ScalingStacks

4. Tensor products and integral transforms

In this section, we study the relation between the geometry and algebra of perfect stacks. We begin with some basic properties of tensor products of ∞\infty-categories. Then we prove (Theorem 4.7) that the ∞\infty-category of sheaves on a product of perfect stacks, and more generally a derived fiber product, is given by the tensor product of the ∞\infty-categories of sheaves on the factors. This implies that the ∞\infty-category of sheaves on a perfect stack is self-dual, which in turn allows us to describe ∞\infty-categories of functors by ∞\infty-categories of integral kernels (Corollary 4.10). Finally, we extend this last result to a relative setting where the base is an arbitrary derived stack with affine diagonal.

4.1. Tensor products of ∞\infty-categories

In this section, we consider some properties of tensor products of ∞\infty-categories that will be used in what follows. First we prove (Proposition 4.1) that ∞\infty-categories of modules over associative algebra objects are dualizable, with duals given by modules for the opposite algebra, and that tensor product of algebras induces tensor products on module categories. We then discuss the tensor product of small stable categories, and its compatibility with passing to the corresponding presentable stable ∞\infty-categories of Ind\operatorname{Ind}-objects.

4.1.1. Algebras and modules

Recall the tensor product of presentable stable ∞\infty-categories developed in [L4], [L5]. Namely, the ∞\infty-category 𝒫​rL\mathcal{P}r^{\rm L} of presentable ∞\infty-categories (with morphisms given by left adjoints) carries a natural symmetric monoidal tensor product that preserves stable objects.

Let 𝒞\mathcal{C} be a monoidal ∞\infty-category. For two presentable ∞\infty-categories ℳ\mathcal{M} and ℳ′\mathcal{M}^{\prime} left tensored over 𝒞\mathcal{C}, we denote by Fun𝒞L⁡(ℳ,ℳ′)\Fun^{\rm L}_{\mathcal{C}}(\mathcal{M},\mathcal{M}^{\prime}) the ∞\infty-category of left adjoints from ℳ\mathcal{M} to ℳ′\mathcal{M}^{\prime} that preserve the tensor over 𝒞\mathcal{C}.

Similarly, the opposite category of 𝒫​rL\mathcal{P}r^{\rm L} is the ∞\infty-category 𝒫​rR\mathcal{P}r^{\rm R} of presentable ∞\infty-categories with morphisms given by right adjoints. For two presentable ∞\infty-categories ℳ\mathcal{M} and ℳ′\mathcal{M}^{\prime} left cotensored over 𝒞\mathcal{C}, we denote by Fun𝒞R⁡(ℳ,ℳ′)\Fun^{\rm R}_{\mathcal{C}}(\mathcal{M},\mathcal{M}^{\prime}) the ∞\infty-category of right adjoints from ℳ\mathcal{M} to ℳ′\mathcal{M}^{\prime} that preserve the cotensor over 𝒞\mathcal{C}.

0NXL

Proposition 4.1. Let 𝒞\mathcal{C} be a stable presentable symmetric monoidal ∞\infty-category, and A∈𝒞A\in\mathcal{C} an associative algebra object.

  1. (1)

    For any 𝒞\mathcal{C}-module ℳ\mathcal{M}, there is a canonical equivalence of ∞\infty-categories

    ModA⁡(𝒞)⊗𝒞ℳ≃ModA⁡(ℳ).\Mod_{A}(\mathcal{C})\otimes_{\mathcal{C}}\mathcal{M}\simeq\Mod_{A}(\mathcal{M}).
  2. (2)

    For A′∈𝒞A^{\prime}\in\mathcal{C} a second associative algebra, there is a canonical equivalence of ∞\infty-categories

    ModA⊗A′⁡(𝒞)≃ModA⁡(𝒞)⊗𝒞ModA′⁡(𝒞).\Mod_{A\otimes A^{\prime}}(\mathcal{C})\simeq\Mod_{A}(\mathcal{C})\otimes_{\mathcal{C}}\Mod_{A^{\prime}}(\mathcal{C}).
  3. (3)

    The ∞\infty-category of modules ModA⁡(𝒞)\Mod_{A}(\mathcal{C}) is dualizable as a 𝒞\mathcal{C}-module with dual given by the ∞\infty-category of modules ModAop⁡(𝒞)\Mod_{A^{\rm op}}(\mathcal{C}) over the opposite algebra.

The proof will depend on the following two lemmas:

0NXM

Lemma 4.2. Let ℳ\mathcal{M} and ℳ′\mathcal{M}^{\prime} be stable presentable ∞\infty-categories that are left tensored and cotensored over 𝒞\mathcal{C}. Let G:ℳ′→ℳG:\mathcal{M}^{\prime}\rightarrow\mathcal{M} be a right adjoint that is tensored and cotensored over 𝒞\mathcal{C}. Assume further that GG is colimit preserving. Then GG is conservative if the induced functor Fun𝒞R⁡(𝒟,ℳ′)→Fun𝒞R⁡(𝒟,ℳ)\Fun_{\mathcal{C}}^{\rm R}(\mathcal{D},\mathcal{M}^{\prime})\rightarrow\Fun_{\mathcal{C}}^{\rm R}(\mathcal{D},\mathcal{M}) is conservative for any 𝒟\mathcal{D}.

0NXN

Proof. Suppose GG is not conservative. Then to prove the lemma, it suffices to exhibit a presentable ∞\infty-category 𝒟\mathcal{D} also cotensored over 𝒞\mathcal{C} and a nontrivial right adjoint j:𝒟→ℳ′j:\mathcal{D}\rightarrow\mathcal{M}^{\prime} cotensored over 𝒞\mathcal{C} such that j∘Gj\circ G is trivial.

Define 𝒟\mathcal{D} to be the full ∞\infty-subcategory of ℳ′\mathcal{M}^{\prime} of GG-acyclic objects, that is, objects m∈ℳm\in\mathcal{M} such that G⁡(m)G(m) is trivial. Our first task is to show that 𝒟\mathcal{D} is indeed presentable.

Observe that 𝒟\mathcal{D} is equivalent to the fiber product 𝒟≃0×ℳℳ′\mathcal{D}\simeq 0\times_{\mathcal{M}}\mathcal{M}^{\prime}, where the limit is computed in the ∞\infty-category Cat∞\rm Cat_{\infty} of ∞\infty-categories. Recall by [L2, Proposition 5.5.3.13], the natural functor 𝒫​rL→Cat∞\mathcal{P}r^{\rm L}\rightarrow{\rm Cat}_{\infty} preserves limits. Furthermore, the forgetful functor Mod𝒞​(𝒫​rL)→𝒫​rL\mathrm{Mod}_{\mathcal{C}}(\mathcal{P}r^{\rm L})\to\mathcal{P}r^{\rm L} also preserves limits since it has a left adjoint (given by induction).

Since the functor GG preserves colimits and is 𝒞\mathcal{C}-linear, we may regard it as a morphism in Mod𝒞​(𝒫​rL)\mathrm{Mod}_{\mathcal{C}}(\mathcal{P}r^{\rm L}). Thus 𝒟\mathcal{D} can be computed as a limit in Mod𝒞​(𝒫​rL)\mathrm{Mod}_{\mathcal{C}}(\mathcal{P}r^{\rm L}), and so can be regarded as an object of 𝒫​rL\mathcal{P}r^{\rm L}. In other words, 𝒟\mathcal{D} is presentable and furthermore tensored over 𝒞\mathcal{C}. Finally, since 𝒟\mathcal{D} is tensored over 𝒞\mathcal{C}, it is automatically cotensored as well.

Now it remains to show that the inclusion j:𝒟→ℳ′j:\mathcal{D}\rightarrow\mathcal{M}^{\prime} is indeed a right adjoint and cotensored over 𝒞\mathcal{C}. Since jj preserves all limits and colimits (and in particular κ\kappa-filtered colimits), the adjoint functor theorem applies. Finally, since GG is cotensored over 𝒞\mathcal{C}, jj is as well. ∎

0NXP

Lemma 4.3. Let ℳ\mathcal{M} be a stable presentable ∞\infty-category which is left tensored and cotensored over 𝒞\mathcal{C}, and let AA be an associative algebra in 𝒞\mathcal{C}. Then the forgetful functor G:ModA​(𝒞)⊗𝒞ℳ→ℳG:\mathrm{Mod}_{A}(\mathcal{C})\otimes_{\mathcal{C}}\mathcal{M}\to\mathcal{M} is conservative.

