Goodwillie Towers of \infty-Categories and Desuspension

This new condition is clearly stronger than (2'), so we will show that condition (2') suffices. Consider a pushout square in \Delta_n

If S and S' intersect in one element then we are done by condition (2'). Otherwise they intersect in at least two objects, call one of them i so that S\cap S' - \{i\} is nonempty. Consider the diagram

By condition (2') the outside and left hand squares are pushouts, so that the right hand square is also a pushout.

The idea is that an injection [k]\to [n] designates a unique subset of [n], and this allows us to pass between the two categories.

Consider a chain of inclusions [m_0]\xrightarrow{f_1}[m_1]\xrightarrow{f_2}\dots\xrightarrow{f_k}[m_k] in \Delta_n which does not belong to \Delta_n^0. We want to show that such a k-simplex lifts to one in \square[n+1]_0. Necessarily [m_k] = [n]. The lift is given by the image of the chain of inclusions J_0 \subseteq J_1 \subseteq \dots\subseteq [n] where J_i = im(f_k\circ\cdots\circ f_{i+1}). This is similar to the proof of [HA, Proposition 4.1.5.7].

There is a fully faithful left Kan extension functor

from diagrams defined on subsets of [n-1] of cardinality less than two to diagrams defined on the entire powerset of [n-1]. Fun^{sco}(\square[n], \mathcal{C}) can be identified with the image of this functor on the subcategory of Fun(P_{\le1}(\{0,\dots,n-1\}, \mathcal{C}) of diagrams with initial vertex a 0 object. This category is equivalent to \mathcal{C}^n.

The case for n=2 was proved already in Example 1.13. If n>2 we can choose four distinct elements \{a,b,c,d\}. Consider the subsets \{a,b\}, \{b,c\} and form the special punctured 3-cube

Any special punctured n-cube \mathcal{X} takes this subdiagram to a special punctured 3-cube, providing an equivalence \mathcal{X}\{a,c\}\simeq \mathcal{X}\{b,c\} via the argument of Example 1.13. Likewise we can form a punctured 3-cube amongst \{b,c,d\} and obtain an equivalence \mathcal{X}\{b,c\}\simeq \mathcal{X}\{b,d\}, giving an equivalence \mathcal{X}\{a,c\}\simeq\mathcal{X}\{b,d\}. In this way we can connect any two cardinality two subsets of [n] via an equivalence.

This is clearly the case for n=1. We will show that the square

is a pullback, proving the claim by induction. This will be accomplished by showing that for any A_{n-1}-cogroup object A in \mathcal{C} there is an equivalence of fibers

By [HA, Theorem 4.1.5.8] there is an essentially unique extension of A to an A_n-precogroup object A^0:\Delta^0_n\to \mathcal{C} and so an equivalence of fibers coGrp_n(\mathcal{C})_A\simeq coGrp_n(\mathcal{C})_{A^0}. From the pushout square of Lemma 1.17 we obtain a pullback square

and hence an equivalence of fibers coGrp_n(\mathcal{C})_{A^0} \simeq T_n(\mathcal{C})_{\pi^0_nA^0}. It suffices to show now that the natural map of fibers T_n(\mathcal{C})_{\pi^0_nA^0}\to T_n(\mathcal{C})_{\pi_{n-1} A} is an equivalence.

By abuse of notation, let A denote the object A[1], The functor \pi_n^0A^0 factors via the pullback in the diagram below since the cardinality two vertices are all equal to A. Note also that the pullback in the bottom left vertex of the diagram below is a homotopy pullback, so that the square is an actual pullback.

The weak equivalence Fun^{sco}(\square[n], \mathcal{C})\xrightarrow{\sim} \mathcal{C}^n of Lemma 1.20 gives a weak equivalence

which allows us to greatly simplify the bottom left vertex of the above diagram.

