(CMon(C)) by direct inspection as the full subcategory of Fun(Δop×Γop,C). To prove theorem 0.6, we need a lemma
引理 0.7. Let C be a category with finite products. Then C is semi-additive if the following conditions hold:
1.
The terminal objects ∗∈C is also initial.
2.
We have a natural transformation μx:(Δ:x↦x×x)↦idC such that both compositions xx≃x×∗⟶idx×0x×x→μxx≃∗×x⟶0×idxx×x→μxxare homotopic to idx for all x∈C, and the diagram
证明. We first prove that CMon(C) is semi-additive by lemma 0.7. The constant functor const∗ is the final object. To see it is initial, let X∈CMon(C) and compute Nat(const∗,X)≃HomC(∗,<n>∈ΓoplimXn)≃HomC(∗,X0)≃HomC(∗,∗)≃∗Now we need to construct functorial maps μM:M×M→M for all M∈CMon(C). Consider the map ×:Γop×Γop→Γop by taking the product of sets and morphisms. It induces a functor Fun(Γop,C)→Fun(Γop,Fun(Γop,C)). We can see that the Segal conditions are preserved. So we get a functor Double:CMon(C)→CMon(CMon(C))By construction, Double1 is just the identity and Double2 is naturally equivalent to the diagonal Δ:M→M×M. Hence we define μM to be the multiplication m:Double2→Double1 induced by m:<2>→<1>,1,2↦1. This transformation is functoral and these conditons also hold. Thus CMon(C) is semi-additive by the lemma 0.7. To show CGrp(C) is additive, we can know that for G∈CGrp(C), the codiagonal map G×G→G is induced by μG, hence by the multiplication on Double(G). Thus we only need to show that Double(G) is a cartesian commutative group it self. This is easy. Finally, to show the adjunctions, just notice that using CMon(C) is semi-additive, then for R=CMon(−) and η:R⇒id, we have ηR:R∘R⇒R and Rη:R∘R⇒R are both equivaleces. So by some nonsenses, we have the adjunctions.
□
推论 0.8. There is a commutative diagram of horizontal adjunction
We also have:
1.
Restricting B to CGrp(Ani) gives a fully faithful functor B:CGrp(Ani)→CMon(Ani).
2.
Both functors B,Ω actually take values in CGrp(Ani).
3.
ΩB is left adjoint to the inclusion CGrp(C)⊆CMon(C).
In particular, the top row adjunction restricts to the adjunction
推论 0.9. We get a commutative solid diagram
Finally we talk about free E∞-monoids. Similarly, the forgetful functor CMon(Ani)→Ani has a left adjoint FreeCMon:Ani→CMon(Ani). For X∈Ani, it is given by FreeCMon(X)1≃n⩾0∐XhSnnWe need to define homotopy-quotient under the action of a group. In general, given a functor F:BG→C, where C is an ∞-cateogry with colimits and G is a discrete group, we put FhG:=colimBGFOn each Xn, to get this homotopy-quotient, we give a functor BS→core(Ani) sending the 0-complex of BSn to Xn. This is the same as giving a functor S→ΩXn(core(Ani))≃Homcore(Ani)(Xn,Xn). Then we just conder the permutation actions.