0NXQ

Proof. Observe that for any 𝒟\mathcal{D} tensored over 𝒞\mathcal{C}, the pullback

Fun𝒞L⁡(ModA​(𝒞)⊗𝒞ℳ,𝒟)→Fun𝒞L⁡(ℳ,𝒟)\Fun^{\rm L}_{\mathcal{C}}(\mathrm{Mod}_{A}(\mathcal{C})\otimes_{\mathcal{C}}\mathcal{M},\mathcal{D})\rightarrow\Fun^{\rm L}_{\mathcal{C}}(\mathcal{M},\mathcal{D})

induced by the induction F:ℳ→ModA⊗𝒞ℳF:\mathcal{M}\to\mathrm{Mod}_{A}\otimes_{\mathcal{C}}\mathcal{M} is conservative. In other words, if a functor out of ModA×ℳ\mathrm{Mod}_{A}\times\mathcal{M} (which preserves colimits in each variable) is trivial when restricted to ℳ\mathcal{M}, then it is necessarily trivial.

Consequently, switching to opposite categories, we have that the corresponding functor

Fun𝒞R⁡(𝒟,ModA⊗ℳ)→Fun𝒞R⁡(𝒟,ℳ)\Fun^{\rm R}_{\mathcal{C}}(\mathcal{D},\mathrm{Mod}_{A}\otimes\mathcal{M})\rightarrow\Fun^{\rm R}_{\mathcal{C}}(\mathcal{D},\mathcal{M})

induced by the forgetful functor G:ModA⊗𝒞ℳ→ℳG:\mathrm{Mod}_{A}\otimes_{\mathcal{C}}\mathcal{M}\to\mathcal{M} is conservative.

Now we can apply Lemma 4.2 with ℳ′=ModA⊗ℳ\mathcal{M}^{\prime}=\mathrm{Mod}_{A}\otimes\mathcal{M} to obtain that GG is conservative. ∎

0NXR

Proof of Proposition 4.1. We first prove that ModA​(𝒞)⊗𝒞ℳ\mathrm{Mod}_{A}(\mathcal{C})\otimes_{\mathcal{C}}\mathcal{M} is equivalent to ModA​(ℳ)\mathrm{Mod}_{A}(\mathcal{M}) by the natural evaluation functor. Consider the adjunction

𝒞\textstyle{\mathcal{C}\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}F\scriptstyle{F}ModA​(𝒞)\textstyle{\mathrm{Mod}_{A}(\mathcal{C})\ignorespaces\ignorespaces\ignorespaces\ignorespaces}G\scriptstyle{G}

where F(−)=A⊗−F(-)=A\otimes- is the induction, and GG is the forgetful functor.

The above adjunction induces an adjunction

ℳ\textstyle{\mathcal{M}\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}F⊗id\scriptstyle{F\otimes{\rm id}}ModA​(𝒞)⊗𝒞ℳ\textstyle{\mathrm{Mod}_{A}(\mathcal{C})\otimes_{\mathcal{C}}\mathcal{M}\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}G⊗id\scriptstyle{G\otimes{\rm id}}ModT​(ℳ)\textstyle{\mathrm{Mod}_{T}(\mathcal{M})}

and thus a functor to modules over the monad T=(G⊗id)∘(F⊗id)T=(G\otimes{\rm id})\circ(F\otimes{\rm id}) acting on ℳ\mathcal{M}. The functor underlying TT is given by tensoring with AA, so we also have an equivalence ModT​(ℳ)≃ModA​(ℳ)\mathrm{Mod}_{T}(\mathcal{M})\simeq\mathrm{Mod}_{A}(\mathcal{M}).

By its universal characterization, the functor G⊗idG\otimes{\rm id} is colimit preserving. Note as well that GG and hence G⊗idG\otimes{\rm id} is also 𝒞\mathcal{C}-linear (or in other words, the adjunction satisfies an analogue of the projection formula). Thus it follows from Lemma 4.3 that G⊗idG\otimes{\rm id} is also conservative. Thus G⊗idG\otimes{\rm id} satisfies the monadic Barr-Beck conditions, and we obtain the desired equivalence ModA​(𝒞)⊗𝒞ℳ≃ModA​(ℳ)\mathrm{Mod}_{A}(\mathcal{C})\otimes_{\mathcal{C}}\mathcal{M}\simeq\mathrm{Mod}_{A}(\mathcal{M}).

Next, we can apply this to the instance where ℳ\mathcal{M} is the ∞\infty-category of left modules over another associative algebra A′A^{\prime} to conclude that there is a natural equivalence ModA​(𝒞)⊗𝒞ModA′​(𝒞)≃ModA​(ModA′​(𝒞))\mathrm{Mod}_{A}(\mathcal{C})\otimes_{\mathcal{C}}\mathrm{Mod}_{A^{\prime}}(\mathcal{C})\simeq\mathrm{Mod}_{A}(\mathrm{Mod}_{A^{\prime}}(\mathcal{C})). We now have a chain of adjunctions

𝒞\textstyle{\mathcal{C}\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}F′\scriptstyle{F^{\prime}}ModA′​(𝒞)\textstyle{\mathrm{Mod}_{A^{\prime}}(\mathcal{C})\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}F′′\scriptstyle{F^{\prime\prime}}G′\scriptstyle{G^{\prime}}ModA​(ModA′​(𝒞))\textstyle{\mathrm{Mod}_{A}(\mathrm{Mod}_{A^{\prime}}(\mathcal{C}))\ignorespaces\ignorespaces\ignorespaces\ignorespaces}G′′\scriptstyle{G^{\prime\prime}}

in which the composite G′∘G′′G^{\prime}\circ G^{\prime\prime} is colimit preserving and conservative, and hence satisfies the monadic Barr-Beck conditions.

Furthermore, the above adjunction naturally extends to a diagram in which the cycle of left adjoints (denoted by bowed arrows), and hence also the cycle of right adjoints (denoted by straight arrows), commute

𝒞\textstyle{\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\mathcal{C}\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}F\scriptstyle{F}F′′​F′\scriptstyle{F^{\prime\prime}F^{\prime}}ModA​(ModA′​(𝒞))\textstyle{\mathrm{Mod}_{A}(\mathrm{Mod}_{A^{\prime}}(\mathcal{C}))\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}G′​G′′\scriptstyle{G^{\prime}G^{\prime\prime}}f\scriptstyle{f}ModA⊗A′​(𝒞)\textstyle{\ignorespaces\ignorespaces\ignorespaces\ignorespaces\mathrm{Mod}_{A\otimes A^{\prime}}(\mathcal{C})\ignorespaces\ignorespaces\ignorespaces\ignorespaces}G\scriptstyle{G}g\scriptstyle{g}

Here F(−)=A1⊗A2⊗−F(-)=A_{1}\otimes A_{2}\otimes- is the induction, GG is the forgetful functor, ff is the natural functor factoring through ModA​(𝒞)⊗𝒞ModA′​(𝒞)\mathrm{Mod}_{A}(\mathcal{C})\otimes_{\mathcal{C}}\mathrm{Mod}_{A^{\prime}}(\mathcal{C}), and gg is its right adjoint. From this diagram, we obtain a morphism of monads

G′​G′′​F′′​F′\textstyle{G^{\prime}G^{\prime\prime}F^{\prime\prime}F^{\prime}\ignorespaces\ignorespaces\ignorespaces\ignorespaces}G′​G′′​g​f​F′′​F′≃G​F.\textstyle{G^{\prime}G^{\prime\prime}gfF^{\prime\prime}F^{\prime}\simeq GF.}

Now the underlying functors of the monads G​F​(−)GF(-) and G′​G′′​F′′​F′​(−)G^{\prime}G^{\prime\prime}F^{\prime\prime}F^{\prime}(-) are both equivalent to the tensor A⊗A′⊗(−)A\otimes A^{\prime}\otimes(-), so the above morphism of monads is an equivalence. Thus we obtain the promised equivalence ModA​(𝒞)⊗𝒞ModA′​(𝒞)≃ModA​(ModA′​(𝒞))≃ModA⊗A′​(𝒞)\mathrm{Mod}_{A}(\mathcal{C})\otimes_{\mathcal{C}}\mathrm{Mod}_{A^{\prime}}(\mathcal{C})\simeq\mathrm{Mod}_{A}(\mathrm{Mod}_{A^{\prime}}(\mathcal{C}))\simeq\mathrm{Mod}_{A\otimes A^{\prime}}(\mathcal{C}).

