Introduction
Statement of main result
Statement of main result. Kashiwara’s Lemma for D \mathcal{D} D -modules asserts that, for a closed embedding of smooth schemes Z ↪ X Z \hookrightarrow X Z ↪ X over an algebraically closed characteristic 0 field k k k , the pushforward i + : D Z –mod → D X –mod i_+ : \mathcal{D}_Z\text{--mod} \to\mathcal{D}_X\text{--mod} i + : D Z –mod → D X –mod 1 is fully faithful, and identifies its essential image as D X –mod Z \mathcal{D}_X\text{--mod}_Z D X –mod Z , the full subcategory of D X –mod \mathcal{D}_X\text{--mod} D X –mod spanned by D \mathcal{D} D -modules set-theoretically supported on Z Z Z (cf. [HTT08, Theorem 1.6.1 and Corollary 1.6.2]). Since the category D X –mod Z \mathcal{D}_X\text{--mod}_Z D X –mod Z is insensitive to the scheme structure, Kashiwara’s Lemma in particular shows that the derived category of D \mathcal{D} D -modules is invariant under nilthickening. In [BZN12], Ben-Zvi–Nadler demonstrate that the (2-periodicized) category of D \mathcal{D} D -modules on X X X can be recovered, using Koszul duality and the Hochschild–Kostant–Rosenberg (HKR) isomorphism, as the category of cyclic sheaves on the derived loop stack of X X X 2 —more precisely, they establish an equivalence
IndCoh ( L X ) t T a t e ≃ D X –mod Z / 2 , \operatorname{IndCoh}(\mathcal{L}X)^{t\mathrm{Tate}} \simeq\mathcal{D}_X\text{--mod}_{\mathbb{Z}/2}, IndCoh ( L X ) t Tate ≃ D X –mod Z /2 ,
where D X –mod Z / 2 \mathcal{D}_X\text{--mod}_{\mathbb{Z}/2} D X –mod Z /2 denotes the 2-periodicization of D X –mod . W h e n \mathcal{D}_X\text{--mod}. When D X –mod . W h e n X$ is a geometric stack, rather than a scheme, the natural HKR morphism T X [ − 1 ] → L X T_X[-1] \to\mathcal{L}X T X [ − 1 ] → L X furnishes an isomorphism only upon taking the formal completion, along the zero section X ↪ T X [ − 1 ] X \hookrightarrow T_X[-1] X ↪ T X [ − 1 ] and along the constant loops X ↪ L X X \hookrightarrow\mathcal{L}X X ↪ L X , respectively, on each side: T ^ X [ − 1 ] → ∼ L ^ X \widehat{T}_X[-1] \xrightarrow{\sim} \widehat{\mathcal{L}}_X T X [ − 1 ] ∼ L X . In the same paper, Ben-Zvi–Nadler show ([BZN12, Theorem 6.13]) that for a smooth geometric stack X X X , the (2-periodicized) category of D \mathcal{D} D -modules on X X X can be recovered, using this equivalence of completions, as the category of cyclic sheaves on L ^ X \widehat{\mathcal{L}}X L X .
Thus, the category of cyclic sheaves IndCoh ( L X ) t T a t e \operatorname{IndCoh}(\mathcal{L}X)^{t\mathrm{Tate}} IndCoh ( L X ) t Tate on the full loop space $\mathcal{L}X is a distinct entity—a kind of enlargement of the (2-periodicized) category of D \mathcal{D} D -modules on X X X .3 This category, dubbed the category of potent D \mathcal{D} D -modules on X X X , and denoted by D p o t ( X ) \mathcal{D}^{\mathrm{pot}}(X) D pot ( X ) , is a central object in the program of potent categorical representation theory proposed in the recent ICM address of Ben-Zvi–Nadler ([BZN25]). In the philosophy expounded therein, and explained below in § 1.2 , D p o t ( X ) \mathcal{D}^{\mathrm{pot}}(X) D pot ( X ) is a decompletion of the ordinary 2-periodicized category of D \mathcal{D} D -modules.
For a map of stacks f : X → Y f : X \to Y f : X → Y , there is an induced S 1 S^1 S 1 -equivariant map of loop stacks, L f : L X → L Y \mathcal{L}f : \mathcal{L}X \to\mathcal{L}Y L f : L X → L Y . If X X X and Y Y Y are reasonable, one can show that the ! ! ! -pullback L f ! : IndCoh ( L Y ) → IndCoh ( L X ) \mathcal{L}f^! : \operatorname{IndCoh}(\mathcal{L}Y) \to\operatorname{IndCoh}(\mathcal{L}X) L f ! : IndCoh ( L Y ) → IndCoh ( L X ) induces a map on regularized Tate invariants, giving a functor f ! : D p o t ( Y ) → D p o t ( X ) f^! : \mathcal{D}^{\mathrm{pot}}(Y) \to\mathcal{D}^{\mathrm{pot}}(X) f ! : D pot ( Y ) → D pot ( X ) . In forthcoming work [ES], we construct a six functor formalism extending the assignment
( X → f Y ) ⇝ ( D p o t ( Y ) → f ! D p o t ( X ) ) , (X \xrightarrow{f} Y) \leadsto(\mathcal{D}^{\mathrm{pot}}(Y) \xrightarrow{f^!} \mathcal{D}^{\mathrm{pot}}(X)), ( X f Y ) ⇝ ( D pot ( Y ) f ! D pot ( X )) ,
and compatible with the six functor formalism on the full subcategory of ordinary D \mathcal{D} D -modules. In particular, given a closed immersion of stacks, i : Z ↪ X i : Z \hookrightarrow X i : Z ↪ X , there is a well-defined pushforward, i ∗ : D p o t ( Z ) → D p o t ( X ) i_* : \mathcal{D}^{\mathrm{pot}}(Z) \to\mathcal{D}^{\mathrm{pot}}(X) i ∗ : D pot ( Z ) → D pot ( X ) , as well as an appropriate notion of objects in D p o t ( X ) \mathcal{D}^{\mathrm{pot}}(X) D pot ( X ) supported on Z Z Z . The aim of this paper is to show that, for a broad class of stacks, D p o t \mathcal{D}^{\mathrm{pot}} D pot satisfies the same invariance under nilthickening as its completed counterpart.
Theorem 1.1 (“Kashiwara’s Lemma for D p o t \mathcal{D}^{\mathrm{pot}} D pot ”, Theorem 7.17 ). Let X X X be a linearly scalloped stack almost of finite type and with affine diagonal and i : Z ↪ X i : Z \hookrightarrow X i : Z ↪ X a closed immersion almost of finite presentation. Then i ∗ : D p o t ( Z ) → D p o t ( X ) i_* : \mathcal{D}^{\mathrm{pot}}(Z) \to\mathcal{D}^{\mathrm{pot}}(X) i ∗ : D pot ( Z ) → D pot ( X ) is fully faithful and its essential image is D Z p o t ( X ) \mathcal{D}^{\mathrm{pot}}_Z(X) D Z pot ( X ) , the full subcategory of D p o t ( X ) \mathcal{D}^{\mathrm{pot}}(X) D pot ( X ) spanned by objects set-theoretically supported on Z Z Z .
Remark 1.2. The class of linearly scalloped stacks, defined in [KR24, Definition A.1], encompasses many of the most prominently featured finite-dimensional stacks in geometric representation theory. For example, any stack of the form Y / H Y/H Y / H , where H H H is a subgroup of a reductive group G G G and Y Y Y is an H H H -quasi-projective derived scheme, is a linearly scalloped stack. In particular, the Hecke stack B \ G / B B\backslash G/B B \ G / B is linearly scalloped. Roughly speaking, they are stacks built successively using Nisnevich squares out of global quotient stacks of the form Y / G Y/G Y / G for G G G a reductive group and Y Y Y a G G G -quasi-projective scheme.
Motivation
Explanation of the name “potent D \mathcal{D} D -modules”
The name “potent D \mathcal{D} D -module” is inspired by the following observation.5 Given a stack X X X , the unipotent loop space of X X X , denoted L u X \mathcal{L}^{u}X L u X , is defined to be the mapping stack Map ( B G a , X ) \operatorname{Map}(B\mathbb{G}_{a},X) Map ( B G a , X ) . Precomposition with the affinization map S 1 → B G a S^{1} \to B\mathbb{G}_{a} S 1 → B G a gives an inclusion L u X ↪ L X \mathcal{L}^{u}X \hookrightarrow\mathcal{L}X L u X ↪ L X . The inclusion L ^ X ↪ L X \widehat{\mathcal{L}}X \hookrightarrow\mathcal{L}X L X ↪ L X factors through this latter map, giving an inclusion L ^ X ↪ L u X \widehat{\mathcal{L}}X \hookrightarrow\mathcal{L}^{u}X L X ↪ L u X . By a theorem of [BZCHN] (recalled as Theorem 5.18 below), when X X X is a geometric stack, pushforward along this inclusion induces an equivalence
IndCoh ( L ^ X ) t T a t e ⟶ ∼ IndCoh ( L u X ) t T a t e . \operatorname{IndCoh}(\widehat{\mathcal{L}}X)^{t\mathrm{Tate}} \overset{\sim}{\longrightarrow} \operatorname{IndCoh}(\mathcal{L}^{u}X)^{t\mathrm{Tate}}. IndCoh ( L X ) t Tate ⟶ ∼ IndCoh ( L u X ) t Tate .
For a smooth stack X X X , the left-hand category is identified with D X - m o d Z / 2 \mathcal{D}_{X}\text{-}\mathrm{mod}_{\mathbb{Z}/2} D X - mod Z /2 , as explained in the previous section (and, for a general stack X X X , it is identified with the 2-periodicized category of (right) crystals, Crys ( X ) Z / 2 \operatorname{Crys}(X)_{\mathbb{Z}/2} Crys ( X ) Z /2 , by [3 ] ). The right-hand category, on the other hand, is obtained by taking the regularized Tate invariants t T a t e t\mathrm{Tate} t Tate of the category IndCoh ( L u X ) \operatorname{IndCoh}(\mathcal{L}^{u}X) IndCoh ( L u X ) , whose S 1 S^{1} S 1 -action factors through the affinization S 1 → B G a S^{1} \to B\mathbb{G}_{a} S 1 → B G a . In Proposition 3.31 below, it is shown that the automorphism of a compact object in IndCoh ( L u X ) \operatorname{IndCoh}(\mathcal{L}^{u}X) IndCoh ( L u X ) induced by its S 1 S^{1} S 1 -action is unipotent. Hence, D X - m o d Z / 2 \mathcal{D}_{X}\text{-}\mathrm{mod}_{\mathbb{Z}/2} D X - mod Z /2 is obtained as t T a t e t\mathrm{Tate} t Tate of a unipotent categorical S 1 S^{1} S 1 -action. In contrast, the automorphism of an object of IndCoh ( L X ) \operatorname{IndCoh}(\mathcal{L}X) IndCoh ( L X ) induced by its S 1 S^{1} S 1 -action is not necessarily unipotent—the loop rotation S 1 S^{1} S 1 -action on IndCoh ( L X ) \operatorname{IndCoh}(\mathcal{L}X) IndCoh ( L X ) is “potent.” As D p o t ( X ) \mathcal{D}^{\mathrm{pot}}(X) D pot ( X ) is obtained as t T a t e t\mathrm{Tate} t Tate of a category with “potent” categorical S 1 S^{1} S 1 -action, its objects are called potent D \mathcal{D} D -modules.
Decompletion and genuine equivariance
One way to motivate the invention of genuine equivariant spectra is the following. Given a compact Lie group G G G acting on a CW complex X X X , one might study G G G -equivariant vector bundles on X X X . In particular, one might want to consider the Grothendieck group of such, K 0 ( Vect G ( X ) ) K_{0}(\operatorname{Vect}_{G}(X)) K 0 ( Vect G ( X )) . In order to do so, it is natural to consider the complex K K K -theory of the Borel construction ∣ X / G ∣ |X/G| ∣ X / G ∣ —the so-called Borel G G G -equivariant K K K -theory of X X X —but this unfortunately gives the wrong object. Indeed, while K 0 ( Vect G ( ∗ ) ) K_{0}(\operatorname{Vect}_{G}(*)) K 0 ( Vect G ( ∗ )) is the representation ring of G G G , R ( G ) R(G) R ( G ) , K U 0 KU^{0} K U 0 of the classifying space B G BG B G is the completion of the representation ring R ( G ) R(G) R ( G ) at the augmentation ideal I G = ker ( R ( G ) → Z ) I_{G}=\ker(R(G)\to\mathbb{Z}) I G = ker ( R ( G ) → Z ) .
After some reflection, one realizes the issue is that K U KU K U , as a spectrum, is determined by descent from its value on CW complexes. Instead, one should consider an enriched object whose value is specified for objects like B G BG B G —orbifold test objects of the form G / H G/H G / H —by hand. The appropriate definition for such an object is that of a genuine equivariant spectrum. The genuine G G G -equivariant K K K -theory spectrum K U G KU_{G} K U G is such an object, and its value on the G G G -homogeneous space G / H G/H G / H is rigged precisely so that K U G ( G / H ) ≃ R ( H ) KU_{G}(G/H)\simeq R(H) K U G ( G / H ) ≃ R ( H ) . More generally, the Atiyah–Segal completion theorem [AS69] tells us that the completion of genuine equivariant K K K -theory at the augmentation ideal I G I_{G} I G is isomorphic to Borel equivariant K K K -theory.
The category of D \mathcal{D} D -modules (or crystals) on a stack suffers from the same kind of deficiency. Given a stack X X X , the category of crystals on X X X is obtained by descent from affines:
Crys ( X ) ≃ lim S → X Crys ( S ) . \operatorname{Crys}(X) \simeq\lim_{S\to X}\operatorname{Crys}(S). Crys ( X ) ≃ S → X lim Crys ( S ) .
The result is a category which sees only the formal completion of L X \mathcal{L}X L X at constant loops; indeed (cf. [3 BZN12] ), for a sufficiently nice stack, we have
Crys ( X ) Z / 2 ≃ IndCoh ( L X ^ ) t T a t e . \operatorname{Crys}(X)_{\mathbb{Z}/2} \simeq\operatorname{IndCoh}(\widehat{\mathcal{L}X})^{t\mathrm{Tate}}. Crys ( X ) Z /2 ≃ IndCoh ( L X ) t Tate .
Our perspective—the one expounded in [BZN25]—is that one should rather consider a category which accesses the full geometry of the loop space L X \mathcal{L}X L X . Namely, one should consider potent D \mathcal{D} D -modules on X X X .
The analogy between Borel vs. genuine equivariant K K K -theory on the one hand, and ordinary vs. potent D \mathcal{D} D -modules on the other, becomes more precise when X X X is the global quotient of a derived scheme U U U by a reductive group G C G_{\mathbb{C}} G C . Then L ( U / G C ) \mathcal{L}(U/G_{\mathbb{C}}) L ( U / G C ) admits a canonical map μ : L ( U / G C ) → G C / / G C \mu:\mathcal{L}(U/G_{\mathbb{C}})\to G_{\mathbb{C}}/\!/G_{\mathbb{C}} μ : L ( U / G C ) → G C / / G C to the GIT quotient G C / / G C G_{\mathbb{C}}/\!/G_{\mathbb{C}} G C / / G C . In this case, the unipotent loop space L u ( U / G C ) : = Map ( B G a , U / G C ) \mathcal{L}^{u}(U/G_{\mathbb{C}}):=\operatorname{Map}(B\mathbb{G}_{a},U/G_{\mathbb{C}}) L u ( U / G C ) := Map ( B G a , U / G C ) has an alternative expression as the formal completion of L ( U / G C ) \mathcal{L}(U/G_{\mathbb{C}}) L ( U / G C ) along the fiber of μ \mu μ over the conjugacy class of the identity [ 1 G C ] [1_{G_{\mathbb{C}}}] [ 1 G C ] . Since U / G C U/G_{\mathbb{C}} U / G C is a geometric stack, pullback along the canonical map L ^ ( U / G C ) ↪ L u ( U / G C ) \widehat{\mathcal{L}}(U/G_{\mathbb{C}})\hookrightarrow\mathcal{L}^{u}(U/G_{\mathbb{C}}) L ( U / G C ) ↪ L u ( U / G C ) induces an equivalence,
IndCoh ( L u ( U / G C ) ) t T a t e → ≃ IndCoh ( L ^ ( U / G C ) ) t T a t e , \operatorname{IndCoh}(\mathcal{L}^{u}(U/G_{\mathbb{C}}))^{t\mathrm{Tate}} \xrightarrow{\simeq} \operatorname{IndCoh}(\widehat{\mathcal{L}}(U/G_{\mathbb{C}}))^{t\mathrm{Tate}}, IndCoh ( L u ( U / G C ) ) t Tate ≃ IndCoh ( L ( U / G C ) ) t Tate ,
so we have, in a literal sense (cf. the proof of Lemma 4.12), that objects of Crys ( U / G C ) Z / 2 \operatorname{Crys}(U/G_{\mathbb{C}})_{\mathbb{Z}/2} Crys ( U / G C ) Z /2 live on the formal completion at [ 1 G C ] ∈ G C / / G C [1_{G_{\mathbb{C}}}]\in G_{\mathbb{C}}/\!/G_{\mathbb{C}} [ 1 G C ] ∈ G C / / G C . On the other hand, given a compact Lie group G G G , the complexification of the representation ring R ( G ) R(G) R ( G ) —which we identified with K U G 0 ( ∗ ) KU^{0}_{G}(\ast) K U G 0 ( ∗ ) above—is isomorphic to algebraic global functions on the GIT quotient G C / / G C G_{\mathbb{C}}/\!/G_{\mathbb{C}} G C / / G C of the complexification G C G_{\mathbb{C}} G C . Because K U 0 ( B G ) KU^{0}(BG) K U 0 ( B G ) is a formal completion of K G 0 ( ∗ ) K^{0}_{G}(\ast) K G 0 ( ∗ ) , we say that K G 0 ( ∗ ) K^{0}_{G}(\ast) K G 0 ( ∗ ) is a decompletion of K U 0 ( B G ) KU^{0}(BG) K U 0 ( B G ) . In this way, we see that the natural functor Crys ( U / G C ) Z / 2 → D p o t ( U / G C ) \operatorname{Crys}(U/G_{\mathbb{C}})_{\mathbb{Z}/2}\to\mathcal{D}^{\mathrm{pot}}(U/G_{\mathbb{C}}) Crys ( U / G C ) Z /2 → D pot ( U / G C ) exhibits D p o t ( U / G C ) \mathcal{D}^{\mathrm{pot}}(U/G_{\mathbb{C}}) D pot ( U / G C ) as a decompletion of Crys ( U / G C ) Z / 2 \operatorname{Crys}(U/G_{\mathbb{C}})_{\mathbb{Z}/2} Crys ( U / G C ) Z /2 along the map from genuine to Borel equivariant K K K -theory,
K U G 0 ( ∗ ) ≃ O ( G C / / G C ) → O ( G C / / G C ) I G ∧ ≃ K U 0 ( B G ) . KU^{0}_{G}(\ast)\simeq\mathcal{O}(G_{\mathbb{C}}/\!/G_{\mathbb{C}})\to\mathcal{O}(G_{\mathbb{C}}/\!/G_{\mathbb{C}})^{\wedge}_{I_{G}}\simeq KU^{0}(BG). K U G 0 ( ∗ ) ≃ O ( G C / / G C ) → O ( G C / / G C ) I G ∧ ≃ K U 0 ( B G ) .
Remark 1.3. In light of the preceding discussion, one way to understand Theorem 1.1 is as a genuine equivariant version of Kashiwara’s Lemma for D \mathcal{D} D -modules.
Remark 1.4. Points of G / / G G/\!/G G / / G are often called “equivariant parameters.” By working with D p o t ( X ) \mathcal{D}^{\mathrm{pot}}(X) D pot ( X ) , we remember information about the loop space at every equivariant parameter [ α ] ∈ G / / G [\alpha]\in G/\!/G [ α ] ∈ G / / G , rather than just [ 1 G ] [1_{G}] [ 1 G ] . These parameters appear naturally in many different contexts, such as in the study of 3d mirror symmetry and geometric Langlands, as explained in [BZN25, §§3, 4]. The proof of our main theorem, outlined below in § 1.3 , makes extensive use of the ability to consider the restriction of potent D \mathcal{D} D -modules to individual equivariant parameters.
Remark 1.5. Another sense in which D p o t ( U / G C ) \mathcal{D}^{\mathrm{pot}}(U/G_{\mathbb{C}}) D pot ( U / G C ) is a decompletion of Crys ( U / G C ) Z / 2 \operatorname{Crys}(U/G_{\mathbb{C}})_{\mathbb{Z}/2} Crys ( U / G C ) Z /2 is furnished by the observation that the functor Crys ( U / G C ) Z / 2 → D p o t ( U / G C ) \operatorname{Crys}(U/G_{\mathbb{C}})_{\mathbb{Z}/2}\to\mathcal{D}^{\mathrm{pot}}(U/G_{\mathbb{C}}) Crys ( U / G C ) Z /2 → D pot ( U / G C ) is obtained as the induced map on regularized Tate invariants given by pushforward along the “decompletion” L ^ ( U / G C ) ↪ L ( U / G C ) \widehat{\mathcal{L}}(U/G_{\mathbb{C}})\hookrightarrow\mathcal{L}(U/G_{\mathbb{C}}) L ( U / G C ) ↪ L ( U / G C ) of the formal loop space. Decompletions of the formal loop, however, are not unique. For example, L X \mathcal{L}X L X and T X [ − 1 ] T_{X}[-1] T X [ − 1 ] have equivalent formal completions T ^ X [ − 1 ] ≃ L X ^ \widehat{T}_{X}[-1]\simeq\widehat{\mathcal{L}X} T X [ − 1 ] ≃ L X . In this sense, each of L X \mathcal{L}X L X and T X [ − 1 ] T_{X}[-1] T X [ − 1 ] can be seen as particular decompletions of the formal stack L X ^ \widehat{\mathcal{L}X} L X .
Decompletion and genuine S 1 S^{1} S 1 -equivariance
We are actually performing two decompletions by considering potent D \mathcal{D} D -modules: In the previous section, we explained decompletion with respect to the equivariant parameters of an algebraic group action. In this section, we will explain how the passage from ordinary (homotopy) S 1 S^{1} S 1 -invariants (resp. Tate invariants) to regularized S 1 S^{1} S 1 -invariants (resp. t t t Tate invariants) is a decompletion of the action of the topological circle S 1 S^{1} S 1 . Let C \mathcal{C} C be a stable, presentable ∞ \infty ∞ -category with regular t t t -structure6 and equipped with an S 1 S^{1} S 1 -action. For example, we could take C = IndCoh ( X ) \mathcal{C}=\operatorname{IndCoh}(X) C = IndCoh ( X ) , where X X X is a finite type algebraic stack with affine diagonal and S 1 S^1 S 1 -action. Denote the regularization D : = R ( C S 1 ) : = Ind ( Coh ( C S 1 ) ) \mathcal{D}:=\mathcal{R}(\mathcal{C}^{S^1}):=\operatorname{Ind}(\operatorname{Coh}(\mathcal{C}^{S^1})) D := R ( C S 1 ) := Ind ( Coh ( C S 1 )) . There is a pair of adjoint functors
I : C ⇄ C S 1 : U , I:\mathcal{C}\rightleftarrows\mathcal{C}^{S^1}:U, I : C ⇄ C S 1 : U ,
where U U U is the forgetful functor (forgetting S 1 S^1 S 1 -equivariance) and I I I is the induction functor. If c ∈ C c\in\mathcal{C} c ∈ C is compact, then I ( c ) I(c) I ( c ) is also compact in C S 1 \mathcal{C}^{S^1} C S 1 , because U U U preserves colimits. Moreover, the objects I ( c ) I(c) I ( c ) for compact c ∈ C c\in\mathcal{C} c ∈ C generate C S 1 \mathcal{C}^{S^1} C S 1 , since U U U is conservative. Thus, this defines a fully faithful and continuous inclusion
i : C S 1 ↪ D . (1) i:\mathcal{C}^{S^1}\hookrightarrow\mathcal{D}. \tag*{(1)} i : C S 1 ↪ D . (1)
Now, all objects I ( c ) I(c) I ( c ) are u u u -torsion with respect to the k [ u ] k[u] k [ u ] -module structure on C S 1 \mathcal{C}^{S^1} C S 1 , where u u u is in cohomological degree 2. Indeed, the u u u -action is induced by tensoring an object c ∈ C S 1 c\in\mathcal{C}^{S^1} c ∈ C S 1 with the degree 2 map
u : k ⟶ k [ 2 ] u:k\longrightarrow k[2] u : k ⟶ k [ 2 ]
in Vect S 1 ≃ k [ u ] - mod u -tors \operatorname{Vect}^{S^1}\simeq k[u]\text{-}\operatorname{mod}_{u\text{-tors}} Vect S 1 ≃ k [ u ] - mod u -tors . Since forgetting the equivariance is compatible with the tensor action by Vect S 1 \operatorname{Vect}^{S^1} Vect S 1 and sends u u u to a nullhomotopic map in Vect \operatorname{Vect} Vect , we find that the u u u -action on U ( c ~ ) U(\widetilde{c}) U ( c ) is nullhomotopic for any c ~ ∈ C S 1 \widetilde{c}\in\mathcal{C}^{S^1} c ∈ C S 1 . Suppose now that c ~ = I ( c ) \widetilde{c}=I(c) c = I ( c ) for some c ∈ C c\in\mathcal{C} c ∈ C . By adjunction,
Hom C S 1 ( I ( c ) , I ( c ) [ 2 ] ) ≃ Hom C ( c , U I ( c ) [ 2 ] ) , \operatorname{Hom}_{\mathcal{C}^{S^1}}(I(c),I(c)[2])\simeq\operatorname{Hom}_{\mathcal{C}}(c,UI(c)[2]), Hom C S 1 ( I ( c ) , I ( c ) [ 2 ]) ≃ Hom C ( c , U I ( c ) [ 2 ]) ,
and the map u I ( c ) : I ( c ) → I ( c ) [ 2 ] u_{I(c)}:I(c)\to I(c)[2] u I ( c ) : I ( c ) → I ( c ) [ 2 ] corresponds to the composition
c ⟶ U I ( c ) → U ( u I ( c ) ) U ( I ( c ) ) [ 2 ] , c\longrightarrow UI(c)\xrightarrow{U(u_{I(c)})}U(I(c))[2], c ⟶ U I ( c ) U ( u I ( c ) ) U ( I ( c )) [ 2 ] ,
which we know is nullhomotopic. Hence, u I ( c ) u_{I(c)} u I ( c ) is nullhomotopic as well. Conversely, if c ~ ∈ C S 1 \widetilde{c}\in\mathcal{C}^{S^1} c ∈ C S 1 is an arbitrary S 1 S^1 S 1 -equivariant object, then
ker ( u : c ~ → c ~ [ 2 ] ) \operatorname{ker}(u:\widetilde{c}\to\widetilde{c}[2]) ker ( u : c → c [ 2 ])
is precisely I U ( c ~ ) IU(\widetilde{c}) I U ( c ) , up to a shift, and these generate the full subcategory of u u u -torsion objects of D \mathcal{D} D :
D u -tors : = ker ( D → D [ u − 1 ] ≃ C t T a t e ) . \mathcal{D}_{u\text{-tors}}:=\operatorname{ker}(\mathcal{D}\to\mathcal{D}[u^{-1}]\simeq\mathcal{C}^{t\mathrm{Tate}}). D u -tors := ker ( D → D [ u − 1 ] ≃ C t Tate ) .
