6
$\begingroup$

This is in part motivated from my attempt to understand tate diagonal in III.1 of Thomas Nikolaus, Peter Scholze, On topological cyclic homology, arXiv:1707.01799. I just want to make my understanding precise.


Particular goal:

How the map (III.1) $T_p: Sp \rightarrow Sp$ $$ X \mapsto (X \otimes \cdots \otimes X)^{tC_p}$$ is defined rigorously. What I could define: I could define a map $$ Sp \rightarrow Sp^{\times n} \rightarrow Sp$$ $$ X \mapsto (X,\ldots, X) \mapsto X \otimes \cdots \otimes X $$ using monoidal structure of $Sp^\otimes$ of spectra.

Question How do I rigorously lift this to map $Sp^{BC_p}$? ( allowing me to apply the Tate functor $(-)^{tC_p} : Sp^{BC_p} \rightarrow Sp$. )


EDIT: Most of my question of this goal have been resolved in the replies below (of which all are nice answers). I still have the following confusion

how does one prove formulas fo the underlying(under the notation of Maxime) the adjunction: $$ U:Sp^{BG} \rightarrow Sp:Ind, CoInd $$ of "forgetful"/"inclusion"? where Ind and CoInd are left and right adjoint respectively. i.e. It seems that $$ \bigoplus_g X \simeq UInd X $$ $$ U CoInd X \simeq ?? $$

In particular I am confused about the computation $CoInd(Sp) \simeq Sp^{\times n}$.

$\endgroup$

3 Answers 3

6
$\begingroup$

Let $C$ be a complete $\infty$-category.

Let $U:Fun(BC_n,C)\to C$ denote the forgetful functor, $\mathrm{CoInd}$ its right adjoint, and $(-)^{triv}$ the functor given by precomposition along $BC_n\to *$.

Then we have a canonical equivalence $U(X^{triv})\to X$ which yields, by adjunction, a map $X^{triv}\to \mathrm{CoInd}(X)$ which is $C_n$-equivariant.

Apply this to $C= Cat_\infty$ and $X=Sp$ yields a $C_n$-equivariant map $Sp \to \mathrm{CoInd}(Sp)$. Now $\mathrm{CoInd}(Sp) = Sp^{\times n}$ with the permutation action.

Now $Sp$ can be canonically seen as a commutative monoid in $Cat_\infty$, that is, a certain type of functor $Fin_*\to Cat_\infty$, which we can then obviously restrict to $Fin$ to get $Fin\to Cat_\infty$, informally given by $n\mapsto Sp^{\times n}$.

In particular, we get a $\Sigma_n$-equivariant map $Sp^{\times n}\to Sp$ corresponding to the smash product, and the action of $\Sigma_n$ on $Sp^{\times n}$ restricts to the permutation action of $C_n$

You can prove this by dealing with the universal case.

Another way to do that, which certainly agrees, is to note that in $CAlg(Cat_\infty)$, products and coproducts agree, i.e. it is preadditive and so induction and co-induction agree. In particular you get for free a $C_n$-equivariant map $\mathrm{CoInd}(Sp)\to Sp$ (from $Sp\to U(Sp^{triv})$) which is also given by smash product.

Anyways, it follows that both $Sp\to Sp^{\times n}$ and $Sp^{\times n}\to Sp$ are $C_n$-equivariant

Your left Kan extension construction will not work. Left Kan extending along $*\to BG$ is left adjoint to the forgetful functor, i.e. it is induction - when composed with the forgetful map, this looks like $\bigoplus_{g\in G}$, so if you left Kan extend $X\mapsto X\otimes ... \otimes X$, you will get $\bigoplus_{g\in C_p}X\otimes... \otimes X$, and no permutation action.

As Harry already pointed out, the answer to your side question is "yes", the inclusion has both a left and a right adjoint, in particular it preserves limits and colimits.

$\endgroup$
4
  • 1
    $\begingroup$ Oh man, Ind(C) for this is not the best notation lol. I did upvote though =]. $\endgroup$ Nov 16, 2020 at 12:06
  • $\begingroup$ @HarryGindi : rats, you are absolutely right ! I didn't even notice, ha - let me try to rephrase that $\endgroup$ Nov 16, 2020 at 12:10
  • $\begingroup$ Wow, I've been learning a lot from these posts. @MaximeRamzi do you mind elaboraitng on CoInd computation? or perhaps a reference? I updated my question. $\endgroup$
    – Bryan Shih
    Nov 17, 2020 at 0:06
  • 1
    $\begingroup$ Very generally, if $C$ is a category with finite (co)products and $G$ is a finite group, $CoInd^G(X) = X^G$ with permutation action; here the underlying object of $X^G$ is just $\prod_{g\in G}X$ and $G$ really acts by permuting factors. To prove this, you may want to use the fact that co-induction (resp. induction) is right (resp. left) Kan extension along $*\to BG$, and the fibers are just $G$. $\endgroup$ Nov 17, 2020 at 8:35
4
$\begingroup$

I actually worked this out a few months ago (with a hint from Denis Nardin) and wrote this in a message to a friend of mine:

Consider SymmMonCat as a symmetric monoidal category with the cocartesian monoidal structure. I filled in the details for my own sake: SymmMonCat itself has a symmetric monoidal structure, the cocartesian symmetric monoidal structure. Every symmetric monoidal category is canonically an algebra in this symmetric monoidal category wrt the coproduct. Choose our object C as a functor Δ^0→SymmMonCat. Since SymmMonCat has finite coproducts, the functor extends to a functor FinSet→SymmMonCat . Let BΣ_n×Δ^1→FinSet be the map from yesterday. Then composing BΣ_n×Δ^1→FinSet→SymmMonCat, we get an arrow Δ^1→SymmMonCat^{BΣ_n} classifying the equivariant fold map. Finally, to obtain the arrow we wanted, take the composite with the limit functor SymmMonCat^BΣ_n→SymmMonCat. This gives a map Δ^1→SymmMonCat classifying the map C→C^{BΣ_n} that does what we wanted. That's the excruciating detail. The short of it is consider the fold map for an object C in SymmMonCat, which is BΣ_n-equivariant. Then take homotopy fixed points levelwise.

Getting from here to the cyclic version is the obvious thing (consider the canonical permutation representation to get your map by restriction). I don't mind elaborating more, but I don't have time right this second to flesh this out.

Edit: Oh, also the mysterious map $BΣ_n\times \Delta^1\to FinSet$ is the map corresponding to the map sending $0$ to $\langle n \rangle$ with the obvious $\Sigma_n$-action and 1 to $\langle 1 \rangle$ with the trivial $\Sigma_n$-action. It's the map that canonically factors through the cone $B\Sigma_n \star \Delta^0\to FinSet$.

Also, yes, the inclusion of ∞-groupoids in ∞-categories has both a left and right adjoint (the left adjoint is 'invert everything' and the right adjoint is taking the core ∞-groupoid).

$\endgroup$
4
$\begingroup$

There are several ways to do this, depending on how much technology one is interested in using.

One way to do it is to use the fact that the $\infty$-category of commutative monoid objects in a category with finite products, $\mathsf{CMon}(\mathcal{C})$, can be computed as $\mathsf{Fun}^{\times}(\mathsf{Span}(\mathrm{Fin})^{op}, \mathcal{C})$- i.e. the $\infty$-category of product-preserving presheaves on the $(2,1)$-category of spans of finite sets, with values in $\mathcal{C}$. (See, e.g., Theorem 6.5 in the paper of Nardin https://arxiv.org/pdf/1608.07704.pdf for a proof in a more general context; essentially the proof is by right Kan extending from the restriction to finite pointed sets.) Now, $\mathrm{Map}_{\mathsf{Span}(\mathrm{Fin})}(\bullet, \bullet)$ is then the groupoid of finite sets, and in particular receives a map $\mathrm{B}\Sigma_n \to \mathrm{Map}_{\mathsf{Span}(\mathrm{Fin})}(\bullet, \bullet)$. Now take $\mathcal{C}=\mathsf{Cat}_{\infty}$ and consider the functor $J \mapsto \mathsf{Sp}^{\times J}$ given by the symmetric monoidal structure on $\mathsf{Sp}$. Composing we get: $\mathrm{B}\Sigma_n \to \mathrm{Map}_{\mathsf{Span}(\mathrm{Fin})}(\bullet, \bullet) \to \mathrm{Map}_{\mathsf{Cat}_{\infty}}(\mathsf{Sp}, \mathsf{Sp})$. That's the same as a functor $\mathsf{Sp} \to \mathsf{Sp}^{\mathrm{B}\Sigma_n} = \mathsf{Sp}^{h\Sigma_n}$, which is what you're after (you can take $n=p$ and restrict to $C_p$ if you want).

But maybe you don't want to use that fact about spans. That's fine. You can follow the approach from the beginning of section 2.2 in DAGXIII (https://www.math.ias.edu/~lurie/papers/DAG-XIII.pdf). The point is this: if $\mathcal{D}$ is a symmetric monoidal $\infty$-category, i.e. a commutative monoid/algebra object in $\mathsf{Cat}_{\infty}$, then we automatically get a functor $\mathrm{Sym}(\mathcal{D}) \to \mathcal{D}$ from the free commutative algebra object. The free commutative algebra object is computed as $\coprod \mathcal{D}^{\times n}_{h\Sigma_n}$. Restricting to the $n$th summand gives a map $\mathcal{D}^{\times n}_{h\Sigma_n} \to \mathcal{D}$ which refines the functor $(X_1, ..., X_n) \mapsto X_1\otimes \cdots \otimes X_n$. We may view it just as well as a functor $\mathcal{D}^{\times n} \to \mathcal{D}$ in $\mathsf{Fun}(\mathrm{B}\Sigma_n, \mathsf{Cat}_{\infty})$ (since taking colimits is adjoint to the constant diagram functor). On the other hand, we also have a diagonal map $\mathcal{D} \to \mathcal{D}^{\times n}$ which is obtained by applying $\mathrm{Fun}(-, \mathcal{D})$ to the $\Sigma_n$-equivariant map of sets $\{1, ..., n\} \to \bullet$, and hence is $\Sigma_n$-equivariant. Composing gives $\mathcal{D} \to \mathcal{D}^{\times n} \to \mathcal{D}$ with a $\Sigma_n$-equivariant structure that does what you want.


Also:

  • This functor $X \mapsto (X^{\otimes p})^{tC_p}$ is not 'the Tate diagonal'. The Tate diagonal is a natural transformation (the unique lax symmetric monoidal natural transformation) $X \to (X^{\otimes p})^{tC_p}$.
$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.