Finally, we show that the ∞\infty-category of left AA-modules ModA​(𝒞)\mathrm{Mod}_{A}(\mathcal{C}) is a dualizable 𝒞\mathcal{C}-module by directly exhibiting the ∞\infty-category of right AA-modules ModAop​(𝒞)\mathrm{Mod}_{A^{\rm op}}(\mathcal{C}) as its dual. The trace map is given by the two-sided bar construction

τ:ModA​(𝒞)⊗𝒞ModAop​(𝒞)→𝒞M,N↦M⊗AN\tau:\mathrm{Mod}_{A}(\mathcal{C})\otimes_{\mathcal{C}}\mathrm{Mod}_{A^{\rm op}}(\mathcal{C})\to\mathcal{C}\qquad M,N\mapsto M\otimes_{A}N

The unit map is given by the induction

u:𝒞→ModAop​(𝒞)⊗𝒞ModA​(𝒞)≃ModAop⊗A​(𝒞)c↦A⊗cu:\mathcal{C}\to\mathrm{Mod}_{A^{\rm op}}(\mathcal{C})\otimes_{\mathcal{C}}\mathrm{Mod}_{A}(\mathcal{C})\simeq\mathrm{Mod}_{A^{\rm op}\otimes A}(\mathcal{C})\qquad c\mapsto A\otimes c

where we regard A⊗cA\otimes c as an AA-bimodule.

One can verify directly that the composition

ModA​(𝒞)\textstyle{\mathrm{Mod}_{A}(\mathcal{C})\ignorespaces\ignorespaces\ignorespaces\ignorespaces}id⊗u\scriptstyle{{\rm id}\otimes u}ModA(𝒞)⊗𝒞ModAop(𝒞)⊗𝒞ModA(𝒞)\textstyle{\mathrm{Mod}_{A}(\mathcal{C})\otimes_{\mathcal{C}}\mathrm{Mod}_{A^{\rm op}}(\mathcal{C})\otimes_{\mathcal{C}}\mathrm{Mod}_{A}(\mathcal{C})\ignorespaces\ignorespaces\ignorespaces\ignorespaces}τ⊗id\scriptstyle{\tau\otimes{\rm id}}ModA​(𝒞)\textstyle{\mathrm{Mod}_{A}(\mathcal{C})}

is equivalent to the identity. First, (id⊗u)​(M)({\rm id}\otimes u)(M) is equivalent to A⊗MA\otimes M regarded as an A⊗Aop⊗AA\otimes A^{\rm op}\otimes A-module, and second, (τ⊗id)​(A⊗M)(\tau\otimes{\rm id})(A\otimes M) is equivalent to A⊗AM≃MA\otimes_{A}M\simeq M. ∎

4.1.2. Small stable categories

We have been working with the symmetric monoidal structure on the ∞\infty-category 𝒫​rL\mathcal{P}r^{\rm L} of presentable ∞\infty-categories as developed in [L4], [L5]. We will also need the tensor product of small stable idempotent complete ∞\infty-categories, in particular, the ∞\infty-categories of compact objects in presentable stable ∞\infty-categories.

Let s​t{st} be the full ∞\infty-subcategory of the ∞\infty-category of stable categories (with morphisms exact functors) consisting of those ∞\infty-categories that are idempotent complete. Recall that an ∞\infty-category 𝒞\mathcal{C} is idempotent complete if the essential image of the Yoneda embedding 𝒞→𝒫⁡(𝒞)\mathcal{C}\rightarrow\mathcal{P}(\mathcal{C}) is closed under retracts.

0NXS

Proposition 4.4. The ∞\infty-category s​t{st} carries a symmetric monoidal structure characterized by the property that for 𝒞1,𝒞2,𝒟∈s​t\mathcal{C}_{1},\mathcal{C}_{2},\mathcal{D}\in{st}, the ∞\infty-category of exact functors Funs​t⁡(𝒞1⊗𝒞2,𝒟)\Fun_{{st}}(\mathcal{C}_{1}\otimes\mathcal{C}_{2},\mathcal{D}) is equivalent to the full ∞\infty-subcategory of all functors 𝒞1×𝒞2→𝒟\mathcal{C}_{1}\times\mathcal{C}_{2}\to\mathcal{D} that preserve finite colimits in 𝒞1\mathcal{C}_{1} and 𝒞2\mathcal{C}_{2} separately. Furthermore, passing to the corresponding stable presentable ∞\infty-categories of Ind\operatorname{Ind}-objects is naturally a symmetric monoidal functor.

0NXT

Proof. For 𝒞1,𝒞2∈s​t\mathcal{C}_{1},\mathcal{C}_{2}\in{st}, we define their tensor product by

𝒞1⊗𝒞2=(Ind⁡(𝒞1)⊗Ind⁡(𝒞2))c\mathcal{C}_{1}\otimes\mathcal{C}_{2}=(\operatorname{Ind}(\mathcal{C}_{1})\otimes\operatorname{Ind}(\mathcal{C}_{2}))^{c}

where the tensor product of the right hand side is calculated in the ∞\infty-category 𝒫​rL\mathcal{P}r^{\rm L} of presentable ∞\infty-categories (with morphisms left adjoints), and the superscript c denotes the full ∞\infty-subcategory of compact objects of a presentable ∞\infty-category. Since Ind⁡𝒞1⊗Ind⁡𝒞2\operatorname{Ind}\mathcal{C}_{1}\otimes\operatorname{Ind}\mathcal{C}_{2} is idempotent complete and retracts of compact objects are compact, (Ind⁡𝒞1⊗Ind⁡𝒞2)c(\operatorname{Ind}\mathcal{C}_{1}\otimes\operatorname{Ind}\mathcal{C}_{2})^{c} is idempotent complete as well. Thus the tensor product 𝒞1⊗𝒞2\mathcal{C}_{1}\otimes\mathcal{C}_{2} is indeed an object of s​t{st}.

For 𝒞1,𝒞2,𝒟∈s​t\mathcal{C}_{1},\mathcal{C}_{2},\mathcal{D}\in{st}, let Fun′⁡(𝒞1×𝒞2,𝒟)\Fun^{\prime}(\mathcal{C}_{1}\times\mathcal{C}_{2},\mathcal{D}) be the full ∞\infty-subcategory of functors 𝒞1×𝒞2→𝒟\mathcal{C}_{1}\times\mathcal{C}_{2}\to\mathcal{D} that preserve finite colimits in 𝒞1\mathcal{C}_{1} and 𝒞2\mathcal{C}_{2} separately. We claim that For 𝒞1,𝒞2∈s​t\mathcal{C}_{1},\mathcal{C}_{2}\in{st}, the tensor product 𝒞1⊗𝒞2\mathcal{C}_{1}\otimes\mathcal{C}_{2} corepresents the functor Fun′⁡(𝒞1×𝒞2,−)\Fun^{\prime}(\mathcal{C}_{1}\times\mathcal{C}_{2},-) in the sense that for any 𝒟∈s​t\mathcal{D}\in{st}, there is a canonical equivalence

Fun′⁡(𝒞1×𝒞2,𝒟)≃Funs​t⁡(𝒞1⊗𝒞2,𝒟).\Fun^{\prime}(\mathcal{C}_{1}\times\mathcal{C}_{2},\mathcal{D})\simeq\Fun_{{st}}(\mathcal{C}_{1}\otimes\mathcal{C}_{2},\mathcal{D}).

As a consequence, the associativity and symmetry of the tensor product 𝒞1⊗𝒞2\mathcal{C}_{1}\otimes\mathcal{C}_{2} will immediately follow from the analogous properties of Fun′\Fun^{\prime}.