Thus, we see that the image of the functor i i i of (1) is precisely the u u u -torsion subcategory. Moreover, the natural functor of [3 Proposition 4.3.2] is in fact right adjoint to i i i , and it exhibits D u -tors ≃ C S 1 \mathcal{D}_{u\text{-tors}}\simeq\mathcal{C}^{S^1} D u -tors ≃ C S 1 as a completion of D \mathcal{D} D . In other words, we are decompleting with respect to the equivariant parameter u u u , where C S 1 \mathcal{C}^{S^1} C S 1 is supported at u = 0 u=0 u = 0 and D \mathcal{D} D is its decompletion away from 0 0 0 .7
Nil-invariance and Tate fixed points
We note that the analogue of Theorem 1.1 is not true for IndCoh ( L ( − ) ) \operatorname{IndCoh}(\mathcal{L}(-)) IndCoh ( L ( − )) (in place of D p o t ( − ) \mathcal{D}^{\mathrm{pot}}(-) D pot ( − ) ), or even for regularized S 1 S^1 S 1 -invariants, Ind ( Coh ( L ( − ) ) S 1 ) \operatorname{Ind}(\operatorname{Coh}(\mathcal{L}(-))^{S^1}) Ind ( Coh ( L ( − ) ) S 1 ) . This furnishes an instance of a recurring pattern in algebraic geometry wherein the Tate fixed points operation produces objects which are “topological” in nature—that is, depending only on underlying reduced structures—from objects which are not. We recall the primary example of this pattern below.
If X X X is a smooth scheme, O ( L X ) \mathcal{O}(\mathcal{L}X) O ( L X ) 8 may be identified under the Hochschild–Kostant–Rosenberg isomorphism with the algebraic de Rham complex Ω X ∙ \Omega_X^\bullet Ω X ∙ . Moreover, this isomorphism intertwines the canonical S 1 S^1 S 1 -action on O ( L X ) \mathcal{O}(\mathcal{L}X) O ( L X ) given by loop rotation with the action of the de Rham differential on Ω X ∙ \Omega_X^\bullet Ω X ∙ , letting us further identify O ( L X ) S 1 \mathcal{O}(\mathcal{L}X)^{S^1} O ( L X ) S 1 with Ω X ∙ , c l \Omega_X^{\bullet,\mathrm{cl}} Ω X ∙ , cl , the complex of closed forms on X X X , and O ( L X ) t S 1 \mathcal{O}(\mathcal{L}X)^{tS^1} O ( L X ) t S 1 with the de Rham cohomology of X X X . But while both Ω X ∙ \Omega_X^\bullet Ω X ∙ and Ω X ∙ , c l \Omega_X^{\bullet,\mathrm{cl}} Ω X ∙ , cl are sensitive to nilthickenings of X X X , the de Rham cohomology of X X X is not, with H d R ∙ ( X ) H_{\mathrm{dR}}^\bullet(X) H dR ∙ ( X ) depending only on the underlying reduced scheme of X X X .
1.2.5. Extended example. In order to anchor the reader’s understanding of the text, we sketch a toy model illustrating the aspects of the theory found in the preceding discussion.
Example 1.7. Let G = G m G = \mathbb{G}_{\mathrm{m}} G = G m , and consider X = B G m X = B\mathbb{G}_{\mathrm{m}} X = B G m . Then L X ≃ G m × B G m \mathcal{L}X \simeq\mathbb{G}_{\mathrm{m}} \times B\mathbb{G}_{\mathrm{m}} L X ≃ G m × B G m , and L X ^ ≃ ( G m ) 1 × B G m \widehat{\mathcal{L}X} \simeq(\mathbb{G}_{\mathrm{m}})_1 \times B\mathbb{G}_{\mathrm{m}} L X ≃ ( G m ) 1 × B G m . Under these identifications, we see that the pushforward
QCoh ( L X ^ ) ↪ QCoh ( L X ) \operatorname{Q Coh}(\widehat{\mathcal{L}X}) \hookrightarrow\operatorname{Q Coh}(\mathcal{L}X) QCoh ( L X ) ↪ QCoh ( L X ) along the inclusion L X ^ ↪ L X \widehat{\mathcal{L}X} \hookrightarrow\mathcal{L}X L X ↪ L X corresponds to the inclusion of graded k [ t , t − 1 ] k[t,t^{-1}] k [ t , t − 1 ] -modules supported at 1 ∈ G m 1 \in\mathbb{G}_{\mathrm{m}} 1 ∈ G m into the category of all graded k [ t , t − 1 ] k[t,t^{-1}] k [ t , t − 1 ] -modules.
Applying Tate invariants to the small categories Coh ( L X ^ ) \operatorname{Coh}(\widehat{\mathcal{L}X}) Coh ( L X ) and Coh ( L X ) \operatorname{Coh}(\mathcal{L}X) Coh ( L X ) respectively, we obtain (cf. [BZCHN23, Example A.13])
Coh ( L X ) t S 1 ≃ Coh ( k [ t , t − 1 ] ) ⊗ k [ u ± 1 ] \operatorname{Coh}(\mathcal{L}X)^{tS^1} \simeq\operatorname{Coh}(k[t,t^{-1}]) \otimes k[u^{\pm1}] Coh ( L X ) t S 1 ≃ Coh ( k [ t , t − 1 ]) ⊗ k [ u ± 1 ] Coh ( L X ^ ) t S 1 ≃ Coh ( Spf k [ [ s ] ] ) ⊗ k [ u ± 1 ] , \operatorname{Coh}(\widehat{\mathcal{L}X})^{tS^1} \simeq\operatorname{Coh}(\operatorname{Spf} k[[s]]) \otimes k[u^{\pm1}], Coh ( L X ) t S 1 ≃ Coh ( Spf k [[ s ]]) ⊗ k [ u ± 1 ] , where s : = t − 1 s := t - 1 s := t − 1 .9 Thus, we see that we have an inclusion
Coh ( L X ^ ) t S 1 ↪ Coh ( L X ) t S 1 , \operatorname{Coh}(\widehat{\mathcal{L}X})^{tS^1} \hookrightarrow\operatorname{Coh}(\mathcal{L}X)^{tS^1}, Coh ( L X ) t S 1 ↪ Coh ( L X ) t S 1 , whose image is precisely those modules over k [ t , t − 1 ] k[t,t^{-1}] k [ t , t − 1 ] which are supported complete with respect to the ideal ( t − 1 ) (t-1) ( t − 1 ) , exhibiting a concrete instance of “decompletion,” in this instance at the parameter [ 1 G m ] ∈ G m / / G m = G m [1_{\mathbb{G}_{\mathrm{m}}}] \in\mathbb{G}_{\mathrm{m}} // \mathbb{G}_{\mathrm{m}} = \mathbb{G}_{\mathrm{m}} [ 1 G m ] ∈ G m // G m = G m .
As explained in [BZCHN23], the category Coh ( L X ) \operatorname{Coh}(\mathcal{L}X) Coh ( L X ) admits a direct sum decomposition parametrized by the character lattice X ∗ ( G m ) = Z X^*(\mathbb{G}_{\mathrm{m}}) = \mathbb{Z} X ∗ ( G m ) = Z :
Coh ( L ( B G m ) ) ≃ ⨁ λ ∈ Z Coh ( G m ) . \operatorname{Coh}(\mathcal{L}(B\mathbb{G}_{\mathrm{m}})) \simeq\bigoplus_{\lambda\in\mathbb{Z}} \operatorname{Coh}(\mathbb{G}_{\mathrm{m}}). Coh ( L ( B G m )) ≃ λ ∈ Z ⨁ Coh ( G m ) . The S 1 S^1 S 1 -action on this category is a natural transformation of the identity whose component on an object of the λ \lambda λ -summand Coh ( B G m ) \operatorname{Coh}(B\mathbb{G}_{\mathrm{m}}) Coh ( B G m ) is given by multiplication by t λ ∈ k [ t , t − 1 ] t^\lambda\in k[t,t^{-1}] t λ ∈ k [ t , t − 1 ] at the point t ∈ G m t \in\mathbb{G}_{\mathrm{m}} t ∈ G m . Thus, one can think of points in G m \mathbb{G}_{\mathrm{m}} G m as equivariant parameters, and taking Coh ( L X ^ ) t S 1 \operatorname{Coh}(\widehat{\mathcal{L}X})^{tS^1} Coh ( L X ) t S 1 corresponds to completing at the equivariant parameter 1 ∈ G m 1 \in\mathbb{G}_{\mathrm{m}} 1 ∈ G m . If we were instead to consider a torus T T T of rank r r r and X = B T X = BT X = B T , then we would be completing at the ideal ( s 1 , … , s r ) (s_1,\ldots,s_r) ( s 1 , … , s r ) , where the s i ∈ O ( T ) s_i \in\mathcal{O}(T) s i ∈ O ( T ) are equivariant parameters.
The inclusion L X ^ ↪ L X \widehat{\mathcal{L}X} \hookrightarrow\mathcal{L}X L X ↪ L X induces a pullback functor
D p o t ( B G m ) : = Ind ( Coh ( L B G m ) t S 1 ) ⟶ Ind ( Coh ( L B G m ^ ) t S 1 ) ≃ D B G m - mod Z / 2 . \mathcal{D}^{\mathrm{pot}}(B\mathbb{G}_{\mathrm{m}}) := \operatorname{Ind}(\operatorname{Coh}(\mathcal{L}B\mathbb{G}_{\mathrm{m}})^{tS^1}) \longrightarrow\operatorname{Ind}(\operatorname{Coh}(\widehat{\mathcal{L}B\mathbb{G}_{\mathrm{m}}})^{tS^1}) \simeq\mathcal{D}_{B\mathbb{G}_{\mathrm{m}}\text{-}\operatorname{mod}_{\mathbb{Z}/2}.} D pot ( B G m ) := Ind ( Coh ( L B G m ) t S 1 ) ⟶ Ind ( Coh ( L B G m ) t S 1 ) ≃ D B G m - mod Z /2 . Thus, taking endomorphisms of the unit in both categories induces a map
k [ t ± 1 , u ± 1 ] ⟶ k [ [ s ] ] [ u ± 1 ] , k[t^{\pm1},u^{\pm1}] \longrightarrow k[[s]][u^{\pm1}], k [ t ± 1 , u ± 1 ] ⟶ k [[ s ]] [ u ± 1 ] , which exhibits k [ t ± 1 , u ± 1 ] k[t^{\pm1},u^{\pm1}] k [ t ± 1 , u ± 1 ] as a decompletion of k [ [ s ] ] [ u ± 1 ] k[[s]][u^{\pm1}] k [[ s ]] [ u ± 1 ] . This is analogous to, for example, viewing the genuine U ( 1 ) U(1) U ( 1 ) -equivariant K K K -theory of a point as a decompletion of the complex K K K -theory of the classifying space B U ( 1 ) ≃ C P ∞ BU(1) \simeq\mathbb{C}P^\infty B U ( 1 ) ≃ C P ∞ .
Since the functor Ind \operatorname{Ind} Ind preserves colimits,
Ind ( Coh ( L X ) t S 1 ) ≃ Ind ( Coh ( L X ) S 1 ) ⊗ k [ u ] k [ u ± 1 ] . \operatorname{Ind}(\operatorname{Coh}(\mathcal{L}X)^{tS^1}) \simeq\operatorname{Ind}(\operatorname{Coh}(\mathcal{L}X)^{S^1}) \otimes_{k[u]} k[u^{\pm1}]. Ind ( Coh ( L X ) t S 1 ) ≃ Ind ( Coh ( L X ) S 1 ) ⊗ k [ u ] k [ u ± 1 ] . In particular, we can consider
Ind ( Coh ( L X ) S 1 ) ≃ k [ t , t − 1 , u ] - mod \operatorname{Ind}(\operatorname{Coh}(\mathcal{L}X)^{S^1}) \simeq k[t,t^{-1},u]\text{-}\operatorname{mod} Ind ( Coh ( L X ) S 1 ) ≃ k [ t , t − 1 , u ] - mod and
IndCoh ( L X ) S 1 ≃ k [ t , t − 1 , u ] - mod u -tors , \operatorname{IndCoh}(\mathcal{L}X)^{S^1} \simeq k[t,t^{-1},u]\text{-}\operatorname{mod}_{u\text{-tors}}, IndCoh ( L X ) S 1 ≃ k [ t , t − 1 , u ] - mod u -tors ,
before inverting u u u . Moreover, by Section 1.2.3, there are adjoint functors
k [ t , t − 1 , u ] -mod u -tors ⇄ k [ t , t − 1 , u ] -mod , k[t,t^{-1},u]\text{-mod}_{u\text{-tors}} \rightleftarrows k[t,t^{-1},u]\text{-mod}, k [ t , t − 1 , u ] -mod u -tors ⇄ k [ t , t − 1 , u ] -mod ,
which exhibit k [ t , t − 1 , u ] -mod k[t,t^{-1},u]\text{-mod} k [ t , t − 1 , u ] -mod as the decompletion of k [ t , t − 1 , u ] -mod u -tors k[t,t^{-1},u]\text{-mod}_{u\text{-tors}} k [ t , t − 1 , u ] -mod u -tors away from u = 0 u=0 u = 0 .
Outline of proof
Throughout this outline, we assume that all stacks and schemes are of finite type and have affine diagonal.
Initial reduction
Any closed immersion of stacks Z ↪ X Z \hookrightarrow X Z ↪ X , factors through the formal completion as
Z → i X ^ Z → i ^ X . Z \xrightarrow{i} \widehat{X}_Z \xrightarrow{\widehat{i}} X. Z i X Z i X .
Our proof strategy starts with the observation that, as for IndCoh and QCoh, pushforward along i ^ \widehat{i} i induces an equivalence,
i ^ ∗ : D p o t ( X ^ Z ) → ∼ D Z p o t ( X ) , \widehat{i}_*:\mathcal{D}^{\mathrm{pot}}(\widehat{X}_Z)\xrightarrow{\sim}\mathcal{D}^{\mathrm{pot}}_Z(X), i ∗ : D pot ( X Z ) ∼ D Z pot ( X ) ,
with adjoint equivalence i ^ ! \widehat{i}^{!} i ! (Lemma 4.12). This observation reduces the problem to showing that the pushforward i ∗ : D p o t ( Z ) → D p o t ( X ^ Z ) i_*:\mathcal{D}^{\mathrm{pot}}(Z)\to\mathcal{D}^{\mathrm{pot}}(\widehat{X}_Z) i ∗ : D pot ( Z ) → D pot ( X Z ) is an equivalence—in other words, showing nil-invariance for D p o t \mathcal{D}^{\mathrm{pot}} D pot —which we do in this paper for X X X a linearly scalloped stack.
Reduction to quotient stacks
In order to show our main result for a linearly scalloped stack X X X , we reduce to the case when X X X has the form of a global quotient U / G U/G U / G , where G G G is a reductive group and U U U is a derived scheme.
To start this reduction, we show that if X X X is an algebraic stack, then D p o t ( − ) \mathcal{D}^{\mathrm{pot}}(-) D pot ( − ) satisfies Nisnevich descent on the category Rep / X \operatorname{Rep}_{/X} Rep / X of stacks Y Y Y over X X X whose structure morphism Y → X Y\to X Y → X is representable by algebraic spaces (Proposition 7.2). If X X X is linearly scalloped, this property of D p o t ( − ) \mathcal{D}^{\mathrm{pot}}(-) D pot ( − ) allows us to argue inductively using the scallop decomposition of X X X from the case where X X X has the form of a global quotient of a G G G -quasi-projective derived scheme (Theorem 7.17 ).10
α \alpha α -potent loop spaces
Now let X = U / G X=U/G X = U / G , where G G G is a reductive group and U U U is a derived G G G -scheme. In this case, the loop space L X \mathcal{L}X L X admits a “Jordan decomposition,” described in [Che20; BZCHN23], into certain S 1 S^1 S 1 -invariant formal substacks, denoted L α X \mathcal{L}^{\alpha}X L α X , parametrized by closed points α ∈ G / / G \alpha\in G/\!/G α ∈ G / / G ,11 which correspond to conjugacy classes of semisimple elements in G G G . For a given closed point α ∈ G / / G \alpha\in G/\!/G α ∈ G / / G , the space L α X \mathcal{L}^{\alpha}X L α X is defined as the formal completion of the fiber of the canonical map μ : L ( U / G ) → G / / G \mu:\mathcal{L}(U/G)\to G/\!/G μ : L ( U / G ) → G / / G over $\alpha12 a n d i s c a l l e d t h e and is called the an d i sc a l l e d t h e \alpha− p o t e n t l o o p s p a c e o f -potent loop space of − p o t e n tl oo p s p a ceo f X$.13
The reasons for introducing the spaces L α X \mathcal{L}^{\alpha}X L α X are:
Coh ( L α ( U / G ) ) t S 1 \operatorname{Coh}(\mathcal{L}^{\alpha}(U/G))^{tS^1} Coh ( L α ( U / G ) ) t S 1 is nil-invariant in the sense that pushforward along the inclusion i : Z / G → U / G i:Z/G\to U/G i : Z / G → U / G of any closed immersion14 induces an equivalence (Proposition 5.22),
( i ∗ ) t S 1 : Coh ( L α ( Z / G ) ) t S 1 → ∼ Coh ( L α ( U ^ Z / G ) ) t S 1 , (2) (i_*)^{tS^1}:\operatorname{Coh}(\mathcal{L}^{\alpha}(Z/G))^{tS^1}\xrightarrow{\sim}\operatorname{Coh}(\mathcal{L}^{\alpha}(\widehat{U}_Z/G))^{tS^1}, \tag*{(2)} ( i ∗ ) t S 1 : Coh ( L α ( Z / G ) ) t S 1 ∼ Coh ( L α ( U Z / G ) ) t S 1 , (2)
the nil-invariance of D p o t ( X ) : = IndCoh ( L ( U / G ) ) t T a t e \mathcal{D}^{\mathrm{pot}}(X):=\operatorname{IndCoh}(\mathcal{L}(U/G))^{t\mathrm{Tate}} D pot ( X ) := IndCoh ( L ( U / G ) ) t Tate can be deduced from knowing the equivalence (1.8) for each closed point α ∈ G / / G \alpha\in G/\!/G α ∈ G / / G .
Essential Surjectivity
It turns out that in order to show item (2) above, it is sufficient to show that the pushforward
( i ∗ ) t S 1 : Coh ( L ( Z / G ) ) t S 1 → Coh ( L ( U Z / G ) ^ ) t S 1 (i_*)^{tS^1} : \operatorname{Coh}(\mathcal{L}(Z/G))^{tS^1} \to\operatorname{Coh}(\widehat{\mathcal{L}(U_Z/G)})^{tS^1} ( i ∗ ) t S 1 : Coh ( L ( Z / G ) ) t S 1 → Coh ( L ( U Z / G ) ) t S 1
is fully faithful: In Proposition 6.22, we show that the image of ( i ∗ ) t S 1 (i_*)^{tS^1} ( i ∗ ) t S 1 generates Coh ( L ( U Z / G ) ^ ) t S 1 \operatorname{Coh}(\widehat{\mathcal{L}(U_Z/G)})^{tS^1} Coh ( L ( U Z / G ) ) t S 1 , which, combined with fully faithfulness, proves the theorem after ind-extending (Theorem 6.7 ). The proof of generation reduces to showing the corresponding statement on the hearts of the t t t -structures on each of the source and target of ( i ∗ ) t S 1 (i_*)^{tS^1} ( i ∗ ) t S 1 , which we show by a classical argument involving filtrations defined by powers of the ideal sheaf of Z c l Z^{\mathrm{cl}} Z cl .
Leveraging the α \alpha α -potent loop space
To show fully faithfulness (Proposition 6.11 ), we use item (1). We first reduce, by Lemma 3.10 , to showing that ( i ∗ ) t S 1 (i_*)^{tS^1} ( i ∗ ) t S 1 is fully faithful on objects which are in the image of
Coh ( L ( Z / G ) ) S 1 → Coh ( L ( Z / G ) ) t S 1 . \operatorname{Coh}(\mathcal{L}(Z/G))^{S^1} \to\operatorname{Coh}(\mathcal{L}(Z/G))^{tS^1}. Coh ( L ( Z / G ) ) S 1 → Coh ( L ( Z / G ) ) t S 1 .
Denoting two such objects by F \mathcal{F} F and G \mathcal{G} G , we thus reduce to showing that the following fiber is 0
F : = ker ( Hom Coh ( L ( Z / G ) ) t S 1 ( F , G ) → Hom Coh ( L ( U Z / G ) ^ ) t S 1 ( ( i ∗ ) t S 1 F , ( i ∗ ) t S 1 G ) ) . \mathcal{F}:=\ker\left(\operatorname{Hom}_{\operatorname{Coh}(\mathcal{L}(Z/G))^{tS^1}}(\mathcal{F},\mathcal{G}) \to\operatorname{Hom}_{\operatorname{Coh}(\widehat{\mathcal{L}(U_Z/G)})^{tS^1}}((i_*)^{tS^1}\mathcal{F},(i_*)^{tS^1}\mathcal{G})\right). F := ker ( Hom Coh ( L ( Z / G ) ) t S 1 ( F , G ) → Hom Coh ( L ( U Z / G ) ) t S 1 (( i ∗ ) t S 1 F , ( i ∗ ) t S 1 G ) ) .
Our strategy is then to use a theorem of Neeman [NB92, Lemma 2.12], which states that a (possibly unbounded) complex of A A A -modules M M M is 0 if and only if all base changes M ⊗ A k ( p ) ≃ 0 M\otimes_A k(\mathfrak{p})\simeq0 M ⊗ A k ( p ) ≃ 0 for all prime ideals p ∈ Spec A \mathfrak{p}\in\operatorname{Spec} A p ∈ Spec A , where k ( p ) k(\mathfrak{p}) k ( p ) is the residue field of p \mathfrak{p} p . Since the categories in question are all O ( G / / G ) \mathcal{O}(G/\!/G) O ( G / / G ) -linear, we aim to show, for Spec A : = G / / G \operatorname{Spec} A:=G/\!/G Spec A := G / / G , that F ⊗ A k ( p ) ≃ 0 \mathcal{F}\otimes_A k(\mathfrak{p})\simeq0 F ⊗ A k ( p ) ≃ 0 for all prime ideals p ⊂ A \mathfrak{p}\subset A p ⊂ A .15
After choosing a prime ideal p ⊂ A \mathfrak{p}\subset A p ⊂ A , we obtain a closed point α ′ \alpha' α ′ of Spec A k ( p ) : = Spec ( A ⊗ k k ( p ) ) \operatorname{Spec} A_{k(\mathfrak{p})}:=\operatorname{Spec}(A\otimes_k k(\mathfrak{p})) Spec A k ( p ) := Spec ( A ⊗ k k ( p )) by using the point A → k ( p ) A\to k(\mathfrak{p}) A → k ( p ) . However, the closed point α ′ \alpha' α ′ is not geometric; hence, it is not necessarily represented by a semisimple element of G G G . In order to apply Proposition 5.22, we require a semisimple representative. Thus, to correct this, we pass to the algebraic closure k ( p ) ‾ \overline{k(\mathfrak{p})} k ( p ) , and consider instead the geometric point α \alpha α determined by A → k ( p ) ‾ A\to\overline{k(\mathfrak{p})} A → k ( p ) . We then base change all our geometric objects to k ( p ) ‾ \overline{k(\mathfrak{p})} k ( p ) . Denoting ( − ) k ( p ) ‾ : = − ⊗ k k ( p ) ‾ (-)_{\overline{k(\mathfrak{p})}}:=-\otimes_k\overline{k(\mathfrak{p})} ( − ) k ( p ) := − ⊗ k k ( p ) , we next show that Hom complexes appearing in the base-changed fiber F k ( p ) ‾ ∈ A k ( p ) ‾ -mod \mathcal{F}_{\overline{k(\mathfrak{p})}}\in A_{\overline{k(\mathfrak{p})}}\text{-mod} F k ( p ) ∈ A k ( p ) -mod can be written as Hom complexes on the base changed stacks in question; in other words, all calculations can now be performed relative to k ( p ) ‾ \overline{k(\mathfrak{p})} k ( p ) as the base algebraically closed field.