Note now that \pi_n^0 A^0 lands in T_n^0(\mathcal{C})\hookrightarrow Fun(\square[n+1]_0^0, \mathcal{C}) and that under the above equivalence T_{n-1}(\mathcal{C})\times_{\mathcal{C}^n} \{A\} maps into

So that the above square factors and we can compute the fiber T_n(\mathcal{C})_{\pi_n^0A^0} equally as the pullback in the left hand square below, which is seen to be a pullback using two applications of the pullback pasting lemma.

Likewise \pi_{n-1}A factors via T_{n-1}(\mathcal{C})\times_{\mathcal{C}^n}\{A\} and we can compute the fiber T_n(\mathcal{C})_{\pi_{n-1}A} as the left hand pullback in

Now compare the above right square with the square in the definition of P^0.

We consider the difference in these two squares. By construction we demand that the special punctured (n+1)-cubes in P^0 and P both are equal to A on every initial vertex in the subset \square[n]_0\hookrightarrow \square[n+1]_0. In addition, in P^0 we demand that the vertices of cardinality one in \square[n]\hookrightarrow \square[n+1]_0 are also equal to A. By Lemma 1.21 this later condition is automatic in a special punctured (n+1)-cube already: the special n-cube in a special punctured (n+1)-cube provides no additional data. Because of this we have an equivalence P\simeq P^0, and so finally we obtain an equivalence of fibers T_n(\mathcal{C})_{\pi_n^0A^0}\simeq T_n(\mathcal{C})_{\pi_{n-1}A} as desired.

This follows from Theorem 2.5 since P_n\mathcal{C} is a filtered colimit of fully faithful functors \Sigma starting from \mathcal{C}.

This follows from [RVC, Theorem 1.1] or as explained in the (not static!) discussion at the end of the proof of [RVE, Proposition 8.2.11].

There is an equivalence

which we must show lifts to an equivalence

By Lemma 2.10 the legs of the limit cone \mathcal{C}\rightarrow \mathcal{C}_k create colimits and equivalences, so they create the Segal condition (2’) of Definition 1.5.

This is dual to the unital version of [HA, Prop 4.1.4.9].

The equivalences P_n(\mathcal{C})\xrightarrow{\Sigma} coGrp_n(P_n\mathcal{C}) assemble into a tower of equivalences

Providing an equivalence P_{\infty} \xrightarrow{\sim} lim_n coGrp_n(P_n\mathcal{C}). Because the functors P_{\infty}\mathcal{C}\rightarrow P_n\mathcal{C} jointly create colimits we may identify this functor on underlying objects with the suspension \Sigma: P_\infty\mathcal{C}\rightarrow P_{\infty}\mathcal{C}.

On the other hand, the family of functors P_{\infty}\mathcal{C}\xrightarrow{\pi_n}P_n\mathcal{C} satisfy the setup of Notation 2.8, and so Proposition 2.13 provides an equivalence

However, the diagonal functor \Delta: \mathbb{N}\rightarrow \mathbb{N}\times\mathbb{N} is final, inducing an equivalence

Providing the desired equivalence \Sigma: P_{\infty}\mathcal{C}\xrightarrow{\sim} coGrp_{\infty}(P_{\infty}\mathcal{C}) which finishes the proof of Theorem 2.14.

This follows from Lemma 2.12 and the proof of Theorem 2.14.

If \mathcal{C}\xrightarrow{\sim} P_\infty(\mathcal{C}) is an equivalence then \Sigma:\mathcal{C}\xrightarrow{\sim} coGrp_\infty(\mathcal{C}) is an equivalence by Theorem 2.14 and so \mathcal{C}\xrightarrow{\sim}\hat{P}_\infty\mathcal{C} is an equivalence. The result then follows by the two out of three property for equivalences.

Consider the square

By [H, Theorem 1.3] the horizontal functors are equivalences and by Corollary 3.1 the right vertical map is an equivalence, hence also the left hand vertical functor is an equivalence.

In the corresponding square for the case of S_*^{conv}

we have again that the bottom right cospan is an equivalence, but now the top horizontal functor is only fully faithful, so that the left vertical functor is fully faithful.