For 𝒞1,𝒞2,𝒟∈𝒫​rL\mathcal{C}_{1},\mathcal{C}_{2},\mathcal{D}\in\mathcal{P}r^{\rm L}, let FunL×L⁡(𝒞1×𝒞2,𝒟)\Fun^{L\times L}(\mathcal{C}_{1}\times\mathcal{C}_{2},\mathcal{D}) be the full ∞\infty-subcategory of functors 𝒞1×𝒞2→𝒟\mathcal{C}_{1}\times\mathcal{C}_{2}\to\mathcal{D} that preserve colimits in 𝒞1\mathcal{C}_{1} and 𝒞2\mathcal{C}_{2} separately. To prove the claim, observe that the inclusion 𝒟→Ind⁡𝒟\mathcal{D}\rightarrow\operatorname{Ind}\mathcal{D} induces a fully faithful functor

Fun′⁡(𝒞1×𝒞2,𝒟)→Fun′⁡(𝒞1×𝒞2,Ind⁡𝒟)≃FunL×L⁡(Ind⁡𝒞1×Ind⁡𝒞2,Ind⁡𝒟)\Fun^{\prime}(\mathcal{C}_{1}\times\mathcal{C}_{2},\mathcal{D})\rightarrow\Fun^{\prime}(\mathcal{C}_{1}\times\mathcal{C}_{2},\operatorname{Ind}\mathcal{D})\simeq\Fun^{L\times L}(\operatorname{Ind}\mathcal{C}_{1}\times\operatorname{Ind}\mathcal{C}_{2},\operatorname{Ind}\mathcal{D})

Its essential image consists of functors that preserve compact objects. By definition of the monoidal structure on the ∞\infty-category 𝒫​rL\mathcal{P}r^{\rm L} of presentable ∞\infty-categories, we have a further equivalence

FunL×L⁡(Ind⁡𝒞1×Ind⁡𝒞2,Ind⁡𝒟)≃FunL⁡(Ind⁡𝒞1⊗Ind⁡𝒞2,Ind⁡𝒟).\Fun^{L\times L}(\operatorname{Ind}\mathcal{C}_{1}\times\operatorname{Ind}\mathcal{C}_{2},\operatorname{Ind}\mathcal{D})\simeq\Fun^{\rm L}(\operatorname{Ind}\mathcal{C}_{1}\otimes\operatorname{Ind}\mathcal{C}_{2},\operatorname{Ind}\mathcal{D}).

Since the compact objects of Ind⁡𝒞1⊗Ind⁡𝒞2\operatorname{Ind}\mathcal{C}_{1}\otimes\operatorname{Ind}\mathcal{C}_{2} are generated by finite colimits of external products of compacts objects, we obtain an equivalence between Fun′⁡(𝒞1×𝒞2,𝒟)\Fun^{\prime}(\mathcal{C}_{1}\times\mathcal{C}_{2},\mathcal{D}) and the full ∞\infty-subcategory of FunL⁡(Ind⁡𝒞1⊗Ind⁡𝒞2,Ind⁡𝒟)\Fun^{\rm L}(\operatorname{Ind}\mathcal{C}_{1}\otimes\operatorname{Ind}\mathcal{C}_{2},\operatorname{Ind}\mathcal{D}) consisting of functors that preserve compact objects. In other words, we have the asserted equivalence that characterizes the tensor product

Fun′⁡(𝒞1×𝒞2,𝒟)≃Funs​t⁡((Ind⁡𝒞1⊗Ind⁡𝒞2)c,𝒟).\Fun^{\prime}(\mathcal{C}_{1}\times\mathcal{C}_{2},\mathcal{D})\simeq\Fun_{{st}}((\operatorname{Ind}\mathcal{C}_{1}\otimes\operatorname{Ind}\mathcal{C}_{2})^{c},\mathcal{D}).

Finally, the assertion that the functor Ind:s​t→𝒫​rL\operatorname{Ind}:{st}\to\mathcal{P}r^{\rm L} is symmetric monoidal is immediate from the constructions and the natural equivalence Ind⁡(𝒞c)≃𝒞\operatorname{Ind}(\mathcal{C}^{c})\simeq\mathcal{C}, for 𝒞∈𝒫​rL\mathcal{C}\in\mathcal{P}r^{\rm L}. ∎

0NXU

Remark 4.5. Given small stable idempotent complete ∞\infty-categories 𝒞1,𝒞2\mathcal{C}_{1},\mathcal{C}_{2}, by construction their tensor product 𝒞1⊗𝒞2\mathcal{C}_{1}\otimes\mathcal{C}_{2} is again a small stable idempotent complete ∞\infty-category. Though it is possible to consider other versions of a tensor product on small stable ∞\infty-categories that need not preserve idempotent complete ∞\infty-categories, our approach builds it in from the beginning.

4.2. Sheaves on fiber products

In this section, we study the ∞\infty-category of sheaves on the derived fiber product of perfect stacks. The main technical result is that it is equivalent to the tensor product of the ∞\infty-categories of sheaves on the factors (Theorem 4.7). The proof involves first showing that an analogous assertion holds for ∞\infty-categories of perfect complexes. The remainder of the section is devoted to collecting corollaries of the main techinical result.

Recall that for a perfect stack XX, compact (equivalently, perfect or dualizable) objects in the ∞\infty-category QC⁡(X)\qc(X) form a small stable idempotent complete ∞\infty-category QC⁡(X)c\qc(X)^{c}, and there is canonical equivalence QC⁡(X)≃Ind⁡QC⁡(X)c\qc(X)\simeq\operatorname{Ind}\qc(X)^{c}. Recall as well the tensor product of small stable idempotent complete ∞\infty-categories, and that the functor Ind\operatorname{Ind} is symmetric monoidal.

0NXV

Proposition 4.6. Let X1,X2X_{1},X_{2} be perfect stacks. Then external tensor product defines an equivalence

⊠:QC⁡(X1)c⊗QC⁡(X2)c→∼QC⁡(X1×X2)c\boxtimes:\qc(X_{1})^{c}\otimes\qc(X_{2})^{c}\stackrel{{\scriptstyle\sim}}{{\rightarrow}}\qc(X_{1}\times X_{2})^{c}

In other words, the ∞\infty-category of perfect complexes on the product is the (small stable idempotent complete) tensor product of the ∞\infty-categories of perfect complexes on the factors.

0NXW

Proof. Set X=X1×X2X=X_{1}\times X_{2}. By Proposition 3.24, we know that the external product takes compact objects to compact objects, and QC⁡(X)c\qc(X)^{c} is generated by external products.

Thus it suffices to verify that for Mi,Ni∈QC⁡(Xi)cM_{i},N_{i}\in\qc(X_{i})^{c} we have an equivalence

HomX⁡(M1⊠M2,N1⊠N2)≃HomX1⁡(M1,N1)⊗HomX2⁡(M2,N2).\Hom_{X}(M_{1}\boxtimes M_{2},N_{1}\boxtimes N_{2})\simeq\Hom_{X_{1}}(M_{1},N_{1})\otimes\Hom_{X_{2}}(M_{2},N_{2}).

Using the fact that each MiM_{i} is dualizable and p2p_{2} satisfies the projection formula (since it is perfect), we calculate

HomX⁡(p1∗​M1⊗p2∗​M2,p1∗​N1⊗p2∗​N2)\displaystyle\Hom_{X}(p_{1}^{*}M_{1}\otimes p_{2}^{*}M_{2},p_{1}^{*}N_{1}\otimes p_{2}^{*}N_{2}) ≃\displaystyle\simeq Γ⁡(X,p1∗​M1∨⊗p1∗​N1⊗p2∗​M2∨⊗p2∗​N2)\displaystyle\Gamma(X,p_{1}^{*}M_{1}^{\vee}\otimes p_{1}^{*}N_{1}\otimes p_{2}^{*}M_{2}^{\vee}\otimes p_{2}^{*}N_{2})
≃\displaystyle\simeq Γ⁡(X2,(p2)∗​(p1∗​ℋ​o​mX1​(M1,N2)⊗p2∗​ℋ​o​mX2​(M2,N2)))\displaystyle\Gamma(X_{2},(p_{2})_{*}(p_{1}^{*}{\mathcal{H}om}_{X_{1}}(M_{1},N_{2})\otimes p_{2}^{*}{\mathcal{H}om}_{X_{2}}(M_{2},N_{2})))
≃\displaystyle\simeq Γ⁡(X2,HomX1⁡(M1,N1)⊗ℋ​o​mX2​(M2,N2))\displaystyle\Gamma(X_{2},\Hom_{X_{1}}(M_{1},N_{1})\otimes{\mathcal{H}om}_{X_{2}}(M_{2},N_{2}))
≃\displaystyle\simeq HomX1⁡(M1,N1)⊗HomX2⁡(M2,N2)\displaystyle\Hom_{X_{1}}(M_{1},N_{1})\otimes\Hom_{X_{2}}(M_{2},N_{2})

∎

We now prove our main theorem which identifies ∞\infty-categories of sheaves on fiber products algebraically. The proof relies on the following consequence [L2, Corollary 5.5.3.4] of the ∞\infty-categorical adjoint functor theorem: there is a canonical equivalence

(𝒫​rL)o​p≃𝒫​rR({\mathcal{P}r}^{\rm L})^{op}\simeq{\mathcal{P}r}^{\rm R}

between the opposite of the ∞\infty-category 𝒫​rL\mathcal{P}r^{\rm L} of presentable ∞\infty-categories with morphisms left adjoints and the ∞\infty-category 𝒫​rR\mathcal{P}r^{\rm R} of presentable ∞\infty-categories with morphisms right adjoints. In other words, we can reverse diagrams of presentable ∞\infty-categories, in which the functors are all left adjoints, by passing to the corresponding right adjoints.