Now, in order to leverage item (1), we tensor the base changed fiber with − ⊗ A k ( p ) ‾ P ⊗ A k ( p ) ‾ P ∨ -\otimes_{A_{\overline{k(\mathfrak{p})}}}P\otimes_{A_{\overline{k(\mathfrak{p})}}}P^\vee − ⊗ A k ( p ) P ⊗ A k ( p ) P ∨ , where P P P is the Koszul complex of the maximal ideal of A k ( p ) ‾ A_{\overline{k(\mathfrak{p})}} A k ( p ) corresponding to the closed point α \alpha α . The purpose of this maneuver is to use Lemma 6.8 , which allows us to rewrite F ⊗ A k ( p ) ‾ P ⊗ A k ( p ) ‾ P ∨ \mathcal{F}\otimes_{A_{\overline{k(\mathfrak{p})}}}P\otimes_{A_{\overline{k(\mathfrak{p})}}}P^\vee F ⊗ A k ( p ) P ⊗ A k ( p ) P ∨ as a kernel of Hom complexes between sheaves on the α \alpha α -potent loop spaces, since, by an argument similar to that in Lemma 4.12, an S 1 S^1 S 1 -equivariant sheaf supported on the inverse image of α \alpha α is an S 1 S^1 S 1 -equivariant sheaf on the formal completion at that inverse image, which is precisely the α \alpha α -potent loop space. Then, we use item (1) to conclude that
F k ( p ) ‾ ⊗ A k ( p ) ‾ P ⊗ A k ( p ) ‾ P ∨ ≃ 0. \mathcal{F}_{\overline{k(\mathfrak{p})}}\otimes_{A_{\overline{k(\mathfrak{p})}}}P\otimes_{A_{\overline{k(\mathfrak{p})}}}P^\vee\simeq0. F k ( p ) ⊗ A k ( p ) P ⊗ A k ( p ) P ∨ ≃ 0.
Finally, after base changing this complex along A k ( p ) ‾ → k ( p ) ‾ A_{\overline{k(\mathfrak{p})}}\to\overline{k(\mathfrak{p})} A k ( p ) → k ( p ) , the Koszul differential becomes 0 in the resulting complex; thus, by perfectness of P P P , it is a direct sum of copies of shifts of F ⊗ A k ( p ) ‾ \mathcal{F}\otimes_A\overline{k(\mathfrak{p})} F ⊗ A k ( p ) . In particular, each of these is 0, so that we have F ⊗ A k ( p ) ‾ ≃ 0 \mathcal{F}\otimes_A\overline{k(\mathfrak{p})}\simeq0 F ⊗ A k ( p ) ≃ 0 , hence F ⊗ A k ( p ) ≃ 0 \mathcal{F}\otimes_A k(\mathfrak{p})\simeq0 F ⊗ A k ( p ) ≃ 0 , as desired.
Structure of paper
In § 2 , we establish the basic conventions and notation used throughout the paper. In particular, this section contains a glossary of categories that appear in the work. We encourage the reader to consult this section first whenever unfamiliar notation appears.
In § 3 , we collect the foundational definitions and results on group actions, loop spaces, and formal completions used throughout the text. We advise the reader to skip this section at first and return to it as necessary while reading the main text.
In § 4 , we develop the theory of potent D \mathcal{D} D -modules necessary to formulate and prove our main result. We also recall various elements of the six functor formalism on potent D \mathcal{D} D -modules constructed in the forthcoming work [ES].
In Section 5, we discuss the α \alpha α -potent loop space L α \mathcal{L}^{\alpha} L α , A A A -trivializations, and the α \alpha α -shifted action on the ordinary unipotent loop space. Much of the content in this section are recollections of the material in [BZCHN23, Appendix A], as well as compatibility statements between operations like formal completion, fixed points, and A A A -trivialization. The main result of this section is the nil-invariance of cyclic sheaves on the α \alpha α -potent loop space.
In § 6 we prove Kashiwara’s Lemma (Theorem 1.1) holds for quotient stacks. We start by proving some algebraic properties of the interaction between S 1 S_1 S 1 -invariants and Hom complexes. Using these comparisons, we prove Kashiwara’s Lemma for global quotient stacks.
Finally, in § 7 , we use the results of the previous section to prove Kashiwara’s Lemma for all linearly scalloped stacks.
Acknowledgements
JE would first like to thank his advisor, David Nadler, for his continued support and encouragement during this project, and Zhiwei Yun for related discussions which lead to the method of proof of Lemma 3.48 .
KS thanks Rok Gregoric for illuminating discussions on the topics of homotopy theory and categorical group actions.
The authors together would like to thank Pavel Safronov, Sanath Devalapurkar, Swapnil Garg, Yuji Okitani, Jeremy Taylor, Jens Eberhardt, Spencer Tamagni, Germán Stefanich, Sam Gunningham, Sam Raskin, and Justin Hilburn for discussions related to the subject of this work.
The authors thank David Nadler for discussions that led to the creation of this paper. The authors also thank Harrison Chen for answering numerous questions about his previous work, and for generously sharing work-in-progress. Lastly, we thank David Ben-Zvi for drawing our attention to the Jordan loop space decomposition, as well as for patiently explaining the philosophy of decompletion to us.
AI use statement
AI was used for literature search throughout this project, and for locating errors as the project neared completion. It was also used to fix some issues with LaTeX packages, and to help with typesetting.
Conventions and notation
2.0.1. Fix an algebraically closed field k k k of characteristic zero, once and for all.
2.0.2. Given a functor F : C → D F:\mathcal{C} \to\mathcal{D} F : C → D between ∞ \infty ∞ -categories, im ( F ) \operatorname{im}(F) im ( F ) denotes the essential image of F F F .
2.0.3. Hom-objects and mapping spaces. The reader will encounter ∞ \infty ∞ -categories in this paper equipped with a variety of different structures, each of which we use at different times. Here we establish notation for the various hom-objects and mapping spaces associated with these structures.
Let C \mathcal{C} C be an ∞ \infty ∞ -category, and let a , b ∈ C a,b \in\mathcal{C} a , b ∈ C be two objects. We adopt the following notation.
Map C ( a , b ) ∈ Grpd ∞ \operatorname{Map}_{\mathcal{C}}(a,b) \in\operatorname{Grpd}_{\infty} Map C ( a , b ) ∈ Grpd ∞ denotes the mapping space from a a a to b b b . We drop the subscript “C \mathcal{C} C ” when the category is clear from context.
(2) If ( C , ⊗ ) (\mathcal{C},\otimes) ( C , ⊗ ) is closed symmetric monoidal, then H o m C ( a , b ) ∈ C \mathcal{H}om_{\mathcal{C}}(a,b)\in\mathcal{C} H o m C ( a , b ) ∈ C denotes the internal hom-object from a a a to b b b , i.e. the object determined by,
Map C ( c , H o m C ( a , b ) ) ≃ Map C ( c ⊗ a , b ) , \operatorname{Map}_{\mathcal{C}}(c,\mathcal{H}om_{\mathcal{C}}(a,b))\simeq\operatorname{Map}_{\mathcal{C}}(c\otimes a,b), Map C ( c , H o m C ( a , b )) ≃ Map C ( c ⊗ a , b ) ,
for c ∈ C c\in\mathcal{C} c ∈ C . As above, we drop the subscript “C \mathcal{C} C ” when the category is clear from context.
(3) If C \mathcal{C} C is a module category over a rigid symmetric monoidal category R \mathcal{R} R , then Hom R ( a , b ) ∈ R \operatorname{Hom}_{\mathcal{R}}(a,b)\in\mathcal{R} Hom R ( a , b ) ∈ R is the object determined by
Map R ( V , Hom R ( a , b ) ) ≃ Map C ( V ⊗ a , b ) , \operatorname{Map}_{\mathcal{R}}(V,\operatorname{Hom}_{\mathcal{R}}(a,b))\simeq\operatorname{Map}_{\mathcal{C}}(V\otimes a,b), Map R ( V , Hom R ( a , b )) ≃ Map C ( V ⊗ a , b ) ,
for V ∈ R V\in\mathcal{R} V ∈ R , a , b ∈ C a,b\in\mathcal{C} a , b ∈ C .
When R \mathcal{R} R is R R R –mod for a ring spectrum R R R , we alternatively write Hom R ( a , b ) \operatorname{Hom}_{R}(a,b) Hom R ( a , b ) . For example, if C \mathcal{C} C is an object of dgcat k _k k , then Hom k ( a , b ) \operatorname{Hom}_k(a,b) Hom k ( a , b ) is the usual k k k -module spectrum of maps between a a a and b b b .
(4) The space of n n n -morphisms in C \mathcal{C} C is given by Map C a t ∞ ( [ n ] , C ) \operatorname{Map}_{\mathrm{Cat}_{\infty}}([n],\mathcal{C}) Map Cat ∞ ([ n ] , C ) where [ n ] [n] [ n ] denotes the n n n -simplex category. We denote by Ob ( C ) \operatorname{Ob}(\mathcal{C}) Ob ( C ) , Mor ( C ) \operatorname{Mor}(\mathcal{C}) Mor ( C ) , and Mor n ( C ) \operatorname{Mor}_n(\mathcal{C}) Mor n ( C ) , respectively, the spaces 0-morphisms (i.e. objects), 1-morphisms, and n n n -morphisms, respectively.
Warning 2.1. Given two objects X , Y ∈ S t k k X,Y\in\mathrm{Stk}_k X , Y ∈ Stk k , we will denote the internal mapping object of S t k k \mathrm{Stk}_k Stk k between X X X and Y Y Y , i.e. the mapping stack, by Map ( X , Y ) \operatorname{Map}(X,Y) Map ( X , Y ) .
We will mainly be interested in categories of sheaves in this paper, for which we have the following alternative notation.
Let f : X → S f:X\to S f : X → S be a derived stack over S S S . Pullback along f f f gives a natural symmetric monoidal functor f ∗ : Q C o h ( S ) → Q C o h ( X ) f^*:\mathrm{QCoh}(S)\to\mathrm{QCoh}(X) f ∗ : QCoh ( S ) → QCoh ( X ) , making Q C o h ( X ) \mathrm{QCoh}(X) QCoh ( X ) into an algebra over Q C o h ( S ) \mathrm{QCoh}(S) QCoh ( S ) . Composing f ∗ f^* f ∗ with the canonical symmetric monoidal embedding Υ : Q C o h ( X ) → I n d C o h ( X ) \Upsilon:\mathrm{QCoh}(X)\to\mathrm{IndCoh}(X) Υ : QCoh ( X ) → IndCoh ( X ) obtains a symmetric monoidal functor Q C o h ( S ) → I n d C o h ( X ) \mathrm{QCoh}(S)\to\mathrm{IndCoh}(X) QCoh ( S ) → IndCoh ( X ) , making I n d C o h ( X ) \mathrm{IndCoh}(X) IndCoh ( X ) into an algebra over Q C o h ( S ) \mathrm{QCoh}(S) QCoh ( S ) , as well.
(1) We use H o m X ( − , − ) \mathcal{H}om_X(-,-) H o m X ( − , − ) to denote the hom-object (2) for ( C , ⊗ ) (\mathcal{C},\otimes) ( C , ⊗ ) taken to be either ( Q C o h ( X ) , ⊗ ) (\mathrm{QCoh}(X),\otimes) ( QCoh ( X ) , ⊗ ) or ( I n d C o h ( X ) , ⊗ ) (\mathrm{IndCoh}(X),\otimes) ( IndCoh ( X ) , ⊗ ) . When there is potential for confusion, we write H o m Q C o h ( X ) ( − , − ) \mathcal{H}om_{\mathrm{QCoh}(X)}(-,-) H o m QCoh ( X ) ( − , − ) or H o m I n d C o h ( X ) ( − , − ) \mathcal{H}om_{\mathrm{IndCoh}(X)}(-,-) H o m IndCoh ( X ) ( − , − ) .
(2) We use Hom S ( − , − ) \operatorname{Hom}_S(-,-) Hom S ( − , − ) to denote the hom-object of (3) above for R = Q C o h ( S ) \mathcal{R}=\mathrm{QCoh}(S) R = QCoh ( S ) and C \mathcal{C} C taken to be either Q C o h ( X ) \mathrm{QCoh}(X) QCoh ( X ) or I n d C o h ( X ) \mathrm{IndCoh}(X) IndCoh ( X ) .
Inclusions
2.0.4. Inclusions. In order to minimize verbiage, we will use i ? i? i ? as standing, all-purpose notation to denote the inclusion of some space (“?”) into another, whenever said inclusion is clear from context.
2.0.5. Tate invariants. Let C \mathcal{C} C be a k k k -linear ∞ \infty ∞ -category equipped with an S 1 S^1 S 1 -action. The invariant category C S 1 \mathcal{C}^{S^1} C S 1 is linear over k [ u ] k[u] k [ u ] , where u u u has cohomological degree 2. We set
C t S 1 : = C S 1 ⊗ k [ u ] k [ u ± 1 ] , \mathcal{C}^{tS^1}:=\mathcal{C}^{S^1}\mathbin{\otimes_{k[u]}}k[u^{\pm1}], C t S 1 := C S 1 ⊗ k [ u ] k [ u ± 1 ] ,
and call the resulting category the Tate invariants of C \mathcal{C} C .
When C \mathcal{C} C is a large category, the Tate invariants might not be the correct notion for a given purpose, so we often use the following construction of Preygel instead. Assume that C \mathcal{C} C is a k k k -linear stable ∞ \infty ∞ -category with a coherent t t t -structure ([3 Definition 4.2.7] ). Let Coh ( C ) ⊂ C \operatorname{Coh}(\mathcal{C}) \subset\mathcal{C} Coh ( C ) ⊂ C be the full subcategory spanned by bounded almost perfect objects of C \mathcal{C} C . Then Preygel defines the t t t Tate invariants of C \mathcal{C} C to be the ind-completion of the Tate invariants of the induced S 1 S^1 S 1 -action on Coh ( C ) \operatorname{Coh}(\mathcal{C}) Coh ( C ) :
C t T a t e : = Ind ( Coh ( C ) t S 1 ) . \mathcal{C}^{t\mathrm{Tate}} := \operatorname{Ind}\left(\operatorname{Coh}(\mathcal{C})^{tS^1}\right). C t Tate := Ind ( Coh ( C ) t S 1 ) .
2.0.6. Stacks. We use the notions of geometric and ind-geometric stacks of [CW23]. In particular, if X X X is a coherent ind-geometric stack with S 1 S^1 S 1 -action, it makes sense to write
IndCoh ( X ) t T a t e \operatorname{IndCoh}(X)^{t\mathrm{Tate}} IndCoh ( X ) t Tate
by [CW23, Proposition 5.30] and the results of § 3.1 below. In fact, all our stacks will be coherent ind-geometric stacks locally almost of finite type. By Remark 5.20 of op. cit. , for such stacks the categories of ind-coherent sheaves defined in [CW23] and in [Gai13; GR17a; GR17b] are the same.
By closed immersion , we mean a closed immersion in the sense of [CW23, Section 3.7] that is also schematic.
2.0.7. Symmetric monoidal categories. If ( C , ⊗ ) (\mathcal{C}, \otimes) ( C , ⊗ ) is a symmetric monoidal category, we denote the tensor unit by 1 C 1_{\mathcal{C}} 1 C .
Glossary of categories
Cat ∞ \operatorname{Cat}_{\infty} Cat ∞ denotes the ∞ \infty ∞ -category of ∞ \infty ∞ -categories.
Cat ∞ t \operatorname{Cat}_{\infty}^{t} Cat ∞ t denotes the category of stable ∞ \infty ∞ -categories equipped with t t t -structures, whose 1 1 1 -morphisms are t t t -exact functors, and whose higher morphisms are as in Cat ∞ \operatorname{Cat}_{\infty} Cat ∞ .
Cat ∞ s m \operatorname{Cat}_{\infty}^{\mathrm{sm}} Cat ∞ sm denotes the full subcategory of Cat ∞ \operatorname{Cat}_{\infty} Cat ∞ spanned by small, idempotent-complete ∞ \infty ∞ -categories.
Cat ∞ , s t \operatorname{Cat}_{\infty,\mathrm{st}} Cat ∞ , st denotes the subcategory of Cat ∞ \operatorname{Cat}_{\infty} Cat ∞ spanned by stable ∞ \infty ∞ -categories, whose 1 1 1 -morphisms are exact functors, and whose higher morphisms are as in Cat ∞ \operatorname{Cat}_{\infty} Cat ∞ .
Cat ∞ , s t s m \operatorname{Cat}_{\infty,\mathrm{st}}^{\mathrm{sm}} Cat ∞ , st sm denotes the subcategory of Cat ∞ \operatorname{Cat}_{\infty} Cat ∞ given by Cat ∞ , s t ∩ Cat ∞ s t \operatorname{Cat}_{\infty,\mathrm{st}} \cap\operatorname{Cat}_{\infty}^{\mathrm{st}} Cat ∞ , st ∩ Cat ∞ st .
Pr s t L \operatorname{Pr}_{\mathrm{st}}^{\mathrm{L}} Pr st L denotes the ∞ \infty ∞ -subcategory of Cat ∞ \operatorname{Cat}_{\infty} Cat ∞ spanned by presentable, stable ∞ \infty ∞ -categories, whose 1 1 1 -morphisms are exact colimit-preserving functors, and whose higher morphisms are as in Cat ∞ \operatorname{Cat}_{\infty} Cat ∞ .
( Pr s t L ) t (\operatorname{Pr}_{\mathrm{st}}^{\mathrm{L}})^{t} ( Pr st L ) t denotes the ∞ \infty ∞ -category of objects in Pr s t L \operatorname{Pr}_{\mathrm{st}}^{\mathrm{L}} Pr st L equipped with t t t -structures, whose 1 1 1 -morphisms are t t t -exact functors, and whose higher morphisms are as in Pr s t L \operatorname{Pr}_{\mathrm{st}}^{\mathrm{L}} Pr st L .
DGCat k \operatorname{DGCat}_{k} DGCat k denotes Mod Vect k ( Pr s t L ) \operatorname{Mod}_{\operatorname{Vect}_{k}}(\operatorname{Pr}_{\mathrm{st}}^{\mathrm{L}}) Mod Vect k ( Pr st L ) .
DGCat k t \operatorname{DGCat}_{k}^{t} DGCat k t denotes the ∞ \infty ∞ -category of objects in DGCat k \operatorname{DGCat}_{k} DGCat k equipped with t t t -structures, whose 1 1 1 -morphisms are t t t -exact functors, and whose higher morphisms are as in DGCat k \operatorname{DGCat}_{k} DGCat k . Equivalently, DGCat k t ≃ Mod ( Vect k , t ) ( ( Pr s t L ) t ) \operatorname{DGCat}_{k}^{t} \simeq\operatorname{Mod}_{(\operatorname{Vect}_{k},t)}((\operatorname{Pr}_{\mathrm{st}}^{\mathrm{L}})^{t}) DGCat k t ≃ Mod ( Vect k , t ) (( Pr st L ) t ) , where ( Vect k , t ) (\operatorname{Vect}_{k},t) ( Vect k , t ) denotes Vect k \operatorname{Vect}_{k} Vect k equipped with its standard t t t -structure. There is an obvious forgetful functor oblv : DGCat k t → DGCat k \operatorname{oblv}:\operatorname{DGCat}_{k}^{t}\to\operatorname{DGCat}_{k} oblv : DGCat k t → DGCat k which forgets the t t t -structure.
d g c a t k \mathbf{dgcat}_{k} dgcat k denotes Mod Perf ( k ) ( Cat ∞ s m ) \operatorname{Mod}_{\operatorname{Perf}(k)}(\operatorname{Cat}_{\infty}^{\mathrm{sm}}) Mod Perf ( k ) ( Cat ∞ sm ) .
S 1 S^{1} S 1 -mod denotes Fun ( B S 1 , d g c a t k ) \operatorname{Fun}(BS^{1},\mathbf{dgcat}_{k}) Fun ( B S 1 , dgcat k ) .
If A A A is a commutative dg algebra, then A A A -mod denotes A A A -mod( d g c a t k ) (\mathbf{dgcat}_{k}) ( dgcat k ) .
Ch k \operatorname{Ch}_{k} Ch k denote the dg-category of strict chain complexes over the field k k k . There is a localization map Ch k → Vect k \operatorname{Ch}_{k}\to\operatorname{Vect}_{k} Ch k → Vect k inverting quasi-isomorphisms.
QCoh ( X ) \operatorname{QCoh}(X) QCoh ( X ) denotes the k k k -linear (stable cocomplete) ∞ \infty ∞ -category of quasicoherent complexes on a derived prestack X X X . It is equipped with a natural t t t -structure, whose heart QCoh ( X ) ♡ \operatorname{QCoh}(X)^{\heartsuit} QCoh ( X ) ♡ is equivalent to the (ordinary) category of quasicoherent sheaves on π 0 ( X ) \pi_{0}(X) π 0 ( X ) .
Perf ( X ) ⊂ QCoh ( X ) \operatorname{Perf}(X) \subset\operatorname{QCoh}(X) Perf ( X ) ⊂ QCoh ( X ) is the full subcategory spanned by perfect complexes. If X X X is perfect (e.g. a quasi-compact, quasi-separated derived scheme) in the sense of [BZFN10], then QCoh ( X ) = Ind ( Perf ( X ) ) \operatorname{QCoh}(X) = \operatorname{Ind}(\operatorname{Perf}(X)) QCoh ( X ) = Ind ( Perf ( X )) .
Let X X X be a derived stack. Then Coh ( X ) ⊂ QCoh ( X ) \operatorname{Coh}(X) \subset\operatorname{QCoh}(X) Coh ( X ) ⊂ QCoh ( X ) is the full subcategory spanned by objects with locally bounded, coherent cohomology sheaves.
For laft prestacks X X X , IndCoh ( X ) \operatorname{IndCoh}(X) IndCoh ( X ) is defined in detail in [Gai13], but for X X X as in Section 2.0.6, it is simply the ind-completion of Coh ( X ) \operatorname{Coh}(X) Coh ( X ) defined above.
Spc \operatorname{Spc} Spc denotes the ∞ \infty ∞ -category of ∞ \infty ∞ -groupoids, whose objects are alternatively called spaces or animae .
Given an ∞ \infty ∞ -category C \mathcal{C} C with a t-structure, C + \mathcal{C}^{+} C + will denote ⋃ n ≥ 0 C ≥ − n \bigcup_{n\geq0}\mathcal{C}^{\geq-n} ⋃ n ≥ 0 C ≥ − n , and C − \mathcal{C}^{-} C − will denote ⋃ n ≥ 0 C ≤ n \bigcup_{n\geq0}\mathcal{C}^{\leq n} ⋃ n ≥ 0 C ≤ n .
We let Coh t \operatorname{Coh}_{t} Coh t denote the ∞ \infty ∞ -category whose objects are ∞ \infty ∞ -categories equipped with coherent t-structure, whose 1-morphisms are continuous and left t-exact up to a shift functors, and whose higher morphisms are those in Cat ∞ \operatorname{Cat}_{\infty} Cat ∞ .16
We let Reg t ⊂ Coh t \operatorname{Reg}_{t} \subset\operatorname{Coh}_{t} Reg t ⊂ Coh t denote the full subcategory whose objects are ∞ \infty ∞ -categories with regular t-structures.
Preliminaries
Group actions
Definition 3.1. Let G G G be a group object in Ani \operatorname{Ani} Ani (i.e. an ∞ \infty ∞ -group), and let C \mathcal{C} C be an ∞ \infty ∞ -category. A G G G -action on an object c ∈ C c \in\mathcal{C} c ∈ C is defined to be a functor17 ^{17} 17 B G → C BG \to\mathcal{C} B G → C such that the composition ∗ → B G → C \ast\to BG \to\mathcal{C} ∗ → B G → C is the object c c c .
In particular, one may take C \mathcal{C} C to be Cat ∞ \operatorname{Cat}_{\infty} Cat ∞ in the definition above to obtain the notion of an ∞ \infty ∞ -category equipped with a G G G -action, which call a G G G -category. Given an object c ∈ C c \in\mathcal{C} c ∈ C equipped with a G G G -action, we often write c : B G → C c : BG \to\mathcal{C} c : B G → C by abuse of notation.
Given a G G G -category C \mathcal{C} C , one may consider objects of C \mathcal{C} C equipped with G G G -actions which are compatible with that on C \mathcal{C} C .
Definition 3.2. Let C \mathcal{C} C be a G G G -category. The space of G G G -equivariant objects of C \mathcal{C} C is defined to be Map G ( ∗ , C ) : = Map Fun ( B G , Cat ∞ ) ( ∗ , C ) \operatorname{Map}_{G}(\ast,\mathcal{C}) := \operatorname{Map}_{\operatorname{Fun}(BG,\operatorname{Cat}_{\infty})}(\ast,\mathcal{C}) Map G ( ∗ , C ) := Map Fun ( B G , Cat ∞ ) ( ∗ , C ) , i.e. the space of natural transformations from the point (equipped with its only possible G G G -action) to C : B G → Cat ∞ \mathcal{C} : BG \to\operatorname{Cat}_{\infty} C : B G → Cat ∞ .
Remark 3.3. An object with G G G -action in a plain ∞ \infty ∞ -category C \mathcal{C} C is equivalently a G G G -equivariant object of the trivial G G G -category C \mathcal{C} C .
