ScalingStacks

[0MIG]

Proof. The claim is that for any kk-correspondence N→CkN\to C_{k}, the functor

−×CkN:(Gauntn/Ck)→(Gauntn/Ck)-\times_{C_{k}}N\colon(\gaunt_{n}/C_{k})\to(\gaunt_{n}/C_{k})

admits a right adjoint Hom¯Ck⁡(N,−)\uHom_{C_{k}}(N,-).

Since Gauntn/Ck\gaunt_{n}/C_{k} is locally finitely presentable, it is cocomplete and admits a strong generator [1, Th. 1.20] and is co-wellpowered [1, Th. 1.58]. Thus by the special adjoint functor theorem [30, Sect. V.8] the functor

−×CkN:(Gauntn/Ck)→(Gauntn/Ck)-\times_{C_{k}}N\colon(\gaunt_{n}/C_{k})\to(\gaunt_{n}/C_{k})

admits a right adjoint precisely if it commutes with colimits.

When k=0k=0, this follows from the fact that Gauntn\gaunt_{n} itself is cartesian closed.

For k>0k>0, suppose (Gauntn−1/Ck−1)(\gaunt_{n-1}/C_{k-1}) is cartesian closed. To prove that the category (Gauntn/Ck)(\gaunt_{n}/C_{k}) is cartesian closed, we require a description of colimits in terms of the equivalent category Corrnk\corr_{n}^{k}. For any small category Λ\Lambda and any diagram X:Λ→CorrnkX\colon\Lambda\to\corr_{n}^{k} with

Xλ=(Xλ,0,Xλ,1,Fλ),X_{\lambda}=(X_{\lambda,0},X_{\lambda,1},F_{\lambda}),

it is easy to see that the colimit is given by the triple (X0,X1,F)(X_{0},X_{1},F) where

X0=colimλ∈ΛXλ,0​ and ​X1=colimλ∈ΛXλ,1,X_{0}=\colim_{\lambda\in\Lambda}X_{\lambda,0}\text{\quad and\quad}X_{1}=\colim_{\lambda\in\Lambda}X_{\lambda,1},

and F:X0op×X1→Corrn−1k−1F\colon X_{0}^{\mathrm{op}}\times X_{1}\to\corr_{n-1}^{k-1} is the enriched left Kan extension of

colimλ∈ΛFλ:colimλ∈Λ(Xλ,0op×Xλ,1)→Corrn−1k−1\colim_{\lambda\in\Lambda}F_{\lambda}\colon\colim_{\lambda\in\Lambda}(X_{\lambda,0}^{\mathrm{op}}\times X_{\lambda,1})\to\corr_{n-1}^{k-1}

along the diagonal

colimλ∈Λ(Xλ,0op×Xλ,1)→X0op×X1.\colim_{\lambda\in\Lambda}(X_{\lambda,0}^{\mathrm{op}}\times X_{\lambda,1})\to X_{0}^{\mathrm{op}}\times X_{1}.

Now for any object Y=(Y0,Y1,G)∈CorrnkY=(Y_{0},Y_{1},G)\in\corr_{n}^{k}, we wish to compare (colimλ∈ΛXλ)×Y(\colim_{\lambda\in\Lambda}X_{\lambda})\times Y and colimλ∈Λ(Xλ×Y)\colim_{\lambda\in\Lambda}(X_{\lambda}\times Y). In light of our descriptions of products in Corrnk\corr_{n}^{k}, we see that the former is (X0×Y0,X1×Y1,F⊗G)(X_{0}\times Y_{0},X_{1}\times Y_{1},F\otimes G), and the latter is the colimit of the diagram Z:Λ→CorrnkZ\colon\Lambda\to\corr_{n}^{k} that carries λ\lambda to

(Xλ,0×Y1,Xλ,1×Y1,Fλ⊗G).(X_{\lambda,0}\times Y_{1},X_{\lambda,1}\times Y_{1},F_{\lambda}\otimes G).

Note that, since Gauntn\gaunt_{n} is cartesian closed, one has

colimλ∈Λ(Xλ,0×Y0)≅X0×Y0​ and ​colimλ∈Λ(Xλ,1×Y1)≅X1×Y1;\colim_{\lambda\in\Lambda}(X_{\lambda,0}\times Y_{0})\cong X_{0}\times Y_{0}\text{\quad and\quad}\colim_{\lambda\in\Lambda}(X_{\lambda,1}\times Y_{1})\cong X_{1}\times Y_{1};

hence our description of colimits in Corrnk\corr_{n}^{k} exhibits the colimit of ZZ as (X0×Y0,X1×Y1,(F⊗G)′)(X_{0}\times Y_{0},X_{1}\times Y_{1},(F\otimes G)^{\prime}), where (F⊗G)′(F\otimes G)^{\prime} is the enriched left Kan extension of

colimλ∈Λ(Fλ⊗G):colimλ∈Λ((Xλ,0×Y0)op×(Xλ,1×Y1))→Corrn−1k−1\colim_{\lambda\in\Lambda}(F_{\lambda}\otimes G)\colon\colim_{\lambda\in\Lambda}((X_{\lambda,0}\times Y_{0})^{\mathrm{op}}\times(X_{\lambda,1}\times Y_{1}))\to\corr_{n-1}^{k-1}

along the diagonal

colimλ∈Λ((Xλ,0×Y0)op×(Xλ,1×Y1))→(X0×Y0)op×(X1×Y1).\colim_{\lambda\in\Lambda}((X_{\lambda,0}\times Y_{0})^{\mathrm{op}}\times(X_{\lambda,1}\times Y_{1}))\to(X_{0}\times Y_{0})^{\mathrm{op}}\times(X_{1}\times Y_{1}).

Our induction hypothesis is that Corrn−1k−1\corr_{n-1}^{k-1} is cartesian closed; so this enriched left Kan extension can be identified with the composition of the enriched left Kan extension of

colimλ∈Λ(Fλ,G):colimλ∈Λ(Xλ,0op×Xλ,1)×(Y0op×Y1)→Corrn−1k−1×Corrn−1k−1\colim_{\lambda\in\Lambda}(F_{\lambda},G)\colon\colim_{\lambda\in\Lambda}(X_{\lambda,0}^{\mathrm{op}}\times X_{\lambda,1})\times(Y_{0}^{\mathrm{op}}\times Y_{1})\to\corr_{n-1}^{k-1}\times\corr_{n-1}^{k-1}

along the diagonal

colimλ∈Λ(Fλ,G):colimλ∈Λ(Xλ,0op×Xλ,1)×(Y0op×Y1)→(X0op×X1)×(Y0op×Y1).\colim_{\lambda\in\Lambda}(F_{\lambda},G)\colon\colim_{\lambda\in\Lambda}(X_{\lambda,0}^{\mathrm{op}}\times X_{\lambda,1})\times(Y_{0}^{\mathrm{op}}\times Y_{1})\to(X_{0}^{\mathrm{op}}\times X_{1})\times(Y_{0}^{\mathrm{op}}\times Y_{1}).

But now this left Kan extension is simply the product of GG with the left Kan extension that defines FF. In other words, we have an isomorphism (F⊗G)′≅F⊗G(F\otimes G)^{\prime}\cong F\otimes G, whence the proof is complete. ∎

Original mathematics by the credited authors. Source collection and HTML conversion remain in progress.

Clark Barwick, Christopher Schommer-Pries

Original source: arXiv:1112.0040v6

    Original source · 1112.0040v6