We will also use the fact that the calculation of small limits of presentable ∞\infty-categories is independent of context. Namely, by [L2, Proposition 5.5.3.13, Theorem 5.5.3.18], the forgetful functors from 𝒫​rL{\mathcal{P}r}^{\rm L}, 𝒫​rR{\mathcal{P}r}^{\rm R} to all ∞\infty-categories preserves small limits. In particular, given a small diagram of both left and right adjoints, the universal maps from the limit to the terms of the diagram are also both left and right adjoints.

0NXX

Theorem 4.7. Let X1X_{1}, X2X_{2}, YY be perfect stacks with maps p1:X1→Yp_{1}:X_{1}\to Y, p2:X2→Yp_{2}:X_{2}\to Y. Then there is a canonical equivalence

QC⁡(X1)⊗QC⁡(Y)QC⁡(X2)→∼QC⁡(X1×YX2).\qc(X_{1})\otimes_{\qc(Y)}\qc(X_{2})\stackrel{{\scriptstyle\sim}}{{\rightarrow}}\qc(X_{1}\times_{Y}X_{2}).
0NXY

Proof. To begin, consider the case Y=Spec⁡kY=\Spec k, so X1×YX2=X1×X2X_{1}\times_{Y}X_{2}=X_{1}\times X_{2}. By Proposition 4.6 and the fact that Ind:s​t→𝒫​rL\operatorname{Ind}:{st}\to\mathcal{P}r^{\rm L} is symmetric monoidal, the external product functor provides an equivalence ⊠:QC⁡(X1)⊗QC⁡(X2)→∼QC⁡(X1×X2)\boxtimes:\qc(X_{1})\otimes\qc(X_{2})\stackrel{{\scriptstyle\sim}}{{\rightarrow}}\qc(X_{1}\times X_{2}).

To begin the case of a general perfect stack YY, consider the augmented cosimplicial diagram

X1×YX2\textstyle{X_{1}\times_{Y}X_{2}\ignorespaces\ignorespaces\ignorespaces\ignorespaces}π\scriptstyle{\pi}X1×X2\textstyle{X_{1}\times X_{2}\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}X1×Y×X2\textstyle{X_{1}\times Y\times X_{2}\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}X1×Y×Y×X2×⋯\textstyle{X_{1}\times Y\times Y\times X_{2}\cdots}

with the obvious maps constructed from the given maps p1,p2p_{1},p_{2}.

Applying the (contravariant) functor QC\qc, we obtain an augmented simplicial ∞\infty-category

QC⁡(X1×YX2)\textstyle{\qc(X_{1}\times_{Y}X_{2})}QC⁡(X1×X2)\textstyle{\ignorespaces\ignorespaces\ignorespaces\ignorespaces\qc(X_{1}\times X_{2})}π∗\scriptstyle{\pi^{*}}QC⁡(X1×Y×X2)\textstyle{\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\qc(X_{1}\times Y\times X_{2})}⋯\textstyle{\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\cdots}

with structure maps given by pullbacks. By the absolute case of the theorem when Y=Spec⁡kY=\Spec k, if we forget the augmentation, we obtain the simplicial ∞\infty-category with simplices

QC(X1)⊗QC(Y)⊗⋯⊗QC(Y)⊗QC(X2)\qc(X_{1})\otimes\qc(Y)\otimes\cdots\otimes\qc(Y)\otimes\qc(X_{2})

and structure maps given by tensor contractions. This is precisely the two-sided bar construction [L4, 4.5] whose geometric realization, by definition [L5, 5], calculates the tensor product of QC⁡(Y)\qc(Y)-modules QC⁡(X1)⊗QC⁡(Y)QC⁡(X2)\qc(X_{1})\otimes_{\qc(Y)}\qc(X_{2}). Furthermore, the augmentation provides the natural map

QC⁡(X1×YX2)\textstyle{\qc(X_{1}\times_{Y}X_{2})}QC⁡(X1)⊗QC⁡(Y)QC⁡(X2)\textstyle{\ignorespaces\ignorespaces\ignorespaces\ignorespaces\qc(X_{1})\otimes_{\qc(Y)}\qc(X_{2})}π~∗\scriptstyle{\tilde{\pi}^{*}}

which we will prove is an equivalence.

The above geometric realization is a colimit in 𝒫​rL\mathcal{P}r^{\rm L}, and hence (as observed prior to the statement of the theorem) may be evaluated as a limit in the opposite category 𝒫​rR\mathcal{P}r^{\rm R}. Thus we find that QC⁡(X1)⊗QC⁡(Y)QC⁡(X2)\qc(X_{1})\otimes_{\qc(Y)}\qc(X_{2}) is also the totalization of the cosimplicial ∞\infty-category

QC⁡(X1×X2)\textstyle{\qc(X_{1}\times X_{2})\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}QC⁡(X1×Y×X2)\textstyle{\qc(X_{1}\times Y\times X_{2})\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}⋯\textstyle{\cdots}

with structure maps given by pushforwards. (We note for future reference that these structure maps are pushforwards along affine morphisms, hence are also colimit preserving, i.e., left adjoints.)

In particular, pushforward along the augmentation provides a natural functor

QC⁡(X1×YX2)\textstyle{\qc(X_{1}\times_{Y}X_{2})\ignorespaces\ignorespaces\ignorespaces\ignorespaces}π~∗\scriptstyle{\tilde{\pi}_{*}}QC⁡(X1)⊗QC⁡(Y)QC⁡(X2)\textstyle{\qc(X_{1})\otimes_{\qc(Y)}\qc(X_{2})}

which is an equivalence if and only if π~∗\tilde{\pi}^{*} is an equivalence.

To summarize some of the above structure, we have a diagram of commuting left (lower arrows) and right (upper arrows) adjoints

QC⁡(X1×YX2)\textstyle{\qc(X_{1}\times_{Y}X_{2})\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}π∗\scriptstyle{\pi_{*}}π~∗\scriptstyle{\tilde{\pi}_{*}}QC⁡(X1)⊗QC⁡(Y)QC⁡(X2)\textstyle{\qc(X_{1})\otimes_{\qc(Y)}\qc(X_{2})\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}τ∗\scriptstyle{\tau_{*}}π~∗\scriptstyle{\tilde{\pi}^{*}}QC⁡(X1×X2)\textstyle{\qc(X_{1}\times X_{2})\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}τ∗\scriptstyle{\tau^{*}}π∗\scriptstyle{\pi^{*}}

where τ∗\tau_{*} is the universal map from the totalization to the zero cosimplices, and likewise, τ∗\tau^{*} is the universal map from the zero simplices to the geometric realization. Thus we obtain a map of monads

Ta​l​g=τ∗​τ∗\textstyle{T_{alg}=\tau_{*}\tau^{*}\ignorespaces\ignorespaces\ignorespaces\ignorespaces}Tg​e​o​m=π∗​π∗\textstyle{T_{geom}=\pi_{*}\pi^{*}}

acting on QC⁡(X1×X2)\qc(X_{1}\times X_{2}).

The geometric pushforward π∗\pi_{*} is conservative and preserves colimits since π\pi is affine. Hence by the Barr-Beck theorem, we have a canonical equivalence

QC⁡(X1×YX2)≃ModTg​e​o​m⁡(QC⁡(X1×X2)).\qc(X_{1}\times_{Y}X_{2})\simeq\Mod_{T_{geom}}(\qc(X_{1}\times X_{2})).

We also claim that the universal map τ∗\tau_{*} is conservative and preserves colimits. For the first assertion, recall that τ∗\tau_{*} is nothing more than the forgetful map from the totalization to the zero cosimplices. Since the ∞\infty-categories involved are all stable, evaluating conservatism of a functor is equivalent to determining if nonzero objects are sent to zero. But an object sent to zero in the zeroth cosimplices is sent to zero in all cosimplices, and hence is equivalent to the zero object in the limit ∞\infty-category.

To see τ∗\tau_{*} preserves colimits, recall that the structure maps of our cosimplicial diagram are both right and left adjoints. It then follows (as observed prior to the statement of the theorem) that the totalization may be evaluated equivalently back in the category 𝒫​rL\mathcal{P}r^{\rm L}. In particular, the universal functor τ∗\tau_{*} is a morphism in 𝒫​rL\mathcal{P}r^{\rm L}, and hence a left adjoint and so preserves colimits.