The G G G -action on mapping spaces
Given a G G G -category C \mathcal{C} C , and a pair of G G G -equivariant objects c , c ′ ∈ C c,c' \in\mathcal{C} c , c ′ ∈ C , there is a canonical G G G -action on the mapping space Map ( c , c ′ ) \operatorname{Map}(c,c') Map ( c , c ′ ) between the underlying objects of C \mathcal{C} C defined as follows.
Given a G G G -category C : B G → Cat ∞ \mathcal{C} : BG \to\operatorname{Cat}_{\infty} C : B G → Cat ∞ , there is an induced action of G op G^{\operatorname{op}} G op on the presheaf category P ( C ) \mathcal{P}(\mathcal{C}) P ( C ) given by the composition B G op → C op Cat ∞ op → P Cat ∞ BG^{\operatorname{op}} \xrightarrow{\mathcal{C}^{\operatorname{op}}} \operatorname{Cat}_{\infty}^{\operatorname{op}} \xrightarrow{\mathcal{P}} \operatorname{Cat}_{\infty} B G op C op Cat ∞ op P Cat ∞ , where P \mathcal{P} P is the functor taking the category C \mathcal{C} C to Fun ( C op , Ani ) \operatorname{Fun}(\mathcal{C}^{\operatorname{op}},\operatorname{Ani}) Fun ( C op , Ani ) . This operation is functorial in C \mathcal{C} C ,
Fun ( B G , Cat ∞ ) op → ( − ) op Fun ( B G op , Cat ∞ op ) → P o- Fun ( B G op , Cat ∞ ) , \operatorname{Fun}(BG,\operatorname{Cat}_{\infty})^{\operatorname{op}} \xrightarrow{(-)^{\operatorname{op}}} \operatorname{Fun}(BG^{\operatorname{op}},\operatorname{Cat}_{\infty}^{\operatorname{op}}) \xrightarrow{\mathcal{P}_{\operatorname{o-}}} \operatorname{Fun}(BG^{\operatorname{op}},\operatorname{Cat}_{\infty}), Fun ( B G , Cat ∞ ) op ( − ) op Fun ( B G op , Cat ∞ op ) P o- Fun ( B G op , Cat ∞ ) ,
giving a functor P ~ : ( Cat ∞ G ) op → Cat ∞ G op \widetilde{\mathcal{P}} : (\operatorname{Cat}_{\infty}^{G})^{\operatorname{op}} \to\operatorname{Cat}_{\infty}^{G^{\operatorname{op}}} P : ( Cat ∞ G ) op → Cat ∞ G op .
Note that, if ∗ \ast ∗ denotes the trivial G G G -category, P ~ ( ∗ ) = A n i \widetilde{\mathcal{P}}(\ast)=\mathrm{Ani} P ( ∗ ) = Ani , equipped with the trivial G o p G^{\mathrm{op}} G op -action. Consider a G G G -equivariant object in C \mathcal{C} C , c : ∗ → C c:\ast\to\mathcal{C} c : ∗ → C . Then P ~ ( c o p ) \widetilde{\mathcal{P}}(c^{\mathrm{op}}) P ( c op ) gives a G o p G^{\mathrm{op}} G op -equivariant functor c ∗ : P ( C ) → A n i c^\ast:\mathcal{P}(\mathcal{C})\to\mathrm{Ani} c ∗ : P ( C ) → Ani . In particular, if c ′ c' c ′ is G G G -equivariant object of C \mathcal{C} C , we see that c ∗ c^\ast c ∗ sends the presheaf represented by c ′ c' c ′ to a G o p G^{\mathrm{op}} G op -equivariant object of A n i \mathrm{Ani} Ani whose underlying space is Map ( c , c ′ ) \operatorname{Map}(c,c') Map ( c , c ′ ) . Imitating this construction for copresheaves, we find a G G G -equivariant object of A n i \mathrm{Ani} Ani whose underlying space is also Map ( c , c ′ ) \operatorname{Map}(c,c') Map ( c , c ′ ) .
Thus, given a pair of G G G -equivariant objects c , c ′ c,c' c , c ′ in C \mathcal{C} C , there is a canonical G o p × G G^{\mathrm{op}}\times G G op × G -action on Map ( c , c ′ ) \operatorname{Map}(c,c') Map ( c , c ′ ) . Restricting along the diagonal ( − ) − 1 × id G : G → G o p × G (-)^{-1}\times\operatorname{id}_{G}:G\to G^{\mathrm{op}}\times G ( − ) − 1 × id G : G → G op × G , we obtain a canonical G G G -action on the mapping space Map ( c , c ′ ) \operatorname{Map}(c,c') Map ( c , c ′ ) .
Remark 3.4. The special case of this construction where C \mathcal{C} C has the trivial G G G -action is more straightforward. In this case, a pair of G G G -equivariant objects is a pair of functors c : B G → C c:BG\to\mathcal{C} c : B G → C and c ′ : B G → C c':BG\to\mathcal{C} c ′ : B G → C . Then the G o p × G G^{\mathrm{op}}\times G G op × G -action on the mapping space Map ( c ( ∗ ) , c ′ ( ∗ ) ) \operatorname{Map}(c(\ast),c'(\ast)) Map ( c ( ∗ ) , c ′ ( ∗ )) is given by the composition
B G o p × B G → c o p × c ′ C o p × C → Map ( − , − ) A n i , BG^{\mathrm{op}}\times BG \xrightarrow{c^{\mathrm{op}}\times c'} \mathcal{C}^{\mathrm{op}}\times\mathcal{C} \xrightarrow{\operatorname{Map}(-,-)} \mathrm{Ani}, B G op × B G c op × c ′ C op × C Map ( − , − ) Ani ,
and corresponding G G G -action comes from precomposing this functor with the diagonal map B G → B G o p × B G BG\to BG^{\mathrm{op}}\times BG B G → B G op × B G .
Invariants and coinvariants
Definition 3.5. Let G G G an ∞ \infty ∞ -group and C \mathcal{C} C an ∞ \infty ∞ -category. Suppose that c : B G → C c:BG\to\mathcal{C} c : B G → C is an object of C \mathcal{C} C equipped with a G G G -action. Then the G G G -invariants and G G G -coinvariants of c c c are defined to be
c G : = lim B G C and c G : = colim B G C , c^{G}:=\lim_{BG}\mathcal{C}\qquad\text{and}\qquad c_{G}:=\operatorname{colim}_{BG}\mathcal{C}, c G := B G lim C and c G := colim B G C ,
respectively.
Remark 3.6. Note that a G G G -equivariant object of a G G G -category C \mathcal{C} C is the same thing as an object of C G \mathcal{C}^{G} C G . Indeed, for any ∞ \infty ∞ -category E \mathcal{E} E , the functor of taking G G G -invariants, ( − ) G : Fun ( B G , E ) → E (-)^{G}:\operatorname{Fun}(BG,\mathcal{E})\to\mathcal{E} ( − ) G : Fun ( B G , E ) → E is right-adjoint to the functor triv : E → Fun ( B G , E ) \operatorname{triv}:\mathcal{E}\to\operatorname{Fun}(BG,\mathcal{E}) triv : E → Fun ( B G , E ) that equips objects with the trivial G G G -action. Taking E \mathcal{E} E to be C a t ∞ \mathrm{Cat}_{\infty} Cat ∞ , we thus obtain,
C G ≃ Fun C a t ∞ ( ∗ , C G ) ≃ Fun Fun ( B G , C a t ∞ ) ( triv ( ∗ ) , C ) = Fun G ( ∗ , C ) . \mathcal{C}^{G}\simeq\operatorname{Fun}_{\mathrm{Cat}_{\infty}}(\ast,\mathcal{C}^{G})\simeq\operatorname{Fun}_{\operatorname{Fun}(BG,\mathrm{Cat}_{\infty})}(\operatorname{triv}(\ast),\mathcal{C})=\operatorname{Fun}_{G}(\ast,\mathcal{C}). C G ≃ Fun Cat ∞ ( ∗ , C G ) ≃ Fun Fun ( B G , Cat ∞ ) ( triv ( ∗ ) , C ) = Fun G ( ∗ , C ) .
In particular, under this equivalence, a morphism in C G \mathcal{C}^{G} C G corresponds to a map between natural transformations from triv ( ∗ ) : B G → ∗ → C a t ∞ \operatorname{triv}(\ast):BG\to\ast\to\mathrm{Cat}_{\infty} triv ( ∗ ) : B G → ∗ → Cat ∞ to C : B G → C a t ∞ \mathcal{C}:BG\to\mathrm{Cat}_{\infty} C : B G → Cat ∞ .
Lemma 3.7. Suppose that C \mathcal{C} C is a G G G -category (i.e. C : B G → C a t ∞ \mathcal{C}:BG\to\mathrm{Cat}_{\infty} C : B G → Cat ∞ ), and that c , c ′ ∈ Map G ( ∗ , C ) c,c'\in\operatorname{Map}_{G}(\ast,\mathcal{C}) c , c ′ ∈ Map G ( ∗ , C ) are G G G -equivariant objects of C \mathcal{C} C . Then there is a natural equivalence
Map C G ( c , c ′ ) ≃ Map C ( c , c ′ ) G , \operatorname{Map}_{\mathcal{C}^{G}}(c,c')\simeq\operatorname{Map}_{\mathcal{C}}(c,c')^{G}, Map C G ( c , c ′ ) ≃ Map C ( c , c ′ ) G , where the G G G -action on Map C ( c , c ′ ) \operatorname{Map}_{\mathcal{C}}(c,c') Map C ( c , c ′ ) is the one defined in § 3.1.1 .
Proof. We begin by outlining some general facts about mapping spaces in limit categories. Consider a diagram of ∞ \infty ∞ -categories, D : I → C a t ∞ D:I\to\mathrm{Cat}_{\infty} D : I → Cat ∞ . Since mapping spaces commute with limits in the second argument, the objects and morphisms of the limit category lim I D \lim_{I}D lim I D are given by lim I Map C a t ∞ ( ∗ , D ( i ) ) \lim_{I}\operatorname{Map}_{\mathrm{Cat}_{\infty}}(\ast,D(i)) lim I Map Cat ∞ ( ∗ , D ( i )) , and lim I Map C a t ∞ ( [ 1 ] , D ( i ) ) \lim_{I}\operatorname{Map}_{\mathrm{Cat}_{\infty}}([1],D(i)) lim I Map Cat ∞ ([ 1 ] , D ( i )) , respectively. The universal property of the limit furnishes an equivalence of spaces between Map A n i ( ∗ , lim I Ob ( D ( i ) ) ) \operatorname{Map}_{\mathrm{Ani}}(\ast,\lim_{I}\operatorname{Ob}(D(i))) Map Ani ( ∗ , lim I Ob ( D ( i ))) and compatible collections of objects in D ( i ) D(i) D ( i ) (here i ∈ I i\in I i ∈ I denotes objects in the index category I I I ). For a given d ∈ Map ( ∗ , lim I Ob ( D ( i ) ) ) d\in\operatorname{Map}(\ast,\lim_{I}\operatorname{Ob}(D(i))) d ∈ Map ( ∗ , lim I Ob ( D ( i ))) , we write (somewhat abusively) d = lim I d ( i ) d=\lim_{I}d(i) d = lim I d ( i ) for d ( i ) d(i) d ( i ) the corresponding object in D ( i ) D(i) D ( i ) . Putting this all together, we see that Map lim I D ( d , d ′ ) ≃ lim I Map D ( i ) ( d ( i ) , d ′ ( i ) ) \operatorname{Map}_{\lim_{I}D}(d,d')\simeq\lim_{I}\operatorname{Map}_{D(i)}(d(i),d'(i)) Map l i m I D ( d , d ′ ) ≃ lim I Map D ( i ) ( d ( i ) , d ′ ( i )) .
Now consider the case of a given G G G -category C : B G → C a t ∞ \mathcal{C}:BG\to\mathrm{Cat}_{\infty} C : B G → Cat ∞ , and let c , c ′ ∈ C G = lim B G C c,c'\in\mathcal{C}^{G}=\lim_{BG}\mathcal{C} c , c ′ ∈ C G = lim B G C . Applying the above discussion to this case, we see that
Map C G ( c , c ′ ) ≃ Map lim B G C ( c , c ′ ) ≃ lim B G Map C ( i ) ( c ( i ) , c ′ ( i ) ) . \operatorname{Map}_{\mathcal{C}^{G}}(c,c')\simeq\operatorname{Map}_{\lim_{BG}\mathcal{C}}(c,c')\simeq\lim_{BG}\operatorname{Map}_{\mathcal{C}(i)}(c(i),c'(i)). Map C G ( c , c ′ ) ≃ Map l i m B G C ( c , c ′ ) ≃ B G lim Map C ( i ) ( c ( i ) , c ′ ( i )) .
Since B G BG B G has only one object ∗ * ∗ (up to isomorphism), this latter space is lim B G Map C ( c , c ′ ) \lim_{BG}\operatorname{Map}_{\mathcal C}(c,c') lim B G Map C ( c , c ′ ) , where we write C \mathcal C C , c c c , and c ′ c' c ′ for C ( ∗ ) \mathcal C(*) C ( ∗ ) , c ( ∗ ) c(*) c ( ∗ ) , and c ′ ( ∗ ) c'(*) c ′ ( ∗ ) by abuse of notation. Thus, Map C ( c , c ′ ) \operatorname{Map}_{\mathcal C}(c,c') Map C ( c , c ′ ) is obtained by taking the limit over a particular diagram B G → A n i BG \to\mathrm{Ani} B G → Ani sending the unique object ∗ ∈ B G * \in BG ∗ ∈ B G to Map C ( c , c ′ ) \operatorname{Map}_{\mathcal C}(c,c') Map C ( c , c ′ ) . Unravelling the construction of this diagram, we see that the functor B G → A n i BG \to\mathrm{Ani} B G → Ani is precisely the one constructed in § 3.1.1 giving the G G G -action on Map C ( c , c ′ ) \operatorname{Map}_{\mathcal C}(c,c') Map C ( c , c ′ ) . The equivalence now follows from the definition of Map C ( c , c ′ ) G \operatorname{Map}_{\mathcal C}(c,c')^G Map C ( c , c ′ ) G , as the limit of this diagram. □
Remark 3.8. Note that there are variants of the above proposition that replace C a t ∞ \mathrm{Cat}_{\infty} Cat ∞ with some suitable category of categories, such as D G C a t k \mathrm{DGCat}_k DGCat k .
We record the following lemma for use later.
Lemma 3.9. Let C \mathcal C C and D \mathcal D D be Perf k k k -linear ∞ \infty ∞ -categories equipped with S 1 S^1 S 1 -actions, and let F : C → D F:\mathcal C \to\mathcal D F : C → D be a fully faithful S 1 S^1 S 1 -equivariant functor. Then the induced functor F t S 1 : C t S 1 → D t S 1 F^{tS^1}:\mathcal C^{tS^1} \to\mathcal D^{tS^1} F t S 1 : C t S 1 → D t S 1 is fully faithful.
Proof. We first show that the induced functor F S 1 : C S 1 → D S 1 F^{S^1}:\mathcal C^{S^1} \to\mathcal D^{S^1} F S 1 : C S 1 → D S 1 is fully faithful. To see this, let a ~ , b ~ ∈ C S 1 \widetilde a,\widetilde b \in\mathcal C^{S^1} a , b ∈ C S 1 , and let a , c ∈ C a,c \in\mathcal C a , c ∈ C denote their images under the canonical functor C S 1 → C \mathcal C^{S^1} \to\mathcal C C S 1 → C . From the discussion in § 3.1.1 , we know that the mapping complex Hom C ( a , b ) \operatorname{Hom}_{\mathcal C}(a,b) Hom C ( a , b ) and Hom D ( F a , F b ) \operatorname{Hom}_{\mathcal D}(Fa,Fb) Hom D ( F a , F b ) have canonical S 1 S^1 S 1 -actions. Moreover, since F : C → D F:\mathcal C \to\mathcal D F : C → D is S 1 S^1 S 1 -equivariant, the natural map Hom C ( a , b ) → Hom D ( F a , F b ) \operatorname{Hom}_{\mathcal C}(a,b) \to\operatorname{Hom}_{\mathcal D}(Fa,Fb) Hom C ( a , b ) → Hom D ( F a , F b ) is S 1 S^1 S 1 -equivariant. By Lemma 3.7 , in order to show that F S 1 F^{S^1} F S 1 is fully faithful, it suffices to show that the induced morphism Hom C ( a , b ) S 1 → Hom D ( F a , F b ) S 1 \operatorname{Hom}_{\mathcal C}(a,b)^{S^1} \to\operatorname{Hom}_{\mathcal D}(Fa,Fb)^{S^1} Hom C ( a , b ) S 1 → Hom D ( F a , F b ) S 1 is an isomorphism.
Let H a , b : B S 1 → A n i H_{a,b}:BS^1 \to\mathrm{Ani} H a , b : B S 1 → Ani and H F a , F b : B S 1 → A n i H_{Fa,Fb}:BS^1 \to\mathrm{Ani} H F a , F b : B S 1 → Ani denote the S 1 S^1 S 1 -actions on Hom C ( a , b ) \operatorname{Hom}_{\mathcal C}(a,b) Hom C ( a , b ) and Hom D ( F a , F b ) \operatorname{Hom}_{\mathcal D}(Fa,Fb) Hom D ( F a , F b ) respectively. We note that the S 1 S^1 S 1 -equivariant map Hom C ( a , b ) → Hom D ( F a , F b ) \operatorname{Hom}_{\mathcal C}(a,b) \to\operatorname{Hom}_{\mathcal D}(Fa,Fb) Hom C ( a , b ) → Hom D ( F a , F b ) is a natural transformation F a , b : H a , b ⟹ H F a , F b F_{a,b}:H_{a,b} \Longrightarrow H_{Fa,Fb} F a , b : H a , b ⟹ H F a , F b . Such a natural transformation is an object in the category, Fun ( [ 1 ] , Fun ( B S 1 , A n i ) ) \operatorname{Fun}([1],\operatorname{Fun}(BS^1,\mathrm{Ani})) Fun ([ 1 ] , Fun ( B S 1 , Ani )) , which under adjunction defines an object of Fun ( B S 1 , Fun ( [ 1 ] , A n i ) ) \operatorname{Fun}(BS^1,\operatorname{Fun}([1],\mathrm{Ani})) Fun ( B S 1 , Fun ([ 1 ] , Ani )) , which is precisely a B S 1 BS^1 B S 1 -shaped diagram in Fun ( [ 1 ] , A n i ) \operatorname{Fun}([1],\mathrm{Ani}) Fun ([ 1 ] , Ani ) . The induced map on S 1 S^1 S 1 -invariants is given by the limit of this diagram,
lim i ∈ B S 1 Hom C ( a , b ) i → Hom D ( F a , F b ) i . \lim_{i\in BS^1}\operatorname{Hom}_{\mathcal C}(a,b)_i \to\operatorname{Hom}_{\mathcal D}(Fa,Fb)_i. i ∈ B S 1 lim Hom C ( a , b ) i → Hom D ( F a , F b ) i .
Note that Hom C ( a , b ) i → Hom D ( F a , F b ) i \operatorname{Hom}_{\mathcal C}(a,b)_i \to\operatorname{Hom}_{\mathcal D}(Fa,Fb)_i Hom C ( a , b ) i → Hom D ( F a , F b ) i is an isomorphism for each the unique object i ∈ B S 1 i \in BS^1 i ∈ B S 1 . By [Lur26, Tag 02YF], the natural transformation is an equivalence if and only if F a , b F_{a,b} F a , b exhibits H a , b H_{a,b} H a , b as a right Kan extension of H F a , F b H_{Fa,Fb} H F a , F b along the identity B S 1 → B S 1 BS^1 \to BS^1 B S 1 → B S 1 . This condition is satisfied precisely when the induced map F a , b ( c ) : H a , b ( c ) → H F a , F b ( c ) F_{a,b}(c):H_{a,b}(c) \to H_{Fa,Fb}(c) F a , b ( c ) : H a , b ( c ) → H F a , F b ( c ) is an equivalence for all objects c ∈ B S 1 c \in BS^1 c ∈ B S 1 . Since there is a unique point ∗ ∈ B S 1 * \in BS^1 ∗ ∈ B S 1 , and since F a , b ( ∗ ) : Hom C ( a , b ) → Hom D ( F a , F b ) F_{a,b}(*) : \operatorname{Hom}_{\mathcal C}(a,b) \to\operatorname{Hom}_{\mathcal D}(Fa,Fb) F a , b ( ∗ ) : Hom C ( a , b ) → Hom D ( F a , F b ) is an equivalence by the assumption that F F F is fully faithful, we see that F a , b F_{a,b} F a , b is actually a natural equivalence of S 1 S^1 S 1 -equivariant animae. Thus, when we apply lim B S 1 : Fun ( B S 1 , A n i ) → A n i \lim_{BS^1}:\operatorname{Fun}(BS^1,\mathrm{Ani}) \to\mathrm{Ani} lim B S 1 : Fun ( B S 1 , Ani ) → Ani , we obtain an equivalence, as desired. Hence, the functor F S 1 : C S 1 → D S 1 F^{S^1}:\mathcal C^{S^1} \to\mathcal D^{S^1} F S 1 : C S 1 → D S 1 is fully faithful.
It remains to show that if A \mathcal A A and B \mathcal B B are small k [ u ] k[u] k [ u ] -linear categories, and if F : A → B F:\mathcal A \to\mathcal B F : A → B is a fully faithful k [ u ] k[u] k [ u ] -linear functor, then F ⊗ k [ u ] k [ u ± 1 ] F\otimes_{k[u]}k[u^{\pm1}] F ⊗ k [ u ] k [ u ± 1 ] is fully faithful. Since A \mathcal A A is small, we have a natural equivalence,
Hom A ( x , y ) ⊗ k [ u ] k [ u ± 1 ] → ∼ Hom A ⊗ k [ u ] k [ u ± 1 ] ( x ⊗ k [ u ± 1 ] , y ⊗ k [ u ± 1 ] ) , \operatorname{Hom}_{\mathcal A}(x,y)\otimes_{k[u]}k[u^{\pm1}] \xrightarrow{\sim} \operatorname{Hom}_{\mathcal A\otimes_{k[u]}k[u^{\pm1}]}(x\otimes k[u^{\pm1}],y\otimes k[u^{\pm1}]), Hom A ( x , y ) ⊗ k [ u ] k [ u ± 1 ] ∼ Hom A ⊗ k [ u ] k [ u ± 1 ] ( x ⊗ k [ u ± 1 ] , y ⊗ k [ u ± 1 ]) ,
for all x , y ∈ A x,y\in\mathcal A x , y ∈ A , and the same for B \mathcal B B . Thus, the induced map on Homs given by F ⊗ k [ u ] k [ u ± 1 ] F\otimes_{k[u]}k[u^{\pm1}] F ⊗ k [ u ] k [ u ± 1 ] is given by base change of the map Hom A ( x , y ) → Hom B ( F x , F y ) \operatorname{Hom}_{\mathcal A}(x,y)\to\operatorname{Hom}_{\mathcal B}(Fx,Fy) Hom A ( x , y ) → Hom B ( F x , F y ) , which by assumption is an isomorphism, so we are done. □
Lemma 3.10. Let C \mathcal C C and D \mathcal D D be small, stable ∞ \infty ∞ -categories equipped with S 1 S^1 S 1 -actions, and let F : C → D F:\mathcal C \to\mathcal D F : C → D be an exact, S 1 S^1 S 1 -equivariant functor. Then
F t S 1 : C t S 1 → D t S 1 F^{tS^1}:\mathcal C^{tS^1} \to\mathcal D^{tS^1} F t S 1 : C t S 1 → D t S 1
is fully faithful if and only if it is fully faithful on the image of the natural functor
θ : C S 1 ⟶ C t S 1 : = C S 1 ⊗ k [ u ] k [ u ± 1 ] . \theta:\mathcal{C}^{S^1}\longrightarrow\mathcal{C}^{tS^1}:=\mathcal{C}^{S^1}\otimes_{k[u]} k[u^{\pm1}]. θ : C S 1 ⟶ C t S 1 := C S 1 ⊗ k [ u ] k [ u ± 1 ] .
Proof. By [Per18, Lemma 2.7], the image of θ \theta θ generates C t S 1 \mathcal{C}^{tS^1} C t S 1 . Thus, it suffices to show that
Hom C t S 1 ( θ c , θ c ′ ) ⟶ Hom D t S 1 ( F t S 1 ( θ c ) , F t S 1 ( θ c ′ ) ) \operatorname{Hom}_{\mathcal{C}^{tS^1}}(\theta c,\theta c')\longrightarrow\operatorname{Hom}_{\mathcal{D}^{tS^1}}(F^{tS^1}(\theta c),F^{tS^1}(\theta c')) Hom C t S 1 ( θ c , θ c ′ ) ⟶ Hom D t S 1 ( F t S 1 ( θ c ) , F t S 1 ( θ c ′ ))
is an isomorphism for all c , c ′ ∈ C S 1 c,c'\in\mathcal{C}^{S^1} c , c ′ ∈ C S 1 . For fixed c ∈ C S 1 c\in\mathcal{C}^{S^1} c ∈ C S 1 , the objects c u ′ ∈ C t S 1 c'_u\in\mathcal{C}^{tS^1} c u ′ ∈ C t S 1 such that
Hom ( θ c , c u ′ ) ⟶ Hom ( F t S 1 ( θ c ) , F t S 1 ( c u ′ ) ) \operatorname{Hom}(\theta c,c'_u)\longrightarrow\operatorname{Hom}(F^{tS^1}(\theta c),F^{tS^1}(c'_u)) Hom ( θ c , c u ′ ) ⟶ Hom ( F t S 1 ( θ c ) , F t S 1 ( c u ′ ))
is an isomorphism form a thick subcategory. Making the same argument for fixed c u ′ c'_u c u ′ extends the isomorphism from all θ c \theta c θ c to all objects c u ∈ C t S 1 c_u\in\mathcal{C}^{tS^1} c u ∈ C t S 1 . □
Lemma 3.11. Let R R R be a k k k -algebra. Let C , D ∈ S 1 –mod ( DGCat R ) \mathcal{C},\mathcal{D}\in S^1\text{--mod}(\operatorname{DGCat}_R) C , D ∈ S 1 –mod ( DGCat R ) such that D \mathcal{D} D is compactly generated (hence dualizable) and equipped with the trivial S 1 S^1 S 1 -action. Then
( C ⊗ R D ) t S 1 ≃ C t S 1 ⊗ R [ u ± 1 ] D [ u ± 1 ] . (\mathcal{C}\otimes_R\mathcal{D})^{tS^1}\simeq\mathcal{C}^{tS^1}\otimes_{R[u^{\pm1}]}\mathcal{D}[u^{\pm1}]. ( C ⊗ R D ) t S 1 ≃ C t S 1 ⊗ R [ u ± 1 ] D [ u ± 1 ] .