We may now apply the Barr-Beck theorem, giving a canonical equivalence

QC⁡(X1)⊗QC⁡(Y)QC⁡(X2)≃ModTa​l​g⁡(QC⁡(X1×X2)).\qc(X_{1})\otimes_{\qc(Y)}\qc(X_{2})\simeq\Mod_{T_{alg}}(\qc(X_{1}\times X_{2})).

Thus it remains to show that the above morphism of monads is an equivalence. It is a straightforward diagram chase to check that the monad Ta​l​g=τ∗​τ∗T_{alg}=\tau_{*}\tau^{*} is nothing more than the composition π1∗π0∗\pi_{1}^{*}\pi_{0*} of the geometric functors associated to the initial cosimplicial maps

X1×X2\textstyle{X_{1}\times X_{2}\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}π0\scriptstyle{\pi_{0}}π1\scriptstyle{\pi_{1}}X1×Y×X2\textstyle{X_{1}\times Y\times X_{2}}

Thus by base change, it is equivalent to the monad Tg​e​o​m=π∗​π∗T_{geom}=\pi_{*}\pi^{*}. This concludes the proof of the theorem. ∎

0NXZ

Corollary 4.8. For π:X→Y\pi:X\to Y any map of perfect stacks, QC⁡(X)\qc(X) is self-dual as a QC⁡(Y)\qc(Y)-module.

0NY0

Proof. By Theorem 4.7, we have a canonical factorization QC⁡(X×YX)≃QC⁡(X)⊗QC⁡(Y)QC⁡(X)\qc(X\times_{Y}X)\simeq\qc(X)\otimes_{\qc(Y)}\qc(X). Using this identification, we can define the unit and trace by the correspondences u=Δ∗​π∗:QC⁡(Y)→QC⁡(X×YX)u=\Delta_{*}\pi^{*}:\qc(Y)\to\qc(X\times_{Y}X) and τ=π∗​Δ∗:QC⁡(X×YX)→QC⁡(Y)\tau=\pi_{*}\Delta^{*}:\qc(X\times_{Y}X)\to\qc(Y), where Δ:X→X×YX\Delta:X\to X\times_{Y}X is the relative diagonal. We need to check that the following composition is the identity:

QC⁡(X)\textstyle{\qc(X)\ignorespaces\ignorespaces\ignorespaces\ignorespaces}u⊗id\scriptstyle{u\otimes\operatorname{id}}QC(X)⊗QC⁡(Y)QC(X)⊗QC⁡(Y)QC(X)\textstyle{\qc(X)\otimes_{\qc(Y)}\qc(X)\otimes_{\qc(Y)}\qc(X)\ignorespaces\ignorespaces\ignorespaces\ignorespaces}id⊗τ\scriptstyle{\operatorname{id}\otimes\tau}QC⁡(X)\textstyle{\qc(X)}

The argument is a chase in the following diagram (with Cartesian square):

X\textstyle{\ignorespaces\ignorespaces\ignorespaces\ignorespaces X\ignorespaces\ignorespaces\ignorespaces\ignorespaces}Δ\scriptstyle{\Delta}Δ\scriptstyle{\Delta}X×YX\textstyle{X\times_{Y}X\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}π1\scriptstyle{\pi_{1}}id1×Δ23\scriptstyle{{\rm id}_{1}\times\Delta_{23}}X\textstyle{X}X\textstyle{X}X×YX\textstyle{\ignorespaces\ignorespaces\ignorespaces\ignorespaces X\times_{Y}X\ignorespaces\ignorespaces\ignorespaces\ignorespaces}π2\scriptstyle{\pi_{2}}Δ12×id3\scriptstyle{\Delta_{12}\times{\rm id}_{3}}X×YX×YX\textstyle{X\times_{Y}X\times_{Y}X}

Applying base change and identities for compositions, we have equivalences of functors

(id⊗τ)∘(u⊗id)\displaystyle(\operatorname{id}\otimes\tau)\circ(u\otimes\operatorname{id}) =\displaystyle= π1∗(id1×Δ23)∗(Δ12×id3)∗π2∗\displaystyle\pi_{1*}({\rm id}_{1}\times\Delta_{23})^{*}(\Delta_{12}\times{\rm id}_{3})_{*}\pi^{*}_{2}
≃\displaystyle\simeq π1∗Δ∗Δ∗π∗2\displaystyle\pi_{1*}\Delta_{*}\Delta^{*}\pi^{*}_{2}
≃\displaystyle\simeq idQC⁡(X)\displaystyle{\rm id}_{\qc(X)}

∎

0NY1

Remark 4.9. The above argument also shows that QC⁡(X)c\qc(X)^{c} is dualizable over perfect kk-modules QC⁡(Spec⁡k)c\qc(\Spec k)^{c} when XX is smooth and proper: smoothness is required for the unit in QC⁡(X×X)\qc(X\times X) to be perfect and properness is required for the trace to land in perfect kk-modules.

For a derived stack XX, let ℳ,ℳ′\mathcal{M},\mathcal{M}^{\prime} be stable presentable QC⁡(X)\qc(X)-modules. To reduce notation, we write FunX⁡(ℳ,ℳ′)\Fun_{X}(\mathcal{M},\mathcal{M}^{\prime}) for the stable presentable ∞\infty-category of QC⁡(X)\qc(X)-linear colimit preserving functors ℳ→ℳ′\mathcal{M}\to\mathcal{M}^{\prime}.

0NY2

Corollary 4.10. Let X,X′X,X^{\prime} and YY be perfect stacks with maps X→Y←X′X\rightarrow Y\leftarrow X^{\prime}. Then there is a natural equivalence of ∞\infty-categories

QC⁡(X×YX′)→∼FunY⁡(QC⁡(X),QC⁡(X′)).\qc(X\times_{Y}X^{\prime})\stackrel{{\scriptstyle\sim}}{{\rightarrow}}\Fun_{Y}(\qc(X),\qc({X^{\prime}})).

In other words, the ∞\infty-category of integral kernels is equivalent to the ∞\infty-category of functors.

0NY3

Proof. The statement is an immediate consequence of Theorem 4.7 and the fact that QC⁡(X)\qc(X) is self-dual (Corollary 4.8). It implies that internal hom of QC⁡(Y)\qc(Y)-modules out of QC⁡(X)\qc(X) is calculated by tensoring with QC⁡(X)∨≃QC⁡(X)\qc(X)^{\vee}\simeq\qc(X). ∎

0NY4

Remark 4.11. The equivalence of Corollary 4.10 is naturally monoidal in the following sense. For perfect stacks X,X′,X′′X,X^{\prime},X^{\prime\prime} mapping to a perfect stack YY, there is a convolution map

QC⁡(X×YX′)⊗QC⁡(X′×YX′′)→QC⁡(X×YX′′)\qc(X\times_{Y}X^{\prime})\otimes\qc(X^{\prime}\times_{Y}X^{\prime\prime})\to\qc(X\times_{Y}X^{\prime\prime})

given by pulling back and pushing forward with respect to the triple product X×YX′×YX′′X\times_{Y}X^{\prime}\times_{Y}X^{\prime\prime} (see Section 5.2). On the other hand, we have a composition map

FunY⁡(QC⁡(X),QC⁡(X′))⊗FunY⁡(QC⁡(X′),QC⁡(X′′))→FunY⁡(QC⁡(X),QC⁡(X′′)),\Fun_{Y}(\qc(X),\qc(X^{\prime}))\otimes\Fun_{Y}(\qc(X^{\prime}),\qc(X^{\prime\prime}))\to\Fun_{Y}(\qc(X),\qc(X^{\prime\prime})),

and the equivalence of the theorem intertwines these composition maps.

Finally, for a finite simplicial set Σ\Sigma, we would like to compare the formation of mapping stacks XΣ=Map⁡(Σ,X)X^{\Sigma}=\Map(\Sigma,X) (this can be viewed as the cotensoring of stacks over simplicial sets) with the formation of tensor products of ∞\infty-categories 𝒞⊗Σ\mathcal{C}\otimes\Sigma (the tensoring of symmetric monoidal ∞\infty-categories over simplicial sets). By the notation 𝒞⊗Σ\mathcal{C}\otimes\Sigma, we mean the geometric realization of the simplicial ∞\infty-category given by the constant assignment of 𝒞\mathcal{C} to each simplex of the simpicial set Σ\Sigma.

0NY5

Corollary 4.12. Let XX be a perfect stack, and let Σ\Sigma be a finite simplicial set. Then there is canonical equivalence

QC⁡(XΣ)≃QC⁡(X)⊗Σ.\qc(X^{\Sigma})\simeq\qc(X)\otimes\Sigma.

In other words, the ∞\infty-category of sheaves on the mapping stack is calculated as the tensor product of the ∞\infty-categories of sheaves on the simplices.

0NY6

Proof. We calculate QC⁡(XΣ)\qc(X^{\Sigma}) by induction on the simplices as an iterated fiber product of copies of XX over perfect stacks (by Proposition 3.24), and apply Theorem 4.7 at each stage. ∎

4.3. General base stacks

In this section, we extend Corollary 4.10 on integral transforms to a relative setting where the base is allowed to be an arbitrary derived stack with affine diagonal (not necessarily perfect). In order to describe integral transforms relative to such a base, we will utilize the simple behavior of ∞\infty-categories of sheaves under affine base change.

0NY7

Proposition 4.13. Let YY be a derived stack with affine diagonal, and let f:Spec⁡A→Yf:\Spec A\rightarrow Y be an affine over YY. Then ModA\Mod_{A} is a self-dual QC⁡(Y)\qc(Y)-module. In particular, for any X→YX\to Y there is a canonical equivalence

QC⁡(X×YSpec⁡A)≃QC⁡(X)⊗QC⁡(Y)ModA.\qc(X\times_{Y}\Spec A)\simeq\qc(X)\otimes_{\qc(Y)}\Mod_{A}.
0NY8

Proof. Since YY has affine diagonal, the map f:Spec⁡A→Yf:\Spec A\rightarrow Y is a relative affine, which implies that f∗f_{*} is colimit preserving and conservative. Hence f∗f_{*} satisfies the monadic Barr-Beck criteria [L4, Theorem 3.4.5], implying that the natural map ModA→Modf∗​f∗⁡(QC⁡(Y))\Mod_{A}\rightarrow\Mod_{f_{*}f^{*}}(\qc(Y)) is an equivalence. By the projection formula, the monad f∗​f∗​(−)f_{*}f^{*}(-) is equivalent to the functor f∗​A⊗(−)f_{*}A\otimes(-), with monad structure given by the algebra structure on AA. As a consequence, we see that ModA\Mod_{A} is equivalent to Modf∗​A⁡(QC⁡(Y))\Mod_{f_{*}A}(\qc(Y)). Now Proposition 4.1 gives that Modf∗​A⁡(QC⁡(Y))\Mod_{f_{*}A}(\qc(Y)) is self-dual as a QC⁡(Y)\qc(Y)-module.

Since the ∞\infty-category ModA\Mod_{A} is a dualizable QC⁡(Y)\qc(Y)-module it follows that the functor

(−)⊗QC⁡(Y)ModA{(-)\otimes_{\qc(Y)}\Mod_{A}}

commutes with limits of QC⁡(Y)\qc(Y)-module categories. Thus we have equivalences

QC⁡(X)⊗QC⁡(Y)ModA≃(limA′∈𝐴𝑓𝑓/XModA′)⊗QC⁡(Y)ModA≃limA′∈𝐴𝑓𝑓/X(ModA′⊗QC⁡(Y)ModA).\qc(X)\otimes_{\qc(Y)}\Mod_{A}\simeq(\lim_{A^{\prime}\in\it{Aff}/X}\Mod_{A^{\prime}})\otimes_{\qc(Y)}\Mod_{A}\simeq\lim_{A^{\prime}\in\it{Aff}/X}(\Mod_{A^{\prime}}\otimes_{\qc(Y)}\Mod_{A}).

Another application of Proposition 4.1 implies the following equivalence

limA′∈𝐴𝑓𝑓/X(ModA′⊗QC⁡(Y)ModA)≃limA′∈𝐴𝑓𝑓/X(Modf∗′​A′⊗f∗​A⁡(QC⁡(Y))).\lim_{A^{\prime}\in\it{Aff}/X}(\Mod_{A^{\prime}}\otimes_{\qc(Y)}\Mod_{A})\simeq\lim_{A^{\prime}\in\it{Aff}/X}(\Mod_{f^{\prime}_{*}A^{\prime}\otimes f_{*}A}(\qc(Y))).

Using that the map f′×f:Spec⁡A′×YSpec⁡A→Yf^{\prime}\times f:\Spec A^{\prime}\times_{Y}\Spec A\rightarrow Y is affine and that (f′×f)∗​(A′⊠A)≃f∗′​A′⊗f∗​A(f^{\prime}\times f)_{*}(A^{\prime}\boxtimes A)\simeq f^{\prime}_{*}A^{\prime}\otimes f_{*}A, we obtain that the above is further equivalent to

limA′∈𝐴𝑓𝑓/X(QC⁡(Spec⁡A′×YSpec⁡A))≃QC⁡(colimA′∈𝐴𝑓𝑓/X⁡(Spec⁡A′×YSpec⁡A))\lim_{A^{\prime}\in\it{Aff}/X}(\qc(\Spec A^{\prime}\times_{Y}\Spec A))\simeq\qc(\colim_{A^{\prime}\in\it{Aff}/X}(\Spec A^{\prime}\times_{Y}\Spec A))
≃QC⁡((colimA′∈𝐴𝑓𝑓/X⁡Spec⁡A′)×YSpec⁡A)≃QC⁡(X×YSpec⁡A)\simeq\qc((\colim_{A^{\prime}\in\it{Aff}/X}\Spec A^{\prime})\times_{Y}\Spec A)\simeq\qc(X\times_{Y}\Spec A)

Here we have used that QC⁡(−)\qc(-) sends all colimits to limits, and that (−)×YSpec⁡A(-)\times_{Y}\Spec A commutes with colimits (since it is the left adjoint to the mapping stack over YY). ∎

We now show that in the general setting where the base is an arbitrary derived stack with affine diagonal, functors continue to be given by integral kernels.

0NY9

Theorem 4.14. Let f:X→Yf:X\rightarrow Y be a perfect map of derived stacks with affine diagonal, and let g:X′→Yg:X^{\prime}\rightarrow Y be an arbitrary map of derived stacks. Then there is a natural map QC⁡(X×YX′)→FunY⁡(QC⁡(X),QC⁡(X′))\qc({X\times_{Y}X^{\prime}})\rightarrow\Fun_{Y}(\qc(X),\qc({X^{\prime}})) that is an equivalence of ∞\infty-categories.

0NYA

Proof. Consider the Cartesian diagram

X×YX′\textstyle{X\times_{Y}X^{\prime}\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}f~\scriptstyle{\tilde{f}}g~\scriptstyle{\tilde{g}}X′\textstyle{X^{\prime}\ignorespaces\ignorespaces\ignorespaces\ignorespaces}g\scriptstyle{g}X\textstyle{X\ignorespaces\ignorespaces\ignorespaces\ignorespaces}f\scriptstyle{f}Y\textstyle{Y}

We define a functor QC⁡(X×YX′)→Fun⁡(QC⁡(X),QC⁡(X′))\qc({X\times_{Y}X^{\prime}})\rightarrow\Fun(\qc(X),\qc({X^{\prime}})) by sending a quasi-coherent sheaf M∈QC⁡(X×YX′)M\in\qc(X\times_{Y}X^{\prime}) to the functor f~∗​(M⊗g~∗−)\tilde{f}_{*}(M\otimes\tilde{g}^{*}-). This is a colimit preserving functor, since f~\tilde{f} is perfect. Furthermore, using the projection formula for the map f~\tilde{f}, this functor naturally admits the extra structure of QC⁡(Y)\qc(Y)-linearity as follows

f~∗​(M⊗g~∗−)⊗g∗​V≃f~∗​(M⊗g~∗​(−)⊗f~∗​g∗​V)\tilde{f}_{*}(M\otimes\tilde{g}^{*}-)\otimes g^{*}V\simeq\tilde{f}_{*}(M\otimes\tilde{g}^{*}(-)\otimes\tilde{f}^{*}g^{*}V)
≃f~∗(M⊗g~∗(−)⊗g~∗f∗V)≃f~∗(M⊗g~∗(−⊗f∗V)),\simeq\tilde{f}_{*}(M\otimes\tilde{g}^{*}(-)\otimes\tilde{g}^{*}f^{*}V)\simeq\tilde{f}_{*}(M\otimes\tilde{g}^{*}(-\otimes f^{*}V)),

for any V∈QC⁡(Y)V\in\qc(Y). Therefore, we in fact obtain a functor QC⁡(X×YX′)→FunY⁡(QC⁡(X),QC⁡(X′))\qc(X\times_{Y}X^{\prime})\rightarrow\Fun_{Y}(\qc(X),\qc(X^{\prime})), and the rest of this proof will be devoted to showing it is an equivalence.