Proof. The symmetric monoidal structure on S 1 –mod ( DGCat R ) : = Fun ( B S 1 , DGCat R ) S^1\text{--mod}(\operatorname{DGCat}_R):=\operatorname{Fun}(BS^1,\operatorname{DGCat}_R) S 1 –mod ( DGCat R ) := Fun ( B S 1 , DGCat R ) is taken to be the pointwise one. Under this structure, the S 1 S^1 S 1 -action on C ⊗ R D \mathcal{C}\otimes_R\mathcal{D} C ⊗ R D is given by the restriction of the diagram
Π : B S 1 × B S 1 → C ( − ) × D ( − ) DGCat R × DGCat R → ⊗ R DGCat R \Pi:BS^1\times BS^1\xrightarrow{\mathcal{C}(-)\times\mathcal{D}(-)}\operatorname{DGCat}_R\times\operatorname{DGCat}_R\xrightarrow{\otimes_R}\operatorname{DGCat}_R Π : B S 1 × B S 1 C ( − ) × D ( − ) DGCat R × DGCat R ⊗ R DGCat R
along the diagonal functor Δ : B S 1 → B S 1 × B S 1 \Delta:BS^1\to BS^1\times BS^1 Δ : B S 1 → B S 1 × B S 1 . Since D \mathcal{D} D is equipped with the trivial S 1 S^1 S 1 -action, the diagram Π : B S 1 × B S 1 → DGCat R \Pi:BS^1\times BS^1\to\operatorname{DGCat}_R Π : B S 1 × B S 1 → DGCat R factors through the map B S 1 × B S 1 → B S 1 × ∗ ≃ B S 1 BS^1\times BS^1\to BS^1\times*\simeq BS^1 B S 1 × B S 1 → B S 1 × ∗ ≃ B S 1 . Thus, ( C ⊗ R D ) S 1 ≃ lim i ∈ B S 1 ( C ( i ) ⊗ R D ) (\mathcal{C}\otimes_R\mathcal{D})^{S^1}\simeq\lim_{i\in BS^1}(\mathcal{C}(i)\otimes_R\mathcal{D}) ( C ⊗ R D ) S 1 ≃ lim i ∈ B S 1 ( C ( i ) ⊗ R D ) . Since D \mathcal{D} D is dualizable, C ⊗ R D ≃ Fun R ( D ∨ , C ) \mathcal{C}\otimes_R\mathcal{D}\simeq\operatorname{Fun}_R(\mathcal{D}^{\vee},\mathcal{C}) C ⊗ R D ≃ Fun R ( D ∨ , C ) , where D ∨ \mathcal{D}^{\vee} D ∨ is the dual to D \mathcal{D} D . Since Fun R ( − , − ) \operatorname{Fun}_R(-,-) Fun R ( − , − ) preserves limits in the second argument, we have
lim B S 1 Fun R ( D ∨ , C ) ≃ Fun R ( D ∨ , lim B S 1 C ) ≃ Fun R ( D ∨ , C S 1 ) ≃ C S 1 ⊗ R D ≃ C S 1 ⊗ R [ u ] R [ u ] –mod ⊗ R D ≃ C S 1 ⊗ R [ u ] D [ u ] , \begin{aligned}
\lim_{BS^1}\operatorname{Fun}_R(\mathcal{D}^{\vee},\mathcal{C})&\simeq\operatorname{Fun}_R\left(\mathcal{D}^{\vee},\lim_{BS^1}\mathcal{C}\right)\\
&\simeq\operatorname{Fun}_R(\mathcal{D}^{\vee},\mathcal{C}^{S^1})\\
&\simeq\mathcal{C}^{S^1}\otimes_R\mathcal{D}\\
&\simeq\mathcal{C}^{S^1}\otimes_{R[u]}R[u]\text{--mod}\otimes_R\mathcal{D}\\
&\simeq\mathcal{C}^{S^1}\otimes_{R[u]}\mathcal{D}[u],
\end{aligned} B S 1 lim Fun R ( D ∨ , C ) ≃ Fun R ( D ∨ , B S 1 lim C ) ≃ Fun R ( D ∨ , C S 1 ) ≃ C S 1 ⊗ R D ≃ C S 1 ⊗ R [ u ] R [ u ] –mod ⊗ R D ≃ C S 1 ⊗ R [ u ] D [ u ] ,
where we have used that the R R R -linear structure on C S 1 \mathcal{C}^{S^1} C S 1 factors through its R [ u ] R[u] R [ u ] -linear structure coming from the action of ( R –mod ) S 1 (R\text{--mod})^{S^1} ( R –mod ) S 1 .
By definition, ( C ⊗ R D ) t S 1 : = ( C ⊗ R D ) S 1 ⊗ R [ u ] R [ u ± 1 ] (\mathcal{C}\otimes_R\mathcal{D})^{tS^1}:=(\mathcal{C}\otimes_R\mathcal{D})^{S^1}\otimes_{R[u]}R[u^{\pm1}] ( C ⊗ R D ) t S 1 := ( C ⊗ R D ) S 1 ⊗ R [ u ] R [ u ± 1 ] , so from the identifications above, we have,
( C ⊗ R D ) t S 1 ≃ ( C S 1 ⊗ R [ u ] D [ u ] ) ⊗ R [ u ] R [ u ± 1 ] ≃ ( C S 1 ⊗ R [ u ] R [ u ± 1 ] ) ⊗ R [ u ± 1 ] ( D [ u ] ⊗ R [ u ] R [ u ± 1 ] ) ≃ C t S 1 ⊗ R [ u ± 1 ] D [ u ± 1 ] , \begin{aligned}
(\mathcal{C}\otimes_R\mathcal{D})^{tS^1}&\simeq(\mathcal{C}^{S^1}\otimes_{R[u]}\mathcal{D}[u])\otimes_{R[u]}R[u^{\pm1}]\\
&\simeq(\mathcal{C}^{S^1}\otimes_{R[u]}R[u^{\pm1}])\otimes_{R[u^{\pm1}]}(\mathcal{D}[u]\otimes_{R[u]}R[u^{\pm1}])\\
&\simeq\mathcal{C}^{tS^1}\otimes_{R[u^{\pm1}]}\mathcal{D}[u^{\pm1}],
\end{aligned} ( C ⊗ R D ) t S 1 ≃ ( C S 1 ⊗ R [ u ] D [ u ]) ⊗ R [ u ] R [ u ± 1 ] ≃ ( C S 1 ⊗ R [ u ] R [ u ± 1 ]) ⊗ R [ u ± 1 ] ( D [ u ] ⊗ R [ u ] R [ u ± 1 ]) ≃ C t S 1 ⊗ R [ u ± 1 ] D [ u ± 1 ] ,
where we have used the fact that the base change functor DGCat R [ u ] → DGCat R [ u ± 1 ] \operatorname{DGCat}_{R[u]}\to\operatorname{DGCat}_{R[u^{\pm1}]} DGCat R [ u ] → DGCat R [ u ± 1 ] is symmetric monoidal (i.e. preserves relative tensor products). □
We record the following result of [BZCHN].
Theorem 3.12 ([BZCHN]). Let
A → F B → G C F R ↑ G R ↑ \begin{CD}
A @>F>> B @>G>> C\\
@A{F^R}AA @A{G^R}AA
\end{CD} A F R ⏐ ↑ F B G R ⏐ ↑ G C
be a localization sequence in Fun ( B S 1 , C o h t ) \operatorname{Fun}(BS^1,\mathbf{Coh}_{t}) Fun ( B S 1 , Coh t ) such that the right adjoints F R F^{R} F R and G R G^{R} G R are t t t -bounded. Then the sequence after applying the t t t Tate construction is a localization sequence in k [ u ± 1 ] k[u^{\pm1}] k [ u ± 1 ] –mod.
Group actions and t t t -structures
Definition 3.13. Let G ∈ Grp ( Spc ) G \in\operatorname{Grp}(\operatorname{Spc}) G ∈ Grp ( Spc ) , and suppose given a G G G -action on the underlying ∞ \infty ∞ -category C \mathcal{C} C of an object in Cat ∞ t \operatorname{Cat}^{t}_{\infty} Cat ∞ t . We say that the group action is compatible with the t t t -structure on C \mathcal{C} C if the functor Act : B G → Cat ∞ \operatorname{Act}: BG \to\operatorname{Cat}_{\infty} Act : B G → Cat ∞ encoding the G G G -action factors through the forgetful functor oblv t : Cat ∞ t → Cat ∞ \operatorname{oblv}^{t}: \operatorname{Cat}^{t}_{\infty} \to\operatorname{Cat}_{\infty} oblv t : Cat ∞ t → Cat ∞ .
In other words, the group action is compatible with the t t t -structure on C \mathcal{C} C if Act ( g ) \operatorname{Act}(g) Act ( g ) is a t t t -exact autoequivalence for each morphism g ∈ B G g \in BG g ∈ B G .
Lemma 3.14. Let ( C , C ≤ 0 ) ∈ Cat ∞ t (\mathcal{C},\mathcal{C}_{\leq0}) \in\operatorname{Cat}^{t}_{\infty} ( C , C ≤ 0 ) ∈ Cat ∞ t , i.e. a stable ∞ \infty ∞ -category C \mathcal{C} C equipped with a t t t -structure. Let Act ∈ Fun ( B G , Cat ∞ ) \operatorname{Act} \in\operatorname{Fun}(BG,\operatorname{Cat}_{\infty}) Act ∈ Fun ( B G , Cat ∞ ) be a G G G -action on C \mathcal{C} C , i.e. Act ( ∗ ) ≃ C \operatorname{Act}(\ast) \simeq\mathcal{C} Act ( ∗ ) ≃ C . If the G G G -action on C \mathcal{C} C is compatible with its t t t -structure in the sense of Definition 3.13 , then the truncation functors τ ≤ n \tau_{\leq n} τ ≤ n and τ ≥ n \tau_{\geq n} τ ≥ n are canonically G G G -equivariant for any n n n .
Proof. Let Fun f f ( [ 1 ] , Cat ∞ ) \operatorname{Fun}_{ff}([1],\operatorname{Cat}_{\infty}) Fun f f ([ 1 ] , Cat ∞ ) denote the full subcategory of Fun ( [ 1 ] , Cat ∞ ) \operatorname{Fun}([1],\operatorname{Cat}_{\infty}) Fun ([ 1 ] , Cat ∞ ) spanned by the fully faithful functors. An object of this category may be seen as pair ( C , C ′ ) (\mathcal{C},\mathcal{C}^{\prime}) ( C , C ′ ) consisting of a category C \mathcal{C} C and a full subcategory C ′ ⊂ C \mathcal{C}^{\prime} \subset\mathcal{C} C ′ ⊂ C . Then Cat ∞ t \operatorname{Cat}^{t}_{\infty} Cat ∞ t may be viewed as the full subcategory of Fun f f ( [ 1 ] , Cat ∞ ) \operatorname{Fun}_{ff}([1],\operatorname{Cat}_{\infty}) Fun f f ([ 1 ] , Cat ∞ ) spanned by pairs ( C , C ′ ) (\mathcal{C},\mathcal{C}^{\prime}) ( C , C ′ ) such that C \mathcal{C} C is stable, and C ′ \mathcal{C}^{\prime} C ′ together with the left orthogonal complement of C ′ [ − 1 ] \mathcal{C}^{\prime}[-1] C ′ [ − 1 ] satisfy the properties of a t t t -structure ([HA, Definition 1.2.1.1]) with C ≤ 0 : = C ′ \mathcal{C}_{\leq0} := \mathcal{C}^{\prime} C ≤ 0 := C ′ and C ≥ 0 : = ⊥ C ′ [ − 1 ] \mathcal{C}_{\geq0} := {}^{\perp}\mathcal{C}^{\prime}[-1] C ≥ 0 := ⊥ C ′ [ − 1 ] . From this description, it is clear that the source and target functors ev 0 : Fun ( [ 1 ] , Cat ∞ ) → Cat ∞ \operatorname{ev}_{0}: \operatorname{Fun}([1],\operatorname{Cat}_{\infty}) \to\operatorname{Cat}_{\infty} ev 0 : Fun ([ 1 ] , Cat ∞ ) → Cat ∞ and ev 1 : Fun ( [ 1 ] , Cat ∞ ) → Cat ∞ \operatorname{ev}_{1}: \operatorname{Fun}([1],\operatorname{Cat}_{\infty}) \to\operatorname{Cat}_{\infty} ev 1 : Fun ([ 1 ] , Cat ∞ ) → Cat ∞ restricts to the functors,
pr ≤ n : Cat ∞ t → Cat ∞ and oblv : Cat ∞ t → Cat ∞ \operatorname{pr}_{\leq n}: \operatorname{Cat}^{t}_{\infty} \to\operatorname{Cat}_{\infty}
\qquad\text{and}\qquad
\operatorname{oblv}: \operatorname{Cat}^{t}_{\infty} \to\operatorname{Cat}_{\infty} pr ≤ n : Cat ∞ t → Cat ∞ and oblv : Cat ∞ t → Cat ∞
which send a category C \mathcal{C} C equipped with t t t -structure to the subcategory of n n n -connective objects C ≤ n \mathcal{C}_{\leq n} C ≤ n and C \mathcal{C} C , respectively.
Denote the lift of Act \operatorname{Act} Act to Cat ∞ t \operatorname{Cat}^{t}_{\infty} Cat ∞ t by Act ~ : B G → Cat ∞ t \widetilde{\operatorname{Act}}: BG \to\operatorname{Cat}^{t}_{\infty} Act : B G → Cat ∞ t , and define Act ≤ n : = pr ≤ n ∘ Act ~ \operatorname{Act}_{\leq n} := \operatorname{pr}_{\leq n} \circ\widetilde{\operatorname{Act}} Act ≤ n := pr ≤ n ∘ Act . Under the adjunction,
Fun ( Fun ( [ 1 ] , Cat ∞ ) × [ 1 ] , Cat ∞ ) ≃ Fun ( [ 1 ] , Fun ( Fun ( [ 1 ] , Cat ∞ ) , Cat ∞ ) ) , \operatorname{Fun}(\operatorname{Fun}([1],\operatorname{Cat}_{\infty}) \times[1],\operatorname{Cat}_{\infty}) \simeq\operatorname{Fun}([1],\operatorname{Fun}(\operatorname{Fun}([1],\operatorname{Cat}_{\infty}),\operatorname{Cat}_{\infty})), Fun ( Fun ([ 1 ] , Cat ∞ ) × [ 1 ] , Cat ∞ ) ≃ Fun ([ 1 ] , Fun ( Fun ([ 1 ] , Cat ∞ ) , Cat ∞ )) ,
the evaluation functor ev : Fun ( [ 1 ] , Cat ∞ ) × [ 1 ] → Cat ∞ \operatorname{ev}: \operatorname{Fun}([1],\operatorname{Cat}_{\infty}) \times[1] \to\operatorname{Cat}_{\infty} ev : Fun ([ 1 ] , Cat ∞ ) × [ 1 ] → Cat ∞ corresponds to a natural transformation ev 0 ⟹ ev 1 \operatorname{ev}_{0} \Longrightarrow\operatorname{ev}_{1} ev 0 ⟹ ev 1 , which restricts to a natural transformation pr ≤ n ⟹ oblv \operatorname{pr}_{\leq n} \Longrightarrow\operatorname{oblv} pr ≤ n ⟹ oblv . Whiskering this natural transformation with the functor Act ~ \widetilde{\operatorname{Act}} Act , we obtain a natural transformation i ≤ n : Act ≤ n → Act i_{\leq n}: \operatorname{Act}_{\leq n} \to\operatorname{Act} i ≤ n : Act ≤ n → Act , whose component at the unique point ∗ ∈ B G \ast\in BG ∗ ∈ B G is the inclusion i ≤ n : C ≤ n ↪ C i_{\leq n}: \mathcal{C}_{\leq n} \hookrightarrow\mathcal{C} i ≤ n : C ≤ n ↪ C . Taking right adjoints, we get a natural transformation τ ≤ n : Act ⟹ Act ≤ n \tau_{\leq n}: \operatorname{Act} \Longrightarrow\operatorname{Act}_{\leq n} τ ≤ n : Act ⟹ Act ≤ n , whose component at ∗ ∈ B G \ast\in BG ∗ ∈ B G is the ordinary truncation functor τ ≤ n \tau_{\leq n} τ ≤ n . Thus, we have constructed a canonical G G G -equivariant structure on τ ≤ n \tau_{\leq n} τ ≤ n for any n ∈ Z n \in\mathbb{Z} n ∈ Z .
The proof for τ ≥ n \tau_{\geq n} τ ≥ n proceeds in exactly the same manner, starting with the data of the coconnective subcategory C ≥ 0 ⊂ C \mathcal{C}_{\geq0} \subset\mathcal{C} C ≥ 0 ⊂ C , rather than C ≤ 0 \mathcal{C}_{\leq0} C ≤ 0 .
Given ( C , C ≤ 0 ) ∈ Cat ∞ t (\mathcal{C},\mathcal{C}_{\leq0}) \in\operatorname{Cat}^{t}_{\infty} ( C , C ≤ 0 ) ∈ Cat ∞ t , and a G G G -action on C \mathcal{C} C compatible with its t t t -structure, we may equip the G G G -invariants C G : = lim B G Act \mathcal{C}^{G} := \lim_{BG}\operatorname{Act} C G := lim B G Act with a canonical t t t -structure (cf. [HP23, Lemma 2.1.9(4)]). Indeed, let U : C G → C U: \mathcal{C}^{G} \to\mathcal{C} U : C G → C be the canonical functor. We may define C ≤ 0 G \mathcal{C}^{G}_{\leq0} C ≤ 0 G and C ≥ 0 G ⊂ C G \mathcal{C}^{G}_{\geq0} \subset\mathcal{C}^{G} C ≥ 0 G ⊂ C G , respectively, as the full subcategory of C G \mathcal{C}^{G} C G spanned by objects c ∈ C G c \in\mathcal{C}^{G} c ∈ C G such that U c ∈ C ≤ 0 Uc \in\mathcal{C}_{\leq0} U c ∈ C ≤ 0 and U c ∈ C ≥ 0 Uc \in\mathcal{C}_{\geq0} U c ∈ C ≥ 0 , respectively. Because morphisms in B G BG B G act by t t t -exact autoequivalences, these subcategories determine a t t t -structure on C G \mathcal{C}^{G} C G such that the projection functor C G = lim B G Act → C \mathcal{C}^{G} = \lim_{BG}\operatorname{Act} \to\mathcal{C} C G = lim B G Act → C is t t t -exact.
In particular, we have that ( C G ) ∨ (\mathcal{C}^{G})^{\vee} ( C G ) ∨ is the full subcategory of C G \mathcal{C}^{G} C G spanned by objects c c c such that U c ∈ C ∨ Uc \in\mathcal{C}^{\vee} U c ∈ C ∨ . On the other hand, one can consider the G G G -invariants ( C ∨ ) G : = lim B G Act ∨ (\mathcal{C}^{\vee})^{G} := \lim_{BG}\operatorname{Act}^{\vee} ( C ∨ ) G := lim B G Act ∨ , where Act ∨ \operatorname{Act}^{\vee} Act ∨ is obtained from the pair ( C , C ∨ ) (\mathcal{C},\mathcal{C}^{\vee}) ( C , C ∨ ) in the same manner as Act ≤ n \operatorname{Act}_{\leq n} Act ≤ n in the proof of Lemma 3.14 above. We also obtain a natural transformation Act ♡ ⟹ Act \operatorname{Act}^{\heartsuit} \Longrightarrow\operatorname{Act} Act ♡ ⟹ Act in the same manner as we obtained the natural transformation Act ≤ n ⟹ Act \operatorname{Act}_{\leq n} \Longrightarrow\operatorname{Act} Act ≤ n ⟹ Act above, whose component at ∗ ∈ B G \ast\in BG ∗ ∈ B G is the inclusion C ♡ ↪ C \mathcal{C}^{\heartsuit} \hookrightarrow\mathcal{C} C ♡ ↪ C . Taking G G G -invariants, we thus obtain from this natural transformation a functor,
( C ♡ ) G ⟶ C G (\mathcal{C}^{\heartsuit})^{G} \longrightarrow\mathcal{C}^{G} ( C ♡ ) G ⟶ C G
which, from our definition of the t t t -structure on C G \mathcal{C}^{G} C G , factors canonically through the inclusion ( C G ) ♡ ↪ C G (\mathcal{C}^{G})^{\heartsuit} \hookrightarrow\mathcal{C}^{G} ( C G ) ♡ ↪ C G .
Lemma 3.15. The canonical functor ( C ♡ ) G ⟶ ( C G ) ♡ (\mathcal{C}^{\heartsuit})^{G} \longrightarrow(\mathcal{C}^{G})^{\heartsuit} ( C ♡ ) G ⟶ ( C G ) ♡ is an equivalence.
Proof. By definition, we have a G G G -equivariant equivalence, C ♡ ≃ C ≤ 0 × C C ≥ 0 \mathcal{C}^{\heartsuit} \simeq\mathcal{C}_{\leq0} \mathop{\times}_{\mathcal{C}} \mathcal{C}_{\geq0} C ♡ ≃ C ≤ 0 × C C ≥ 0 . The operation of taking G G G -invariants, being a limit, commutes with products, so we have,
( C ♡ ) G ≃ ( C ≤ 0 ) G × C G ( C ≥ 0 ) G . (\mathcal{C}^{\heartsuit})^{G} \simeq(\mathcal{C}_{\leq0})^{G} \mathop{\times}_{\mathcal{C}^{G}} (\mathcal{C}_{\geq0})^{G}. ( C ♡ ) G ≃ ( C ≤ 0 ) G × C G ( C ≥ 0 ) G . On the other hand, we have commutative diagrams,
C G → U C ↑ ↑ ( C ≤ 0 ) G → U ≤ 0 C ≤ 0 C G → U C ↑ ↑ ( C ≥ 0 ) G → U ≥ 0 C ≥ 0 \begin{CD}
\mathcal{C}^{G} @>U>> \mathcal{C} \\
@A{}AA @A{}AA \\
(\mathcal{C}_{\leq0})^{G} @>U_{\leq0}>> \mathcal{C}_{\leq0}
\end{CD}
\qquad
\begin{CD}
\mathcal{C}^{G} @>U>> \mathcal{C} \\
@A{}AA @A{}AA \\
(\mathcal{C}_{\geq0})^{G} @>U_{\geq0}>> \mathcal{C}_{\geq0}
\end{CD} C G ⏐ ↑ ( C ≤ 0 ) G U U ≤ 0 C ⏐ ↑ C ≤ 0 C G ⏐ ↑ ( C ≥ 0 ) G U U ≥ 0 C ⏐ ↑ C ≥ 0 implying that ( C ≤ 0 ) G ≃ U − 1 ( C ≤ 0 ) = : ( C G ) ≤ 0 (\mathcal{C}_{\leq0})^{G} \simeq U^{-1}(\mathcal{C}_{\leq0}) =: (\mathcal{C}^{G})_{\leq0} ( C ≤ 0 ) G ≃ U − 1 ( C ≤ 0 ) =: ( C G ) ≤ 0 , and ( C ≥ 0 ) G ≃ U − 1 ( C ≥ 0 ) = : ( C G ) ≥ 0 (\mathcal{C}_{\geq0})^{G} \simeq U^{-1}(\mathcal{C}_{\geq0}) =: (\mathcal{C}^{G})_{\geq0} ( C ≥ 0 ) G ≃ U − 1 ( C ≥ 0 ) =: ( C G ) ≥ 0 . Now, by the definition of the heart, we also have ( C G ) ♡ = ( C G ) ≤ 0 × C G ( C G ) ≥ 0 (\mathcal{C}^{G})^{\heartsuit} = (\mathcal{C}^{G})_{\leq0} \mathop{\times}_{\mathcal{C}^{G}} (\mathcal{C}^{G})_{\geq0} ( C G ) ♡ = ( C G ) ≤ 0 × C G ( C G ) ≥ 0 . Combining these equivalences, we thus obtain
( C ♡ ) G ≃ ( C G ) ♡ . (\mathcal{C}^{\heartsuit})^{G} \simeq(\mathcal{C}^{G})^{\heartsuit}. ( C ♡ ) G ≃ ( C G ) ♡ .
Related notions of G G G -action
Let G G G be a group object in S p c \mathrm{Spc} Spc , and consider the constant stack G ∈ S t k k G \in\mathrm{Stk}_{k} G ∈ Stk k with value G G G , which by abuse of notation we sometimes denote by G G G . G G G is a group object in S t k k \mathrm{Stk}_{k} Stk k , so the dg-category Q C o h ( G ) \mathrm{QCoh}(G) QCoh ( G ) obtains a comonoidal structure under pullback along the group multiplication maps. One often sees the following alternative definition of G G G -action on C \mathcal{C} C in the literature:
Definition 3.16. Let G ∈ G r p ( S t k k ) G \in\mathrm{Grp}(\mathrm{Stk}_{k}) G ∈ Grp ( Stk k ) , and let C ∈ d g C a t k \mathcal{C} \in\mathrm{dgCat}_{k} C ∈ dgCat k . Then a G G G -action on C \mathcal{C} C is the data of a P e r f ( G ) \mathrm{Perf}(G) Perf ( G ) -comodule structure on C \mathcal{C} C . Let D ∈ D G C a t k \mathcal{D} \in\mathrm{DGCat}_{k} D ∈ DGCat k . Then a G G G -action on D \mathcal{D} D is the data of a Q C o h ( G ) \mathrm{QCoh}(G) QCoh ( G ) -comodule structure on D \mathcal{D} D .
In fact, this alternative definition is equivalent to the one we gave in Definition 3.1 in the case when G G G is the Betti stack associated to group object in S p c \mathrm{Spc} Spc .
Proposition 3.17. Let G ∈ G r p ( S p c ) G \in\mathrm{Grp}(\mathrm{Spc}) G ∈ Grp ( Spc ) , and let C ∈ Q C o h ( G ) – c o m o d ( D G C a t k ) \mathcal{C} \in\mathrm{QCoh}(G)\text{--}\mathrm{comod}(\mathrm{DGCat}_{k}) C ∈ QCoh ( G ) – comod ( DGCat k ) . Then the underlying dg-category of C \mathcal{C} C has a canonical G G G -action in the sense of Definition 3.1. Conversely, if the underlying dg-category of C \mathcal{C} C has a G G G -action, then this induces a canonical Q C o h ( G ) \mathrm{QCoh}(G) QCoh ( G ) -comodule structure on C \mathcal{C} C .
Proof. First note that the constant stack G G G is the colimit of the constant functor G → ∗ → S t k k G \to\ast\to\mathrm{Stk}_{k} G → ∗ → Stk k . On the other hand, since Q C o h : S t k k → D G C a t k \mathrm{QCoh} : \mathrm{Stk}_{k} \to\mathrm{DGCat}_{k} QCoh : Stk k → DGCat k takes colimits to limits, Q C o h ( G ) ≃ lim G Q C o h ( ∗ ) ≃ F u n ( G , V e c t k ) \mathrm{QCoh}(G) \simeq\lim_{G} \mathrm{QCoh}(\ast) \simeq\mathrm{Fun}(G,\mathrm{Vect}_{k}) QCoh ( G ) ≃ lim G QCoh ( ∗ ) ≃ Fun ( G , Vect k ) .
Consider the functor L : F u n ( B G , D G C a t k ) → D G C a t k L : \mathrm{Fun}(BG,\mathrm{DGCat}_{k}) \to\mathrm{DGCat}_{k} L : Fun ( B G , DGCat k ) → DGCat k given by evaluation at the unique point in B G BG B G . Note that L L L satisfies the conditions of the Barr--Beck--Lurie theorem: it is conservative and has both a right adjoint R : D G C a t k → F u n ( B G , D G C a t k ) R : \mathrm{DGCat}_{k} \to\mathrm{Fun}(BG,\mathrm{DGCat}_{k}) R : DGCat k → Fun ( B G , DGCat k ) given by cotensoring with G G G (C ↦ F u n ( G , C ) \mathcal{C} \mapsto\mathrm{Fun}(G,\mathcal{C}) C ↦ Fun ( G , C ) ) and a left adjoint given by tensoring with G G G . It remains to identify the comonad L ∘ R L \circ R L ∘ R as the operation of tensoring with Q C o h ( G ) ≃ F u n ( G , V e c t k ) \mathrm{QCoh}(G) \simeq\mathrm{Fun}(G,\mathrm{Vect}_{k}) QCoh ( G ) ≃ Fun ( G , Vect k ) , which we do by proving that F u n ( G , C ) ≃ F u n ( G , V e c t k ) ⊗ k C \mathrm{Fun}(G,\mathcal{C}) \simeq\mathrm{Fun}(G,\mathrm{Vect}_{k}) \otimes_{k} \mathcal{C} Fun ( G , C ) ≃ Fun ( G , Vect k ) ⊗ k C .
Let K K K be any (small) anima. Recall that P ( K o p ) o p = Fun ( K o p ) o p \mathcal{P}(K^{\mathrm{op}})^{\mathrm{op}}=\operatorname{Fun}(K^{\mathrm{op}})^{\mathrm{op}} P ( K op ) op = Fun ( K op ) op is the category obtained by freely adjoining limits to K K K , meaning it satisfies the universal property,
Fun ( K , C ) ≃ Fun R ( P ( K o p ) o p , C ) , \operatorname{Fun}(K,\mathcal{C})\simeq\operatorname{Fun}^{R}(\mathcal{P}(K^{\mathrm{op}})^{\mathrm{op}},\mathcal{C}), Fun ( K , C ) ≃ Fun R ( P ( K op ) op , C ) ,
for any cocomplete category C \mathcal{C} C . Suppose that C \mathcal{C} C is also presentable. Then, by the definition of the Lurie tensor product on presentable ∞ \infty ∞ -categories, we have
Fun ( K , C ) ≃ Fun R ( P ( K o p ) o p , C ) ≃ P ( K ) ⊗ C , \operatorname{Fun}(K,\mathcal{C})\simeq\operatorname{Fun}^{R}(\mathcal{P}(K^{\mathrm{op}})^{\mathrm{op}},\mathcal{C})\simeq\mathcal{P}(K)\otimes\mathcal{C}, Fun ( K , C ) ≃ Fun R ( P ( K op ) op , C ) ≃ P ( K ) ⊗ C ,
where we use that K ≃ K o p K\simeq K^{\mathrm{op}} K ≃ K op and that P ( K ) \mathcal{P}(K) P ( K ) is presentable.
In particular, taking K = G K=G K = G and C = Vect k \mathcal{C}=\operatorname{Vect}_k C = Vect k , we obtain the equivalence Fun ( G , Vect k ) ≃ P ( G ) ⊗ Vect k \operatorname{Fun}(G,\operatorname{Vect}_k)\simeq\mathcal{P}(G)\otimes\operatorname{Vect}_k Fun ( G , Vect k ) ≃ P ( G ) ⊗ Vect k . To conclude, we observe that, for any C ∈ DGCat k \mathcal{C}\in\operatorname{DGCat}_k C ∈ DGCat k , we have
P ( G ) ⊗ C ≃ ( P ( G ) ⊗ Vect k ) ⊗ k C ≃ Fun ( G , Vect k ) ⊗ k C ≃ QCoh ( G ) ⊗ k C . \begin{aligned}
\mathcal{P}(G)\otimes\mathcal{C}\simeq(\mathcal{P}(G)\otimes\operatorname{Vect}_k)\otimes_k\mathcal{C}\simeq\operatorname{Fun}(G,\operatorname{Vect}_k)\otimes_k\mathcal{C} \\
&\simeq\operatorname{QCoh}(G)\otimes_k\mathcal{C}.
\end{aligned} P ( G ) ⊗ C ≃ ( P ( G ) ⊗ Vect k ) ⊗ k C ≃ Fun ( G , Vect k ) ⊗ k C ≃ QCoh ( G ) ⊗ k C .
Thus, the comonad structure on DGCat k \operatorname{DGCat}_k DGCat k is the coalgebra structure on QCoh ( G ) \operatorname{QCoh}(G) QCoh ( G ) . □
For any group prestack G ∈ Grp ( PreStk ) G\in\operatorname{Grp}(\operatorname{PreStk}) G ∈ Grp ( PreStk ) , let B G ∙ BG^\bullet B G ∙ denote the bar construction of G G G . If G ∈ Grp ( PreStk ) G\in\operatorname{Grp}(\operatorname{PreStk}) G ∈ Grp ( PreStk ) has the property that the functor Loc G : QCoh ( G ) – mod → ShvCat ( G ) \operatorname{Loc}_G:\operatorname{QCoh}(G)\text{--}\operatorname{mod}\to\operatorname{ShvCat}(G) Loc G : QCoh ( G ) – mod → ShvCat ( G ) 18 (i.e. the left adjoint to enhanced global sections Γ G e n h \Gamma_G^{\mathrm{enh}} Γ G enh ) is fully faithful, we have
QCoh ( G ) – comod ( DGCat k ) ≃ Tot ( QCoh ( B G ∙ ) – mod ) (3) \operatorname{QCoh}(G)\text{--}\operatorname{comod}(\operatorname{DGCat}_k)\simeq\operatorname{Tot}(\operatorname{QCoh}(BG^\bullet)\text{--}\operatorname{mod}) \tag*{(3)} QCoh ( G ) – comod ( DGCat k ) ≃ Tot ( QCoh ( B G ∙ ) – mod ) (3)
≃ Tot ( ShvCat ( B G ∙ ) ) (4) \simeq\operatorname{Tot}(\operatorname{ShvCat}(BG^\bullet)) \tag*{(4)} ≃ Tot ( ShvCat ( B G ∙ )) (4)
≃ ShvCat ( B G ) , (5) \simeq\operatorname{ShvCat}(BG), \tag*{(5)} ≃ ShvCat ( B G ) , (5)
where equivalence (3) is the content of [Gai15, Corollary 10.1.5]; equivalence (4) is the content of [Gai15, Proposition 10.1.3]; and equivalence (5) follows from the definition of ShvCat ( B G ) \operatorname{ShvCat}(BG) ShvCat ( B G ) . Note that B G BG B G denotes the delooping of G G G in PreStk k \operatorname{PreStk}_k PreStk k (as opposed to in Stk k \operatorname{Stk}_k Stk k if G G G were in Stk k ⊂ PreStk k \operatorname{Stk}_k\subset\operatorname{PreStk}_k Stk k ⊂ PreStk k ). By [PPS26, Proposition 2.5], Loc K B \operatorname{Loc}_{K_B} Loc K B is in fact fully faithful for any K ∈ Spc K\in\operatorname{Spc} K ∈ Spc . In particular, for the constant prestack G G G associated to any G ∈ Grp ( Spc ) G\in\operatorname{Grp}(\operatorname{Spc}) G ∈ Grp ( Spc ) , we obtain an equivalence,
Fun ( B G , DGCat k ) → ∼ ShvCat ( B G ) , \operatorname{Fun}(BG,\operatorname{DGCat}_k)\xrightarrow{\sim}\operatorname{ShvCat}(BG), Fun ( B G , DGCat k ) ∼ ShvCat ( B G ) ,
between G G G -actions on dg-categories and quasi-coherent sheaves of categories on B G BG B G .
Remark 3.21. For our case of interest, G = S 1 G=S^1 G = S 1 , the stack B G BG B G is not 1 1 1 -affine itself, so we cannot further identify Γ e n h : ShvCat ( B G ) → ∼ QCoh ( B G ) – mod ( DGCat k ) \Gamma^{\mathrm{enh}}:\operatorname{ShvCat}(BG)\xrightarrow{\sim}\operatorname{QCoh}(BG)\text{--}\operatorname{mod}(\operatorname{DGCat}_k) Γ enh : ShvCat ( B G ) ∼ QCoh ( B G ) – mod ( DGCat k ) .
Lemma 3.22. Let X ∈ PreStk l a f t X\in\operatorname{PreStk}_{\mathrm{laft}} X ∈ PreStk laft be equipped with an action of the group prestack S B 1 S^1_B S B 1 . Then IndCoh ( X ) ∈ DGCat k \operatorname{IndCoh}(X)\in\operatorname{DGCat}_k IndCoh ( X ) ∈ DGCat k obtains a canonical S 1 S^1 S 1 -action, which moreover is compatible with its t t t -structure.
Proof. Recall that the data of an S B 1 S^1_B S B 1 -action on X X X 19 is a certain simplicial object
X ∙ : = ⋯ ⇉ S B 1 × S B 1 × X ⇉ S B 1 × X ⇉ X . X_\bullet:=\cdots\rightrightarrows S^1_B\times S^1_B\times X \rightrightarrows S^1_B\times X\rightrightarrows X. X ∙ := ⋯ ⇉ S B 1 × S B 1 × X ⇉ S B 1 × X ⇉ X .
There is a natural map X ∙ → ( S B 1 ) ∙ X_\bullet\to(S^1_B)_\bullet X ∙ → ( S B 1 ) ∙ given by the projections omitting the X X X factor at each level. By definition, we have B ( S B 1 ) ≃ ∣ ( S B 1 ) ∙ ∣ : = colim [ n ] ∈ Δ o p ( S B 1 ) ∙ B(S^1_B)\simeq\lvert(S^1_B)_\bullet\rvert:=\operatorname{colim}_{[n]\in\Delta^{\mathrm{op}}}(S^1_B)_\bullet B ( S B 1 ) ≃ ∣( S B 1 ) ∙ ∣ := colim [ n ] ∈ Δ op ( S B 1 ) ∙ , so applying colimits, we obtain a morphism ∣ X ∙ ∣ → B ( S B 1 ) \lvert X_\bullet\rvert\to B(S^1_B) ∣ X ∙ ∣ → B ( S B 1 ) . On the other hand, because ( − ) B (-)_B ( − ) B preserves limits and colimits, we have B ( S B 1 ) ≃ ( B S 1 ) B B(S^1_B)\simeq(BS^1)_B B ( S B 1 ) ≃ ( B S 1 ) B , where … Using the straightening equivalence Spc / B S 1 ≃ Fun ( B S 1 , Spc ) \operatorname{Spc}/BS^1 \simeq\operatorname{Fun}(BS^1,\operatorname{Spc}) Spc / B S 1 ≃ Fun ( B S 1 , Spc ) , we obtain the following sequence of equivalences,
PreStk / ( B S 1 ) B = Fun ( CAlg k c n , Spc ) / ( B S 1 ) B ≃ Fun ( CAlg k c n , Spc / B S 1 ) ≃ Fun ( CAlg k c n , Fun ( B S 1 , Spc ) ) ≃ Fun ( CAlg k c n × B S 1 , Spc ) ≃ Fun ( B S 1 , Fun ( CAlg k c n , Spc ) ) = Fun ( B S 1 , PreStk ) . \begin{aligned}
\operatorname{PreStk}_{/(BS^1)_B} &= \operatorname{Fun}(\operatorname{CAlg}^{\mathrm{cn}}_k,\operatorname{Spc})/(BS^1)_B \simeq\operatorname{Fun}(\operatorname{CAlg}^{\mathrm{cn}}_k,\operatorname{Spc}/BS^1) \\
&\simeq\operatorname{Fun}(\operatorname{CAlg}^{\mathrm{cn}}_k,\operatorname{Fun}(BS^1,\operatorname{Spc})) \\
&\simeq\operatorname{Fun}(\operatorname{CAlg}^{\mathrm{cn}}_k \times BS^1,\operatorname{Spc}) \\
&\simeq\operatorname{Fun}(BS^1,\operatorname{Fun}(\operatorname{CAlg}^{\mathrm{cn}}_k,\operatorname{Spc})) \\
&= \operatorname{Fun}(BS^1,\operatorname{PreStk}).
\end{aligned} PreStk / ( B S 1 ) B = Fun ( CAlg k cn , Spc ) / ( B S 1 ) B ≃ Fun ( CAlg k cn , Spc / B S 1 ) ≃ Fun ( CAlg k cn , Fun ( B S 1 , Spc )) ≃ Fun ( CAlg k cn × B S 1 , Spc ) ≃ Fun ( B S 1 , Fun ( CAlg k cn , Spc )) = Fun ( B S 1 , PreStk ) .
Thus, a S B 1 S^1_B S B 1 -action on X X X is equivalent to a functor B S 1 → PreStk BS^1 \to\operatorname{PreStk} B S 1 → PreStk sending the unique point of B S 1 BS^1 B S 1 to X X X . Moreover, since X ∈ PreStk l a f t X \in\operatorname{PreStk}_{\mathrm{laft}} X ∈ PreStk laft , this determines a functor act X : B S 1 → PreStk l a f t \operatorname{act}_X : BS^1 \to\operatorname{PreStk}_{\mathrm{laft}} act X : B S 1 → PreStk laft . Composing act X o p : ( B S 1 ) o p → PreStk l a f t o p \operatorname{act}^{\mathrm{op}}_X : (BS^1)^{\mathrm{op}} \to\operatorname{PreStk}^{\mathrm{op}}_{\mathrm{laft}} act X op : ( B S 1 ) op → PreStk laft op with the functor IndCoh : PreStk l a f t o p → DGCat k \operatorname{IndCoh} : \operatorname{PreStk}^{\mathrm{op}}_{\mathrm{laft}} \to\operatorname{DGCat}_k IndCoh : PreStk laft op → DGCat k , to obtain a functor A : B S 1 ≃ ( B S 1 ) o p → DGCat k A : BS^1 \simeq(BS^1)^{\mathrm{op}} \to\operatorname{DGCat}_k A : B S 1 ≃ ( B S 1 ) op → DGCat k under which the unique point of B S 1 BS^1 B S 1 is sent to IndCoh ( X ) \operatorname{IndCoh}(X) IndCoh ( X ) . Thus, the S B 1 S^1_B S B 1 -action on X X X induces an S 1 S^1 S 1 -action on IndCoh ( X ) \operatorname{IndCoh}(X) IndCoh ( X ) .
It remains to show that this S 1 S^1 S 1 -action is compatible with the t t t -structure in the sense that A : B S 1 → DGCat k A : BS^1 \to\operatorname{DGCat}_k A : B S 1 → DGCat k factors through DGCat k t \operatorname{DGCat}^{t}_k DGCat k t .21 Under the adjunction,
B : Alg E 1 ( Cat ∞ ) ⇄ ( Cat ∞ ) ∗ / : Ω ∗ , B : \operatorname{Alg}_{\mathbb{E}_1}(\operatorname{Cat}_{\infty}) \rightleftarrows(\operatorname{Cat}_{\infty})_{*/} : \Omega_*, B : Alg E 1 ( Cat ∞ ) ⇄ ( Cat ∞ ) ∗ / : Ω ∗ ,
A A A corresponds to a functor of monoids (i.e. E 1 \mathbb{E}_1 E 1 -algebras), A ~ : S 1 → Ω ∗ ( DGCat k ) ≃ Aut DGCat k ( IndCoh ( X ) ) \widetilde{A}: S^1 \to\Omega_*(\operatorname{DGCat}_k) \simeq\operatorname{Aut}_{\operatorname{DGCat}_k}(\operatorname{IndCoh}(X)) A : S 1 → Ω ∗ ( DGCat k ) ≃ Aut DGCat k ( IndCoh ( X )) , where DGCat k \operatorname{DGCat}_k DGCat k is regarded as an object of ( Cat ∞ ) ∗ / (\operatorname{Cat}_{\infty})_{*/} ( Cat ∞ ) ∗ / via the choice of category IndCoh ( X ) ∈ DGCat k \operatorname{IndCoh}(X) \in\operatorname{DGCat}_k IndCoh ( X ) ∈ DGCat k .
Since S 1 ≃ B Z S^1 \simeq B\mathbb{Z} S 1 ≃ B Z , it suffices to show that the image under A ~ \widetilde{A} A of the unique point of S 1 S^1 S 1 is a t t t -exact autoequivalence of IndCoh ( X ) \operatorname{IndCoh}(X) IndCoh ( X ) . But since A ~ \widetilde{A} A is a monoidal map, ∗ ∈ S 1 * \in S^1 ∗ ∈ S 1 must be sent to the unit id IndCoh ( X ) ∈ Aut DGCat k ( IndCoh ( X ) ) \operatorname{id}_{\operatorname{IndCoh}(X)} \in\operatorname{Aut}_{\operatorname{DGCat}_k}(\operatorname{IndCoh}(X)) id IndCoh ( X ) ∈ Aut DGCat k ( IndCoh ( X )) , which is obviously t t t -exact. □
Algebraic group actions
While the previous discussion centered around actions of group objects in Spc \operatorname{Spc} Spc , actions of algebraic groups—or, more generally, actions of group stacks—play a significant role in this paper. For our purposes, it is convenient to reformulate the definition of G G G -action given in Definition 3.16 by imitating Preygel’s definition for formal group actions found in [2 ] .
In order to formulate the definition, we recall the following. By [HA, Theorem 4.5.3.1], for any symmetric monoidal category C ⊗ \mathcal{C}^{\otimes} C ⊗ , the functor,
p : Mod ( C ) → CAlg ( C ) p : \operatorname{Mod}(\mathcal{C}) \to\operatorname{CAlg}(\mathcal{C}) p : Mod ( C ) → CAlg ( C )
is a coCartesian fibration such that the fiber over R ∈ CAlg ( C ) R \in\operatorname{CAlg}(\mathcal{C}) R ∈ CAlg ( C ) is the category Mod R ( C ⊗ ) \operatorname{Mod}_R(\mathcal{C}^{\otimes}) Mod R ( C ⊗ ) , and the coCartesian lifts of R → S R \to S R → S are given by the extension of scalars functor − ⊗ R S : Mod R ( C ) → Mod S ( C ) - \otimes_R S : \operatorname{Mod}_R(\mathcal{C}) \to\operatorname{Mod}_S(\mathcal{C}) − ⊗ R S : Mod R ( C ) → Mod S ( C ) .
Construction 3.23. For each object C ∈ C C \in\mathcal{C} C ∈ C , we construct a coCartesian section of p p p , denoted by C ‾ \underline{C} C , as follows. Since Mod 1 ( C ) ≃ C \operatorname{Mod}_1(\mathcal{C}) \simeq\mathcal{C} Mod 1 ( C ) ≃ C by [HA, Proposition 3.4.2.1], C ∈ Mod 1 ( C ) C \in\operatorname{Mod}_1(\mathcal{C}) C ∈ Mod 1 ( C ) , meaning we obtain an object in the fiber of p p p over the initial object 1 ∈ CAlg ( C ) 1 \in\operatorname{CAlg}(\mathcal{C}) 1 ∈ CAlg ( C ) . For each R ∈ CAlg ( C ) R \in\operatorname{CAlg}(\mathcal{C}) R ∈ CAlg ( C ) , the unique map u R : 1 → R u_R : 1 \to R u R : 1 → R lifts to the p p p -coCartesian edge u ~ R : C → C ⊗ 1 R \widetilde{u}_R : C \to C \otimes1 R u R : C → C ⊗ 1 R . By dint of p p p being a coCartesian fibration and the universal property of p p p -coCartesian edges, these lifts assemble coherently22 into a coCartesian section,
C ‾ : CAlg ( C ) → Mod ( C ) . \underline{C} : \operatorname{CAlg}(\mathcal{C}) \to\operatorname{Mod}(\mathcal{C}). C : CAlg ( C ) → Mod ( C ) .
We consider two special cases of this construction: (1) Let C ⊗ = Vect ⊗ \mathcal{C}^{\otimes} = \operatorname{Vect}^{\otimes} C ⊗ = Vect ⊗ . Then a given object K ∈ Vect k K \in\operatorname{Vect}_{k} K ∈ Vect k determines a functor K ‾ : CAlg k → Mod ( Vect k ) \underline{K} : \operatorname{CAlg}_{k} \to\operatorname{Mod}(\operatorname{Vect}_{k}) K : CAlg k → Mod ( Vect k ) . Restricting K ‾ \underline{K} K along the inclusion CAlg k c n ↪ CAlg k \operatorname{CAlg}_{k}^{\mathrm{cn}} \hookrightarrow\operatorname{CAlg}_{k} CAlg k cn ↪ CAlg k , and composing with the forgetful functor Mod ( Vect ) → Spc \operatorname{Mod}(\operatorname{Vect}) \to\operatorname{Spc} Mod ( Vect ) → Spc , we obtain a prestack,
K ‾ : CAlg k c n → Spc , (6) \underline{K} : \operatorname{CAlg}_{k}^{\mathrm{cn}} \to\operatorname{Spc}, \tag*{(6)} K : CAlg k cn → Spc , (6)
which, by abuse of notation, we also denote by K ‾ \underline{K} K .
(2) Let C ⊗ = DGCat k ⊗ \mathcal{C}^{\otimes} = \operatorname{DGCat}_{k}^{\otimes} C ⊗ = DGCat k ⊗ . A given dg-category E ∈ DGCat k \mathcal{E} \in\operatorname{DGCat}_{k} E ∈ DGCat k determines a functor
E ‾ : CAlg k ( DGCat k ) → Mod ( DGCat k ) . \underline{\mathcal{E}} : \operatorname{CAlg}_{k}(\operatorname{DGCat}_{k}) \to\operatorname{Mod}(\operatorname{DGCat}_{k}). E : CAlg k ( DGCat k ) → Mod ( DGCat k ) .
There is a fully faithful embedding CAlg k c n → CAlg ( DGCat k ) \operatorname{CAlg}_{k}^{\mathrm{cn}} \to\operatorname{CAlg}(\operatorname{DGCat}_{k}) CAlg k cn → CAlg ( DGCat k ) which assigns to the object R ∈ CAlg k R \in\operatorname{CAlg}_{k} R ∈ CAlg k the symmetric monoidal category Mod R ⊗ ∈ CAlg ( DGCat k ) \operatorname{Mod}_{R}^{\otimes} \in\operatorname{CAlg}(\operatorname{DGCat}_{k}) Mod R ⊗ ∈ CAlg ( DGCat k ) . Restricting along this embedding and composing with the functor oblv ( − ) ≃ : Mod ( DGCat k ) → Spc \operatorname{oblv}(-)^{\simeq} : \operatorname{Mod}(\operatorname{DGCat}_{k}) \to\operatorname{Spc} oblv ( − ) ≃ : Mod ( DGCat k ) → Spc obtains a prestack
E ‾ : CAlg k c n → Spc , (7) \underline{\mathcal{E}} : \operatorname{CAlg}_{k}^{\mathrm{cn}} \to\operatorname{Spc}, \tag*{(7)} E : CAlg k cn → Spc , (7)
which, by abuse of notation, we also denote by E ‾ \underline{\mathcal{E}} E .