Recall that the ∞\infty-category of quasi-coherent sheaves on X′X^{\prime} is given by the limit

QC⁡(X′)≃lim𝐴𝑓𝑓/X′ModA.\qc({X^{\prime}})\simeq\lim_{\it{Aff}/X^{\prime}}\Mod_{A}.

By working locally in the target, this provides a description of the ∞\infty-category of QC⁡(Y)\qc(Y)-linear functors with values in QC⁡(X′)\qc({X^{\prime}}) as the limit

FunY⁡(QC⁡(X),QC⁡(X′))≃FunY⁡(QC⁡(X),lim𝐴𝑓𝑓/X′ModA)≃lim𝐴𝑓𝑓/X′FunY⁡(QC⁡(X),ModA)\Fun_{Y}(\qc(X),\qc({X^{\prime}}))\simeq\Fun_{Y}(\qc(X),\lim_{\it{Aff}/X^{\prime}}\Mod_{A})\simeq\lim_{\it{Aff}/X^{\prime}}\Fun_{Y}(\qc(X),\Mod_{A})

Likewise, we have a description of the ∞\infty-category of quasi-coherent sheaves on the fiber product X×YX′X\times_{Y}X^{\prime} as a limit

QC⁡(X×YX′)≃lim𝐴𝑓𝑓/X′QC⁡(X×YSpec⁡A).\qc({X\times_{Y}X^{\prime}})\simeq\lim_{\it{Aff}/X^{\prime}}\qc({X\times_{Y}\Spec A}).

This follows from the fact that the functor QC⁡(−)\qc(-) takes colimits to limits, and the fiber product functor X×Y(−)X\times_{Y}(-) commutes with all colimits (because it has a right adjoint).

Now one can analyze the functor QC⁡(X×YX′)→FunY⁡(QC⁡(X),QC⁡(X′))\qc({X\times_{Y}X^{\prime}})\rightarrow\Fun_{Y}(\qc(X),\qc(X^{\prime})) by considering the terms in the above two limits. That is, to prove the theorem, it suffices to prove it locally in the target X′X^{\prime}: for any Spec⁡A→X′\Spec A\rightarrow X^{\prime}, we must show that the functor

QC⁡(X×YSpec⁡A)→FunY⁡(QC⁡(X),ModA)\qc({X\times_{Y}\Spec A})\rightarrow\Fun_{Y}(\qc(X),\Mod_{A})

is an equivalence.

We will prove this in two steps. First, we will deal with case that the base YY is affine. Afterward, we will use this case to deal with a general base YY.

So assume for the time being that Y=Spec⁡BY=\Spec B. Then XX is a perfect stack over BB, and by Corollary 4.8, QC⁡(X)\qc(X) is a self-dual ModB\Mod_{B}-module. Thus we have equivalences

FunB⁡(QC⁡(X),ModA)≃FunB⁡(ModB,QC⁡(X)∨⊗BModA)≃QC⁡(X)⊗BModA.\Fun_{B}(\qc(X),\Mod_{A})\simeq\Fun_{B}(\Mod_{B},\qc(X)^{\vee}\otimes_{B}\Mod_{A})\simeq\qc(X)\otimes_{B}\Mod_{A}.

By Proposition 4.13 we know that the functor QC⁡(−)\qc(-) takes affine base change to tensor product of ∞\infty-categories. Therefore we have an equivalence

QC⁡(X×BSpec⁡A)≃QC⁡(X)⊗BModA.\qc({X\times_{B}\Spec A})\simeq\qc(X)\otimes_{B}\Mod_{A}.

Putting together the above equivalences, we conclude that we have equivalences

QC⁡(X×BSpec⁡A)≃QC⁡(X)⊗BModA≃FunB⁡(QC⁡(X),ModA),\qc({X\times_{B}\Spec A})\simeq\qc(X)\otimes_{B}\Mod_{A}\simeq\Fun_{B}(\qc(X),\Mod_{A}),

This proves the theorem when YY is affine.

Working locally in the base YY, we will now use the above discussion to prove the theorem in general. As above, since QC⁡(−)\qc(-) and fiber products behave well with respect to colimits we can calculate the ∞\infty-category of quasi-coherent sheaves on the fiber product X×YSpec⁡AX\times_{Y}\Spec A as the limit

QC(X×YSpecA)≃limB∈𝐴𝑓𝑓/YQC(X×YSpecB×YSpecA).\qc(X\times_{Y}\Spec A)\simeq\lim_{B\in\it{Aff}/Y}\qc(X\times_{Y}\Spec B\times_{Y}\Spec A).

To calculate the ∞\infty-category of functors, we use the following: by Proposition 4.13, the ∞\infty-category ModA\Mod_{A} is a dualizable QC⁡(Y)\qc(Y)-module and hence the functor ModA⊗QC⁡(Y)(−)\Mod_{A}\otimes_{\qc(Y)}(-) commutes with all limits. Therefore we have an equivalence

ModA≃ModA⊗QC⁡(Y)(limB∈𝐴𝑓𝑓/YModB)≃limB∈𝐴𝑓𝑓/YModA⊗QC⁡(Y)ModB,\Mod_{A}\simeq\Mod_{A}\otimes_{\qc(Y)}(\lim_{B\in\it{Aff}/Y}\Mod_{B})\simeq\lim_{B\in\it{Aff}/Y}\Mod_{A}\otimes_{\qc(Y)}\Mod_{B},

and so in particular we obtain equivalences

FunY⁡(QC⁡(X),ModA)≃FunY⁡(QC⁡(X),limB∈𝐴𝑓𝑓/YModA⊗QC⁡(Y)ModB)\Fun_{Y}(\qc(X),\Mod_{A})\simeq\Fun_{Y}(\qc({X}),\lim_{B\in\it{Aff}/Y}\Mod_{A}\otimes_{\qc(Y)}\Mod_{B})
≃limB∈𝐴𝑓𝑓/YFunY⁡(QC⁡(X),ModA⊗QC⁡(Y)ModB).\simeq\lim_{B\in\it{Aff}/Y}\Fun_{Y}(\qc({X}),\Mod_{A}\otimes_{\qc(Y)}\Mod_{B}).

By the adjunction between induction and restriction, and a repeated application of Proposition 4.13, we also have equivalences

FunY⁡(QC⁡(X),ModA⊗QC⁡(Y)ModB)≃FunB⁡(QC⁡(X)⊗QC⁡(Y)ModB,ModA⊗QC⁡(Y)ModB).\Fun_{Y}(\qc({X}),\Mod_{A}\otimes_{\qc(Y)}\Mod_{B})\simeq\Fun_{B}(\qc(X)\otimes_{\qc(Y)}\Mod_{B},\Mod_{A}\otimes_{\qc(Y)}\Mod_{B}).
≃FunB⁡(QC⁡(X×YSpec⁡B),QC⁡(Spec⁡A×YSpec⁡B))\simeq\Fun_{B}(\qc({X\times_{Y}\Spec B}),\qc({\Spec A\times_{Y}\Spec B}))

Finally, by the above discussion and the affine case of the theorem with base Spec⁡B\Spec B, we obtain the following chain of equivalences

QC(X×YSpecA)≃limB∈𝐴𝑓𝑓/YQC(X×YSpecB×YSpecA)\qc({X\times_{Y}\Spec A})\simeq\lim_{B\in\it{Aff}/Y}\qc({X\times_{Y}\Spec B\times_{Y}\Spec A})
≃limB∈𝐴𝑓𝑓/YFunB⁡(QC⁡(X×YB),QC⁡(A×YSpec⁡B))≃FunY⁡(QC⁡(X),ModA).\simeq\lim_{B\in\it{Aff}/Y}\Fun_{B}(\qc({X\times_{Y}B}),\qc({A\times_{Y}\Spec B}))\simeq\Fun_{Y}(\qc(X),\Mod_{A}).

Since we previously reduced the theorem to the case when X′=Spec⁡AX^{\prime}=\Spec A, this completes the proof. ∎

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

David Ben-Zvi, John Francis, David Nadler

Original source: arXiv:0805.0157v5