We also have the following related construction:
Construction 3.26. Let C ⊗ \mathcal{C}^{\otimes} C ⊗ be a symmetric monoidal category, and let C ∈ C C \in\mathcal{C} C ∈ C . Denote by St ( p ) : CAlg ( C ) → Cat ∞ \operatorname{St}(p) : \operatorname{CAlg}(\mathcal{C}) \to\operatorname{Cat}_{\infty} St ( p ) : CAlg ( C ) → Cat ∞ the straightening of the coCartesian fibration p p p recalled above. The coCartesian section of p p p , C \mathcal{C} C , constructed in Construction 3.23 corresponds under the straightening operation to a natural transformation St ( C ) : ∗ → St ( p ) \operatorname{St}(\mathcal{C}) : * \to\operatorname{St}(p) St ( C ) : ∗ → St ( p ) , where ∗ : CAlg ( C ) → Cat ∞ * : \operatorname{CAlg}(\mathcal{C}) \to\operatorname{Cat}_{\infty} ∗ : CAlg ( C ) → Cat ∞ is the constant functor with value the point ∗ ∈ Cat ∞ * \in\operatorname{Cat}_{\infty} ∗ ∈ Cat ∞ . We then define
Aut ( C ) : = ∗ × St ( p ) ∗ , \operatorname{Aut}(\mathcal{C}) := * \times_{\operatorname{St}(p)} * , Aut ( C ) := ∗ × St ( p ) ∗ ,
where the product on the right-hand side is taken in Fun ( CAlg ( C ) , Cat ∞ ) \operatorname{Fun}(\operatorname{CAlg}(\mathcal{C}), \operatorname{Cat}_{\infty}) Fun ( CAlg ( C ) , Cat ∞ ) . The resulting functor sends the object A ∈ CAlg ( C ) A \in\operatorname{CAlg}(\mathcal{C}) A ∈ CAlg ( C ) to the category Aut Mod A ( C ) ( C ( A ) ) \operatorname{Aut}_{\operatorname{Mod}_{A}(\mathcal{C})}(\mathcal{C}(A)) Aut Mod A ( C ) ( C ( A )) .
As a special case, we take C = DGCat k \mathcal{C} = \operatorname{DGCat}_{k} C = DGCat k , and C = E ∈ DGCat k C = \mathcal{E} \in\operatorname{DGCat}_{k} C = E ∈ DGCat k . Composing with the functor ( − ) ≃ : Cat ∞ → Spc (-)^{\simeq} : \operatorname{Cat}_{\infty} \to\operatorname{Spc} ( − ) ≃ : Cat ∞ → Spc and restricting along the inclusion CAlg k c n → CAlg ( DGCat k ) \operatorname{CAlg}_{k}^{\mathrm{cn}} \to\operatorname{CAlg}(\operatorname{DGCat}_{k}) CAlg k cn → CAlg ( DGCat k ) , we obtain a functor res ( − ) ≃ : Fun ( CAlg ( DGCat k ) , Cat ∞ ) → PreStk k \operatorname{res}(-)^{\simeq} : \operatorname{Fun}(\operatorname{CAlg}(\operatorname{DGCat}_{k}), \operatorname{Cat}_{\infty}) \to\operatorname{PreStk}_{k} res ( − ) ≃ : Fun ( CAlg ( DGCat k ) , Cat ∞ ) → PreStk k , which preserves limits. Thus, we have
res ( Aut ( E ) ) ≃ ≃ ∗ B × res ( St ( p ) ) ≃ ∗ B , \operatorname{res}(\operatorname{Aut}(\mathcal{E}))^{\simeq} \simeq*_B \times_{\operatorname{res}(\operatorname{St}(p))^{\simeq}} *_B, res ( Aut ( E ) ) ≃ ≃ ∗ B × res ( St ( p ) ) ≃ ∗ B ,
where ∗ B *_B ∗ B is the Betti prestack with value ∗ ∈ Spc * \in\operatorname{Spc} ∗ ∈ Spc . To simplify notation, we make the following conventions:
(1) DGCat ‾ k \overline{\operatorname{DGCat}}_{k} DGCat k denotes the prestack res ( St ( p ) ) ≃ \operatorname{res}(\operatorname{St}(p))^{\simeq} res ( St ( p ) ) ≃ , and
(2) E ~ \widetilde{\mathcal{E}} E denotes the natural transformation res ( St ( E ) ) ≃ \operatorname{res}(\operatorname{St}(\mathcal{E}))^{\simeq} res ( St ( E ) ) ≃ .
We may consider DGCat ‾ k \overline{\operatorname{DGCat}}_{k} DGCat k as a pointed object of PreStk k \operatorname{PreStk}_{k} PreStk k via the map E ~ \widetilde{\mathcal{E}} E . In which case, ∗ B × res ( St ( p ) ) ≃ ∗ B *_B \times_{\operatorname{res}(\operatorname{St}(p))^{\simeq}} *_B ∗ B × res ( St ( p ) ) ≃ ∗ B may be rewritten as Ω ∗ ( DGCat ‾ k ) \Omega_{*}(\overline{\operatorname{DGCat}}_{k}) Ω ∗ ( DGCat k ) , where Ω ∗ : ( PreStk k ) ∗ B / → Grp ( PreStk ) \Omega_{*} : (\operatorname{PreStk}_{k})_{*_B/} \to\operatorname{Grp}(\operatorname{PreStk}) Ω ∗ : ( PreStk k ) ∗ B / → Grp ( PreStk ) is the based loops functor on pointed prestacks. The value on R ∈ CAlg k c n R \in\operatorname{CAlg}_{k}^{\mathrm{cn}} R ∈ CAlg k cn of the prestack Ω E ~ ( DGCat ‾ k ) \Omega_{\widetilde{\mathcal{E}}}(\overline{\operatorname{DGCat}}_{k}) Ω E ( DGCat k ) is the category Aut DGCat R ( E ⊗ k R ) ≃ \operatorname{Aut}_{\operatorname{DGCat}_{R}}(\mathcal{E} \otimes_{k} R)^{\simeq} Aut DGCat R ( E ⊗ k R ) ≃ . This motivates the following definition of the automorphism prestack of E \mathcal{E} E :
Aut ( E ) : = Ω ∗ ( DGCat ‾ k ) = ∗ B × res ( St ( p ) ) ≃ ∗ B . (8) \operatorname{Aut}(\mathcal{E}) := \Omega_{*}(\overline{\operatorname{DGCat}}_{k}) = *_B \times_{\operatorname{res}(\operatorname{St}(p))^{\simeq}} *_B. \tag*{(8)} Aut ( E ) := Ω ∗ ( DGCat k ) = ∗ B × res ( St ( p ) ) ≃ ∗ B . (8)
We emphasize that Aut ( E ) \operatorname{Aut}(\mathcal{E}) Aut ( E ) is an object of Grp ( PreStk k ) \operatorname{Grp}(\operatorname{PreStk}_{k}) Grp ( PreStk k ) .
The above construction allows us to give an alternative definition of group prestack action on a category.
Definition 3.28. Let E ∈ DGCat k \mathcal{E} \in\operatorname{DGCat}_{k} E ∈ DGCat k , and let G ∈ Grp ( PreStk k ) \mathcal{G} \in\operatorname{Grp}(\operatorname{PreStk}_{k}) G ∈ Grp ( PreStk k ) . A G \mathcal{G} G -action on E \mathcal{E} E is a map of group prestacks G → Aut ( E ) \mathcal{G} \to\operatorname{Aut}(\mathcal{E}) G → Aut ( E ) , where the latter object is the group prestack (8) considered above.
Definition 3.16 and Definition 3.28 do not agree in general, but it turns out that for G ∈ Grp ( PreStk k ) \mathcal{G} \in\operatorname{Grp}(\operatorname{PreStk}_{k}) G ∈ Grp ( PreStk k ) such that Loc G \operatorname{Loc}_{\mathcal{G}} Loc G is a fully faithful—which includes the cases G = S B 1 \mathcal{G}=S^{1}_{B} G = S B 1 and G = B G a \mathcal{G}=B\mathbb{G}_{a} G = B G a —they do.
Lemma 3.29. Let G ∈ Grp ( PreStk k ) \mathcal{G} \in\operatorname{Grp}(\operatorname{PreStk}_{k}) G ∈ Grp ( PreStk k ) such that the functor Loc G : QCoh ( G ) - mod → ShvCat ( G ) \operatorname{Loc}_{\mathcal{G}}:\operatorname{QCoh}(\mathcal{G})\text{-}\operatorname{mod}\to\operatorname{ShvCat}(\mathcal{G}) Loc G : QCoh ( G ) - mod → ShvCat ( G ) is fully faithful, and let C ∈ DGCat k \mathcal{C}\in\operatorname{DGCat}_{k} C ∈ DGCat k . Then the data of a QCoh ( G ) \operatorname{QCoh}(\mathcal{G}) QCoh ( G ) -comodule structure on C \mathcal{C} C is equivalent to the data of a map G → Aut ( C ) ∈ Grp ( PreStk k ) \mathcal{G}\to\operatorname{Aut}(\mathcal{C})\in\operatorname{Grp}(\operatorname{PreStk}_{k}) G → Aut ( C ) ∈ Grp ( PreStk k ) . That is, Definition 3.16 and Definition 3.28 are equivalent definitions of G \mathcal{G} G -action on the category C \mathcal{C} C .
Proof. We begin by producing an equivalence,
Map Grp ( PreStk ) ( G , Aut ( C ) ) → ShvCat ( B G ) ≃ × DGCat k ≃ { C } , \operatorname{Map}_{\operatorname{Grp}(\operatorname{PreStk})}(\mathcal{G},\operatorname{Aut}(\mathcal{C}))\to\operatorname{ShvCat}(B\mathcal{G})^{\simeq}\mathop{\times}_{\operatorname{DGCat}_{k}^{\simeq}}\{\mathcal{C}\}, Map Grp ( PreStk ) ( G , Aut ( C )) → ShvCat ( B G ) ≃ × DGCat k ≃ { C } ,
where the product on the right hand side is taken along the functor ShvCat ( B G ) → ShvCat ( Spec k ) ≃ DGCat k \operatorname{ShvCat}(B\mathcal{G})\to\operatorname{ShvCat}(\operatorname{Spec} k)\simeq\operatorname{DGCat}_{k} ShvCat ( B G ) → ShvCat ( Spec k ) ≃ DGCat k given by pullback along the unique point Spec k → B G \operatorname{Spec} k\to B\mathcal{G} Spec k → B G .
Since the category PreStk \operatorname{PreStk} PreStk is an ∞ \infty ∞ -topos, we have ([HA, Notation 5.3.6.11]) an adjunction,
B : Grp ( PreStk k ) ⇄ ( PreStk k ) Spec k / : Ω ∗ , B:\operatorname{Grp}(\operatorname{PreStk}_{k})\rightleftarrows(\operatorname{PreStk}_{k})_{\operatorname{Spec} k/}:\Omega_{*}, B : Grp ( PreStk k ) ⇄ ( PreStk k ) Spec k / : Ω ∗ ,
where we note that Spec k \operatorname{Spec} k Spec k is the terminal object of PreStk k \operatorname{PreStk}_{k} PreStk k . As Aut ( C ) \operatorname{Aut}(\mathcal{C}) Aut ( C ) is, by definition, the based loop space Ω ∗ ( DGCat ~ k ) \Omega_{*}(\widetilde{\operatorname{DGCat}}_{k}) Ω ∗ ( DGCat k ) , we have the following sequence of equivalences.
Map Grp ( PreStk k ) ( G , Aut ( C ) ) ≃ Map ( PreStk k ) Spec k / ( B G , DGCat ~ k ) ≃ Map PreStk k ( B G , DGCat ~ k ) × Map PreStk k ( Spec k , DGCat ~ k ) { C ~ } ≃ ( lim Spec R → B G , R ∈ CAlg k c n Map PreStk k ( Spec R , DGCat ~ k ) ) × DGCat k ≃ { C } ≃ ( lim Spec R → B G , R ∈ CAlg k c n DGCat ~ k ( R ) ) × DGCat k ≃ C ≃ ( lim Spec R → B G , R ∈ CAlg k c n ( DGCat R ) ≃ ) × DGCat k ≃ C ≃ ( lim Spec R → B G , R ∈ CAlg k c n ( DGCat R ) ) ≃ × DGCat k ≃ C ≃ ShvCat ( B G ) ≃ × ( DGCat k ) ≃ C , \begin{aligned}
\operatorname{Map}_{\operatorname{Grp}(\operatorname{PreStk}_{k})}(\mathcal{G},\operatorname{Aut}(\mathcal{C})) &\simeq\operatorname{Map}_{(\operatorname{PreStk}_{k})_{\operatorname{Spec} k/}}(B\mathcal{G},\widetilde{\operatorname{DGCat}}_{k}) \\
&\simeq\operatorname{Map}_{\operatorname{PreStk}_{k}}(B\mathcal{G},\widetilde{\operatorname{DGCat}}_{k})\mathop{\times}_{\operatorname{Map}_{\operatorname{PreStk}_{k}}(\operatorname{Spec} k,\widetilde{\operatorname{DGCat}}_{k})}\{\widetilde{\mathcal{C}}\} \\
&\simeq\left(\lim_{\substack{\operatorname{Spec} R\to B\mathcal{G},\ R\in\operatorname{CAlg}_{k}^{\mathrm{cn}}}}\operatorname{Map}_{\operatorname{PreStk}_{k}}(\operatorname{Spec} R,\widetilde{\operatorname{DGCat}}_{k})\right)\mathop{\times}_{\operatorname{DGCat}_{k}^{\simeq}}\{\mathcal{C}\} \\
&\simeq\left(\lim_{\substack{\operatorname{Spec} R\to B\mathcal{G},\ R\in\operatorname{CAlg}_{k}^{\mathrm{cn}}}}\widetilde{\operatorname{DGCat}}_{k}(R)\right)\mathop{\times}_{\operatorname{DGCat}_{k}^{\simeq}}\mathcal{C} \\
&\simeq\left(\lim_{\substack{\operatorname{Spec} R\to B\mathcal{G},\ R\in\operatorname{CAlg}_{k}^{\mathrm{cn}}}}(\operatorname{DGCat}_{R})^{\simeq}\right)\mathop{\times}_{\operatorname{DGCat}_{k}^{\simeq}}\mathcal{C} \\
&\simeq\left(\lim_{\substack{\operatorname{Spec} R\to B\mathcal{G},\ R\in\operatorname{CAlg}_{k}^{\mathrm{cn}}}}(\operatorname{DGCat}_{R})\right)^{\simeq}\mathop{\times}_{\operatorname{DGCat}_{k}^{\simeq}}\mathcal{C} \\
&\simeq\operatorname{ShvCat}(B\mathcal{G})^{\simeq}\mathop{\times}_{(\operatorname{DGCat}_{k})^{\simeq}}\mathcal{C},
\end{aligned} Map Grp ( PreStk k ) ( G , Aut ( C )) ≃ Map ( PreStk k ) Spec k / ( B G , DGCat k ) ≃ Map PreStk k ( B G , DGCat k ) × Map PreStk k ( Spec k , DGCat k ) { C } ≃ ( Spec R → B G , R ∈ CAlg k cn lim Map PreStk k ( Spec R , DGCat k ) ) × DGCat k ≃ { C } ≃ ( Spec R → B G , R ∈ CAlg k cn lim DGCat k ( R ) ) × DGCat k ≃ C ≃ ( Spec R → B G , R ∈ CAlg k cn lim ( DGCat R ) ≃ ) × DGCat k ≃ C ≃ ( Spec R → B G , R ∈ CAlg k cn lim ( DGCat R ) ) ≃ × DGCat k ≃ C ≃ ShvCat ( B G ) ≃ × ( DGCat k ) ≃ C ,
where the final equivalence follows from the definition of ShvCat ( − ) \operatorname{ShvCat}(-) ShvCat ( − ) found in ([Gai15]) and the product in the final line is the limit of the span ShvCat ( B G ) → pt ∗ ShvCat ( Spec k ) ≃ ( DGCat k ) ≃ ← { C } \operatorname{ShvCat}(B\mathcal{G})\xrightarrow{\operatorname{pt}^{*}}\operatorname{ShvCat}(\operatorname{Spec} k)\simeq(\operatorname{DGCat}_{k})^{\simeq}\leftarrow\{\mathcal{C}\} ShvCat ( B G ) pt ∗ ShvCat ( Spec k ) ≃ ( DGCat k ) ≃ ← { C } . Now recall from the discussion above that, since the functor Loc G \operatorname{Loc}_{\mathcal{G}} Loc G is fully faithful, we have an equivalence QCoh ( G ) - comod ( DGCat k ) ≃ ShvCat ( B G ) \operatorname{QCoh}(\mathcal{G})\text{-}\operatorname{comod}(\operatorname{DGCat}_{k})\simeq\operatorname{ShvCat}(B\mathcal{G}) QCoh ( G ) - comod ( DGCat k ) ≃ ShvCat ( B G ) . Combining this equivalence with the ones in the chain above, we obtain an equivalence,
Map Grp ( PreStk k ) ( G , Aut ( C ) ) ≃ QCoh ( G ) - comod ( DGCat k ) ≃ ShvCat ( B G ) ≃ × ( DGCat k ) ≃ { C } , \operatorname{Map}_{\operatorname{Grp}(\operatorname{PreStk}_{k})}(\mathcal{G},\operatorname{Aut}(\mathcal{C}))\simeq\operatorname{QCoh}(\mathcal{G})\text{-}\operatorname{comod}(\operatorname{DGCat}_{k})\simeq\operatorname{ShvCat}(B\mathcal{G})^{\simeq}\mathop{\times}_{(\operatorname{DGCat}_{k})^{\simeq}}\{\mathcal{C}\}, Map Grp ( PreStk k ) ( G , Aut ( C )) ≃ QCoh ( G ) - comod ( DGCat k ) ≃ ShvCat ( B G ) ≃ × ( DGCat k ) ≃ { C } ,
as desired. □
Remark 3.30. As noted above, Loc G \operatorname{Loc}_{\mathcal{G}} Loc G is fully faithful when G = G B \mathcal{G}=G_{B} G = G B for any G ∈ Grp ( Spc ) G\in\operatorname{Grp}(\operatorname{Spc}) G ∈ Grp ( Spc ) . In particular, Lemma 3.29 applies to the group prestack S B 1 S^{1}_{B} S B 1 , which utilize in the proof of the proposition below.
Proposition 3.31. Let C ∈ DGCat k \mathcal{C}\in\operatorname{DGCat}_{k} C ∈ DGCat k be a d g dg d g -category equipped with a B G a B\mathbb{G}_{a} B G a -action. Consider C \mathcal{C} C as an S 1 S^{1} S 1 -category via the affinization map S B 1 → B G a S^{1}_{B}\to B\mathbb{G}_{a} S B 1 → B G a . Let c ∈ C c\in\mathcal{C} c ∈ C , and let ρ : c → ≃ c \rho:c\xrightarrow{\simeq}c ρ : c ≃ c denote the automorphism induced by this S 1 S^{1} S 1 -action. Then ρ \rho ρ is a unipotent operator23 if c c c is compact.
Proof. The choice of an object c ∈ C c \in\mathcal{C} c ∈ C defines a map c : Vect k → C c : \operatorname{Vect}_{k} \to\mathcal{C} c : Vect k → C . For each R ∈ CAlg k c n R \in\operatorname{CAlg}_{k}^{\mathrm{cn}} R ∈ CAlg k cn , there is a natural k k k -linear functor − ⊗ k R : C → C ⊗ k R -\otimes_k R : \mathcal{C} \to\mathcal{C} \otimes_k R − ⊗ k R : C → C ⊗ k R given by extension of scalars. Precomposing this functor with c c c , we get a k k k -linear functor which under adjunction corresponds to an R R R -linear functor Mod R ≃ Vect k ⊗ k R → C ⊗ k R \operatorname{Mod}_{R} \simeq\operatorname{Vect}_{k} \otimes_k R \to\mathcal{C} \otimes_k R Mod R ≃ Vect k ⊗ k R → C ⊗ k R , under which the generator R ∈ Mod R R \in\operatorname{Mod}_{R} R ∈ Mod R is sent to the object c ⊗ k R c \otimes_k R c ⊗ k R . Because of the naturality of this construction, these functors assemble into a natural transformation c : Vect k → C c : \operatorname{Vect}_{k} \to\mathcal{C} c : Vect k → C , which we view as a morphism in the ∞ \infty ∞ -category PreStk k = Fun ( CAlg k c n , Spc ) \operatorname{PreStk}_{k} = \operatorname{Fun}(\operatorname{CAlg}_{k}^{\mathrm{cn}}, \operatorname{Spc}) PreStk k = Fun ( CAlg k cn , Spc ) .
On the other hand, a B G a B\mathbb{G}_a B G a -action on C \mathcal{C} C is the data of a map B G a → Aut ( C ) B\mathbb{G}_a \to\operatorname{Aut}(\mathcal{C}) B G a → Aut ( C ) in Grp ( PreStk k ) \operatorname{Grp}(\operatorname{PreStk}_{k}) Grp ( PreStk k ) (i.e. a natural transformation). Note that, by [HA, Remark 2.1.3.4], we have an equivalence
Fun ( CAlg k c n , Grp ( Spc ) ) ≃ Grp ( Fun ( CAlg k c n , Spc ) ) . \operatorname{Fun}(\operatorname{CAlg}_{k}^{\mathrm{cn}}, \operatorname{Grp}(\operatorname{Spc})) \simeq\operatorname{Grp}(\operatorname{Fun}(\operatorname{CAlg}_{k}^{\mathrm{cn}}, \operatorname{Spc})). Fun ( CAlg k cn , Grp ( Spc )) ≃ Grp ( Fun ( CAlg k cn , Spc )) .
As noted above, the category PreStk \operatorname{PreStk} PreStk is an ∞ \infty ∞ -topos, so by [HA, Notation 5.3.6.11], we have an adjunction,
B : Grp ( PreStk k ) ⇄ ( PreStk k ) Spec k / : Ω ∗ . \mathcal{B} : \operatorname{Grp}(\operatorname{PreStk}_{k}) \rightleftarrows(\operatorname{PreStk}_{k})_{\operatorname{Spec} k/} : \Omega_*. B : Grp ( PreStk k ) ⇄ ( PreStk k ) Spec k / : Ω ∗ .
Note that, by [Toë06, Lemme 2.2.5], we have that B G a ≃ B G a B\mathbb{G}_a \simeq B\mathbb{G}_a B G a ≃ B G a , where the latter stack corresponds to the underlying prestack of the delooping of G a \mathbb{G}_a G a as an object of Grp ( Stk k ) \operatorname{Grp}(\operatorname{Stk}_{k}) Grp ( Stk k ) , rather than of Grp ( PreStk k ) \operatorname{Grp}(\operatorname{PreStk}_{k}) Grp ( PreStk k ) . Thus, under the above adjunction, the action E 1 E_1 E 1 -algebra map B G a → Aut ( C ) B\mathbb{G}_a \to\operatorname{Aut}(\mathcal{C}) B G a → Aut ( C ) corresponds to the data of an E 2 E_2 E 2 -algebra map G a → Ω ∗ Aut ( C ) ≃ Aut Aut ( C ) ( id C ) \mathbb{G}_a \to\Omega_*\operatorname{Aut}(\mathcal{C}) \simeq\operatorname{Aut}_{\operatorname{Aut}(\mathcal{C})}(\operatorname{id}_{\mathcal{C}}) G a → Ω ∗ Aut ( C ) ≃ Aut Aut ( C ) ( id C ) , which, in particular, gives a map ρ : G a → Aut ( c ) \rho: \mathbb{G}_a \to\operatorname{Aut}(c) ρ : G a → Aut ( c ) , for each c ∈ C c \in\mathcal{C} c ∈ C . This map induces a map of R R R -points,
R ≃ G a ( R ) → ρ R Aut C ⊗ k R ( c ⊗ k R ) . R \simeq\mathbb{G}_a(R) \xrightarrow{\rho_R} \operatorname{Aut}_{\mathcal{C}\otimes_k R}(c\otimes_k R). R ≃ G a ( R ) ρ R Aut C ⊗ k R ( c ⊗ k R ) .
If c c c is compact, the canonical map
End C ( c ) ⊗ k R ⟶ End C ⊗ k R ( c ⊗ k R ) (9) \operatorname{End}_{\mathcal{C}}(c)\otimes_k R \longrightarrow\operatorname{End}_{\mathcal{C}\otimes_k R}(c\otimes_k R) \tag*{(9)} End C ( c ) ⊗ k R ⟶ End C ⊗ k R ( c ⊗ k R ) (9)
is an equivalence, so we may regard Aut C ⊗ k R ( c ⊗ k R ) \operatorname{Aut}_{\mathcal{C}\otimes_k R}(c\otimes_k R) Aut C ⊗ k R ( c ⊗ k R ) as an E 1 E_1 E 1 -subspace of End C ( c ) ⊗ k R \operatorname{End}_{\mathcal{C}}(c)\otimes_k R End C ( c ) ⊗ k R . Now take R = k [ x ] R=k[x] R = k [ x ] , and consider the element x ∈ k [ x ] x \in k[x] x ∈ k [ x ] , which we regard as a map x : ∗ → k [ x ] x : * \to k[x] x : ∗ → k [ x ] . The composition ρ k [ x ] ( x ) \rho_{k[x]}(x) ρ k [ x ] ( x ) gives a point of End C ( c ) ⊗ k k [ x ] \operatorname{End}_{\mathcal{C}}(c)\otimes_k k[x] End C ( c ) ⊗ k k [ x ] . On the other hand, we have
Map k ( k , End C ( c ) ⊗ k k [ x 1 , … , x n ] ) ≃ Map k ( k , End C ( c ) ⊗ k ( ⨁ i 1 , … , i n ≥ 0 k ⋅ x 1 i 1 ⋯ x n i n ) ) ≃ Map k ( k , ⨁ i 1 , … , i n ≥ 0 ( End C ( c ) ⋅ x 1 i 1 ⋯ x n i n ) ) ≃ ⨁ i 1 , … , i n ≥ 0 Map k ( k , End C ( c ) ⋅ x 1 i 1 ⋯ x n i n ) ≃ ⨁ i 1 , … , i n ≥ 0 Ω ∞ End C ( c ) ⋅ x 1 i 1 ⋯ x n i n \begin{aligned}
\operatorname{Map}_k\left(k,\operatorname{End}_{\mathcal{C}}(c)\otimes_k k[x_1,\ldots,x_n]\right)
&\simeq\operatorname{Map}_k\left(k,\operatorname{End}_{\mathcal{C}}(c)\otimes_k\left(\bigoplus_{i_1,\ldots,i_n\geq0} k\cdot x_1^{i_1}\cdots x_n^{i_n}\right)\right) \\
&\simeq\operatorname{Map}_k\left(k,\bigoplus_{i_1,\ldots,i_n\geq0}\left(\operatorname{End}_{\mathcal{C}}(c)\cdot x_1^{i_1}\cdots x_n^{i_n}\right)\right) \\
&\simeq\bigoplus_{i_1,\ldots,i_n\geq0}\operatorname{Map}_k\left(k,\operatorname{End}_{\mathcal{C}}(c)\cdot x_1^{i_1}\cdots x_n^{i_n}\right) \\
&\simeq\bigoplus_{i_1,\ldots,i_n\geq0}\Omega^\infty\operatorname{End}_{\mathcal{C}}(c)\cdot x_1^{i_1}\cdots x_n^{i_n}
\end{aligned} Map k ( k , End C ( c ) ⊗ k k [ x 1 , … , x n ] ) ≃ Map k ( k , End C ( c ) ⊗ k ( i 1 , … , i n ≥ 0 ⨁ k ⋅ x 1 i 1 ⋯ x n i n ) ) ≃ Map k ( k , i 1 , … , i n ≥ 0 ⨁ ( End C ( c ) ⋅ x 1 i 1 ⋯ x n i n ) ) ≃ i 1 , … , i n ≥ 0 ⨁ Map k ( k , End C ( c ) ⋅ x 1 i 1 ⋯ x n i n ) ≃ i 1 , … , i n ≥ 0 ⨁ Ω ∞ End C ( c ) ⋅ x 1 i 1 ⋯ x n i n
where the notation “x 1 i 1 ⋯ x n i n x_1^{i_1}\cdots x_n^{i_n} x 1 i 1 ⋯ x n i n ” is a suggestive way of writing the index, and the arrow in the third line is an equivalence because k ∈ Vect k k \in\operatorname{Vect}_{k} k ∈ Vect k is compact. Thus, a point of End C ( c ) [ x 1 , … , x n ] : = End C ( c ) ⊗ k k [ x 1 , … , x n ] \operatorname{End}_{\mathcal{C}}(c)[x_1,\ldots,x_n] := \operatorname{End}_{\mathcal{C}}(c)\otimes_k k[x_1,\ldots,x_n] End C ( c ) [ x 1 , … , x n ] := End C ( c ) ⊗ k k [ x 1 , … , x n ] is precisely a sum, or polynomial,
∑ ( i 1 , … , i n ) ∈ N n a i 1 … i n x 1 i 1 ⋯ x n i n , \sum_{(i_1,\ldots,i_n)\in\mathbb{N}^n} a_{i_1\ldots i_n}x_1^{i_1}\cdots x_n^{i_n}, ( i 1 , … , i n ) ∈ N n ∑ a i 1 … i n x 1 i 1 ⋯ x n i n ,
where a i 1 … i n ∈ Ω ∞ ( End C ( c ) ) a_{i_1\ldots i_n} \in\Omega^\infty(\operatorname{End}_{\mathcal{C}}(c)) a i 1 … i n ∈ Ω ∞ ( End C ( c )) . In the following, let u f u_f u f denote the polynomial given by ρ k [ x 1 , … , x n ] ( f ) \rho_{k[x_1,\ldots,x_n]}(f) ρ k [ x 1 , … , x n ] ( f ) , for f ∈ k [ x 1 , … , x n ] f \in k[x_1,\ldots,x_n] f ∈ k [ x 1 , … , x n ] .
Because φ : G a → Aut ( c ) \varphi: \mathbb{G}_a \to\operatorname{Aut}(c) φ : G a → Aut ( c ) is a natural transformation, we have a commutative diagram of E 1 E_1 E 1 -spaces corresponding to any given k k k -algebra map φ : k [ x ] → k [ t , s ] \varphi: k[x] \to k[t,s] φ : k [ x ] → k [ t , s ] :
k [ x ] → ≃ G a ( k [ x ] ) → ρ k [ x ] End C ( c ) [ x ] φ ↓ φ ↓ φ ∗ ↓ k [ t , s ] → ≃ G a ( k [ t , s ] ) → ρ k [ t , s ] End C ( c ) [ t , s ] \begin{CD}
k[x] @>{\simeq}>> \mathbb{G}_a(k[x]) @>{\rho_{k[x]}}>> \operatorname{End}_{\mathcal{C}}(c)[x] \\
@V{\varphi}VV @V{\varphi}VV @V{\varphi^*}VV \\
k[t,s] @>{\simeq}>> \mathbb{G}_a(k[t,s]) @>{\rho_{k[t,s]}}>> \operatorname{End}_{\mathcal{C}}(c)[t,s]
\end{CD} k [ x ] φ ↓ ⏐ k [ t , s ] ≃ ≃ G a ( k [ x ]) φ ↓ ⏐ G a ( k [ t , s ]) ρ k [ x ] ρ k [ t , s ] End C ( c ) [ x ] φ ∗ ↓ ⏐ End C ( c ) [ t , s ]
where we have used the fact that End C ( φ ) \operatorname{End}_{\mathcal C}(\varphi) End C ( φ ) corresponds to the pullback of polynomials φ ∗ \varphi^* φ ∗ under the identification (9) . Now let φ s : k [ x ] → k [ t , s ] \varphi_s:k[x]\to k[t,s] φ s : k [ x ] → k [ t , s ] , φ t : k [ x ] → k [ t , s ] \varphi_t:k[x]\to k[t,s] φ t : k [ x ] → k [ t , s ] , and φ s + t : k [ x ] → k [ t , s ] \varphi_{s+t}:k[x]\to k[t,s] φ s + t : k [ x ] → k [ t , s ] , respectively, denote the maps defined by sending the generator x x x to s s s , t t t , and s + t s+t s + t , respectively. Then the commutative diagram implies that φ s ∗ u x ≃ ρ k [ t , s ] ( s ) = u s \varphi_s^*u_x\simeq\rho_{k[t,s]}(s)=u_s φ s ∗ u x ≃ ρ k [ t , s ] ( s ) = u s , φ t ∗ u x ≃ ρ k [ t , s ] ( t ) = u t \varphi_t^*u_x\simeq\rho_{k[t,s]}(t)=u_t φ t ∗ u x ≃ ρ k [ t , s ] ( t ) = u t , and φ s + t ∗ u x ≃ ρ k [ t , s ] ( s + t ) = u s + t \varphi_{s+t}^*u_x\simeq\rho_{k[t,s]}(s+t)=u_{s+t} φ s + t ∗ u x ≃ ρ k [ t , s ] ( s + t ) = u s + t . On the other hand, the fact that k [ t , s ] ≃ G a ( k [ t , s ] ) → End C [ t , s ] k[t,s]\simeq\mathbb{G}_a(k[t,s])\to\operatorname{End}_{\mathcal C}[t,s] k [ t , s ] ≃ G a ( k [ t , s ]) → End C [ t , s ] is map of E 1 \mathbb{E}_1 E 1 -spaces tells us that u s + t ≃ u s ∘ u t u_{s+t}\simeq u_s\circ u_t u s + t ≃ u s ∘ u t . Thus, we deduce that
φ s + t ∗ u x ≃ φ s ∗ u x ∘ φ t ∗ u x . \varphi_{s+t}^*u_x\simeq\varphi_s^*u_x\circ\varphi_t^*u_x. φ s + t ∗ u x ≃ φ s ∗ u x ∘ φ t ∗ u x .
If u x = a 0 + a 1 x + ⋯ + a n x n u_x=a_0+a_1x+\cdots+a_nx^n u x = a 0 + a 1 x + ⋯ + a n x n , where a i ∈ Ω ∞ ( End C ( c ) ) a_i\in\Omega^\infty(\operatorname{End}_{\mathcal C}(c)) a i ∈ Ω ∞ ( End C ( c )) , then this identity simplifies to the equivalence,
∑ i = 0 n ∑ k = 0 i ( i k ) a i s i − k t k ≃ ∑ i = 0 n ∑ j = 0 n a i a j s i t j , (10) \sum_{i=0}^{n}\sum_{k=0}^{i}\binom{i}{k}a_i s^{i-k}t^k\simeq\sum_{i=0}^{n}\sum_{j=0}^{n}a_i a_j s^i t^j, \tag*{(10)} i = 0 ∑ n k = 0 ∑ i ( k i ) a i s i − k t k ≃ i = 0 ∑ n j = 0 ∑ n a i a j s i t j , (10)
which imposes a relation on the coefficients a i a_i a i . To find this relation, we compare the coefficients of the components on each side of (10) above. On the left-hand side, the coefficient of s α + β t β s^{\alpha+\beta}t^\beta s α + β t β is ( α + β β ) a α + β \binom{\alpha+\beta}{\beta}a_{\alpha+\beta} ( β α + β ) a α + β , while on the right-hand side, s α t β s^\alpha t^\beta s α t β is a α ∘ a β a_\alpha\circ a_\beta a α ∘ a β , so we obtain the relation,
( α + β β ) a α + β ≃ a α ∘ a β , \binom{\alpha+\beta}{\beta}a_{\alpha+\beta}\simeq a_\alpha\circ a_\beta, ( β α + β ) a α + β ≃ a α ∘ a β ,
which can be seen by induction to impose the relation k ! a k ≃ a 1 k k!a_k\simeq a_1^k k ! a k ≃ a 1 k for any 1 ≤ k ≤ n 1\leq k\leq n 1 ≤ k ≤ n . Thus, a 1 a_1 a 1 is seen to be nilpotent (note that the relation above holds even when an endomorphism a i a_i a i is zero, as it will be for sufficiently large i i i ), so u x u_x u x has the form 1 + N x + ⋯ + N n x n 1+Nx+\cdots+N^nx^n 1 + N x + ⋯ + N n x n for some nilpotent operator N ∈ Ω ∞ ( End C ( c ) ) N\in\Omega^\infty(\operatorname{End}_{\mathcal C}(c)) N ∈ Ω ∞ ( End C ( c )) .
The map G a → A u t ( c ) \mathbb{G}_a\to\mathcal{A}ut(c) G a → A u t ( c ) pulls back along the map Z → G a \mathbb{Z}\to\mathbb{G}_a Z → G a to obtain the map of prestacks Z → A u t ( c ) \mathbb{Z}\to\mathcal{A}ut(c) Z → A u t ( c ) . The canonical automorphism ρ : c → ∼ c \rho:c\xrightarrow{\sim}c ρ : c ∼ c given by the image of 1 ∈ Z 1\in\mathbb{Z} 1 ∈ Z under the induced morphism Z → A u t C ( c ) \mathbb{Z}\to\mathcal{A}ut_{\mathcal C}(c) Z → A u t C ( c ) in the usual definition of S 1 \mathbb{S}^1 S 1 -action on C \mathcal C C is obtained by looking at the map on k k k -points induced by the map Z → A u t ( c ) \mathbb{Z}\to\mathcal{A}ut(c) Z → A u t ( c ) . By naturality, we obtain a commutative diagram,
Z ≃ Z ( k [ t ] ) → 1 k [ x ] ≃ G a ( k [ t ] ) → ρ k [ t ] End C ( c ) [ x ] id Z ↓ ev 1 ↓ ev 1 ∗ ↓ Z ≃ Z ( k ) → 1 k ≃ G a ( k ) → ρ k End C ( c ) , \begin{CD}
\mathbb{Z}\simeq\mathbb{Z}(k[t]) @>{1}>> k[x]\simeq\mathbb{G}_a(k[t]) @>{\rho_{k[t]}}>> \operatorname{End}_{\mathcal C}(c)[x]\\
@V{\operatorname{id}_{\mathbb{Z}}}VV @V{\operatorname{ev}_1}VV @V{\operatorname{ev}_1^*}VV\\
\mathbb{Z}\simeq\mathbb{Z}(k) @>{1}>> k\simeq\mathbb{G}_a(k) @>{\rho_k}>> \operatorname{End}_{\mathcal C}(c),
\end{CD} Z ≃ Z ( k [ t ]) id Z ↓ ⏐ Z ≃ Z ( k ) 1 1 k [ x ] ≃ G a ( k [ t ]) ev 1 ↓ ⏐ k ≃ G a ( k ) ρ k [ t ] ρ k End C ( c ) [ x ] ev 1 ∗ ↓ ⏐ End C ( c ) ,
where 1 : Z → k [ x ] 1:\mathbb{Z}\to k[x] 1 : Z → k [ x ] and 1 : Z → k 1:\mathbb{Z}\to k 1 : Z → k denote the maps given by 1 ↦ 1 1\mapsto1 1 ↦ 1 . This diagram tells us that the canonical automorphism ρ : c → c \rho:c\to c ρ : c → c is given by ev 1 ∗ u x ≃ 1 + N + ⋯ + N n ∈ Ω ∞ ( End C ( c ) ) \operatorname{ev}_1^*u_x\simeq1+N+\cdots+N^n\in\Omega^\infty(\operatorname{End}_{\mathcal C}(c)) ev 1 ∗ u x ≃ 1 + N + ⋯ + N n ∈ Ω ∞ ( End C ( c )) . The latter is clearly a unipotent endomorphism, as claimed. □
Stacks
Coherent (ind-)geometric stacks
We seek to apply [CW23, Proposition 5.30] to the stacks which we study in this paper. To do this, we record results below which show that all stacks considered in the paper are coherent ind-geometric in the sense of Definitions 4.1 and 4.3 of op. cit.
Lemma 3.34. Let Z Z Z be a closed substack of a geometric stack X X X almost of finite type in the sense of [CW23]. Then X ^ Z \widehat{X}_Z X Z is a laft reasonable ind-geometric stack.
Proof. Note that X ^ Z \widehat{X}_Z X Z is a laft prestack since X X X (and hence Z r e d Z_{\mathrm{red}} Z red ) is laft. We first check that Z Z Z is a geometric stack. Since the map Z → X Z\to X Z → X is a closed immersion, pulling back the geometric atlas U → X U\to X U → X gives a faithfully flat map U × X Z → Z U\times_X Z\to Z U × X Z → Z . Moreover, the map U × X Z → U U\times_X Z\to U U × X Z → U is a closed immersion; hence, U × X Z U\times_X Z U × X Z is affine by [CW23, Proposition 3.22]. This implies that the map Z → X Z\to X Z → X is affine by loc. cit. Now, since formal completions only depend on the underlying reduced stack, we can assume Z = Z r e d Z = Z_{\mathrm{red}} Z = Z red is reduced and classical. [GR17b, Chapter 9 Corollary 5.2.3] identifies
X ^ Z ≃ colim n Z n , \widehat{X}_Z \simeq\underset{n}{\operatorname{colim}} Z_n, X Z ≃ n colim Z n ,
where the Z n Z_n Z n are the infinitesimal neighborhoods of Z Z Z constructed in loc. cit. The construction therein inductively defines the Z n Z_n Z n by maps
i n − 1 , ∗ ( Sym n ( T Z / X [ 1 ] ) [ − 1 ] ) ⟶ T Z n − 1 / X , i_{n-1,*}(\operatorname{Sym}^{n}(\mathbb{T}_{Z/X}[1])[-1]) \longrightarrow\mathbb{T}_{Z_{n-1}/X}, i n − 1 , ∗ ( Sym n ( T Z / X [ 1 ]) [ − 1 ]) ⟶ T Z n − 1 / X ,
where i n − 1 : Z → Z n − 1 i_{n-1}: Z \to Z_{n-1} i n − 1 : Z → Z n − 1 and Sym \operatorname{Sym} Sym denotes the symmetric algebra with respect to the tensor product ⊗ \otimes ⊗ on IndCoh ( Z ) \operatorname{IndCoh}(Z) IndCoh ( Z ) .24 By Chapter 8 Proposition 5.5.5 of op. cit. , such a map corresponds to a square-zero extension of Z n − 1 Z_{n-1} Z n − 1 by an ideal I n \mathcal{I}_n I n . At this point, we can already argue that the maps Z n → X Z_n \to X Z n → X are affine and the Z n Z_n Z n are geometric. Arguing inductively, each U n = Z n × X U U_n = Z_n \times_X U U n = Z n × X U is a square-zero extension of an affine scheme. Thus, the maps U n → U U_n \to U U n → U are affine, and they descend along the faithfully flat atlas to Z n → X Z_n \to X Z n → X . Similarly, the same argument shows that the maps i n i_n i n are affine. Now, we claim that the ideal I n \mathcal{I}_n I n is precisely
I n ≃ i n − 1 , ∗ ( Sym n ( L Z / X [ − 1 ] ) ) ∈ APerf ( Z n − 1 ) ≤ 0 . [ 2 5 ] \mathcal{I}_n \simeq i_{n-1,*}(\operatorname{Sym}^{n}(\mathbb{L}_{Z/X}[-1])) \in\operatorname{APerf}(Z_{n-1})^{\leq0}.[^25] I n ≃ i n − 1 , ∗ ( Sym n ( L Z / X [ − 1 ])) ∈ APerf ( Z n − 1 ) ≤ 0 . [ 2 5 ]
We will first show how this proves the lemma and then identify I n \mathcal{I}_n I n with this almost perfect complex. Note that I n \mathcal{I}_n I n is indeed connective, since the affine pushforwards i n − 1 , ∗ i_{n-1,*} i n − 1 , ∗ are t t t -exact on QCoh \operatorname{QCoh} QCoh and the closed immersion Z → X Z \to X Z → X implies that L Z / X [ − 1 ] \mathbb{L}_{Z/X}[-1] L Z / X [ − 1 ] is connective. Furthermore, Z Z Z is almost of finite type: It is classical, so it admits a classical closed immersion into X c l X^{\mathrm{cl}} X cl which is finite type. The Z n Z_n Z n are also almost of finite type. Indeed, arguing inductively again, a square-zero extension of an almost finite type stack by a connective almost perfect complex is itself almost of finite type. This can be checked flat locally using the geometric atlas, where it reduces to the statement that if
I ⟶ B ′ ⟶ B I \longrightarrow B' \longrightarrow B I ⟶ B ′ ⟶ B
is a square-zero extension with B B B almost of finite type and I ∈ APerf ( B ) ≤ 0 I \in\operatorname{APerf}(B)^{\leq0} I ∈ APerf ( B ) ≤ 0 , then H 0 ( B ′ ) H^0(B') H 0 ( B ′ ) is a square-zero extension of H 0 ( B ) H^0(B) H 0 ( B ) by a finite module. Thus, it is of finite type and Noetherian. We then use the long exact sequence in cohomology to conclude that each H i ( B ′ ) H^i(B') H i ( B ′ ) is finite over H 0 ( B ′ ) H^0(B') H 0 ( B ′ ) as desired. Thus, we have a presentation
X ^ Z ≃ colim n Z n , \widehat{X}_Z \simeq\underset{n}{\operatorname{colim}} Z_n, X Z ≃ n colim Z n ,
with all structure maps almost finitely presented closed immersions.25 We can furthermore take all Z n Z_n Z n truncated. Indeed, write Z n , m ≃ τ ≤ m Z n Z_{n,m} \simeq\tau_{\leq m}Z_n Z n , m ≃ τ ≤ m Z n . Then we can write
X ^ Z ≃ colim n , m Z n , m , \widehat{X}_Z \simeq\underset{n,m}{\operatorname{colim}} Z_{n,m}, X Z ≃ n , m colim Z n , m ,
which is a reasonable ind-geometric presentation.
We will now show the isomorphism
I n ≃ i n − 1 , ∗ ( Sym n ( L Z / X [ − 1 ] ) ) . \mathcal{I}_n \simeq i_{n-1,*}(\operatorname{Sym}^{n}(\mathbb{L}_{Z/X}[-1])). I n ≃ i n − 1 , ∗ ( Sym n ( L Z / X [ − 1 ])) .
By [GR17b, Chapter 8 Proposition 5.5.5], it is sufficient to show that
D Z n − 1 S e r r e ( i n − 1 , ∗ ( Sym n ( L Z / X [ − 1 ] ) ) ) ≃ i n − 1 , ∗ ( Sym n ( T Z / X [ 1 ] ) ) . \mathbb{D}^{\mathrm{Serre}}_{Z_{n-1}}\left(i_{n-1,*}(\operatorname{Sym}^{n}(\mathbb{L}_{Z/X}[-1]))\right) \simeq i_{n-1,*}(\operatorname{Sym}^{n}(\mathbb{T}_{Z/X}[1])). D Z n − 1 Serre ( i n − 1 , ∗ ( Sym n ( L Z / X [ − 1 ])) ) ≃ i n − 1 , ∗ ( Sym n ( T Z / X [ 1 ])) .
I ⟶ B ′ ⟶ B . I \longrightarrow B' \longrightarrow B. I ⟶ B ′ ⟶ B .
Since I I I is connective, we see that H 0 ( B ′ ) → H 0 ( B ) H^0(B') \to H^0(B) H 0 ( B ′ ) → H 0 ( B ) is surjective as required.
We then compute
D Z S e r r e ( L Z / X [ − 1 ] ) ≃ T Z / X [ 1 ] . \mathbb{D}^{\mathrm{Serre}}_{Z}\left(\mathbb{L}_{Z/X}[-1]\right) \simeq\mathbb{T}_{Z/X}[1]. D Z Serre ( L Z / X [ − 1 ] ) ≃ T Z / X [ 1 ] .
Hence,
D Z n − 1 S e r r e ( i n − 1 , ∗ ( Sym n ( L Z / X [ − 1 ] ) ) ) ≃ i n − 1 , ∗ ( D Z S e r r e ( Sym n ( L Z / X [ − 1 ] ) ) ) ≃ i n − 1 , ∗ ( Sym n ( T Z / X [ 1 ] ) ) , \begin{aligned}
\mathbb{D}^{\mathrm{Serre}}_{Z_{n-1}}\left(i_{n-1,*}\left(\operatorname{Sym}^{n}\left(\mathbb{L}_{Z/X}[-1]\right)\right)\right) \simeq i_{n-1,*}\left(\mathbb{D}^{\mathrm{Serre}}_{Z}\left(\operatorname{Sym}^{n}\left(\mathbb{L}_{Z/X}[-1]\right)\right)\right) \\
&\simeq i_{n-1,*}\left(\operatorname{Sym}^{n}\left(\mathbb{T}_{Z/X}[1]\right)\right),
\end{aligned} D Z n − 1 Serre ( i n − 1 , ∗ ( Sym n ( L Z / X [ − 1 ] ) ) ) ≃ i n − 1 , ∗ ( D Z Serre ( Sym n ( L Z / X [ − 1 ] ) ) ) ≃ i n − 1 , ∗ ( Sym n ( T Z / X [ 1 ] ) ) ,
where we note that Serre duality exchanges the usual tensor product on QCoh ( Z ) \operatorname{QCoh}(Z) QCoh ( Z ) with the tensor product ⊗ \otimes ⊗ on IndCoh ( Z ) \operatorname{IndCoh}(Z) IndCoh ( Z ) . □
Lemma 3.35. Suppose that Z Z Z is a closed substack of an algebraic stack X X X almost of finite type and with affine diagonal. Suppose also that Z → X Z \to X Z → X is almost of finite presentation. Denote by U U U the open complement to Z → X Z \to X Z → X . Then X ^ Z \widehat{X}_{Z} X Z is a coherent ind-geometric stack, and Z Z Z , U U U , and X X X are coherent geometric stacks.
Proof. We seek to apply Lemma 3.34 . By assumption X X X admits a smooth affine atlas U → X \mathcal{U} \to X U → X , which we can use both as the geometric atlas and the coherent atlas. Indeed, by [HA, Proposition 7.2.4.31], U = Spec A \mathcal{U} = \operatorname{Spec} A U = Spec A is Noetherian, since U \mathcal{U} U is almost of finite type. This implies that X X X is coherent by the remark after [CW23, Definition 3.3]. This allows us to apply Lemma 3.34 . Moreover, pulling back the atlas U → X \mathcal{U} \to X U → X to Z Z Z gives a geometric atlas for Z Z Z . Since Z → X Z \to X Z → X is a closed immersion almost of finite presentation, this implies that U × X Z \mathcal{U} \times_X Z U × X Z is also a Noetherian affine scheme, which gives a faithfully flat (in fact, smooth) atlas of Z Z Z . Thus, Z Z Z is coherent.
We next claim that the Z n Z_n Z n defined in the proof of Lemma 3.34 are also coherent geometric stacks. Pulling back the smooth atlas U → X \mathcal{U} \to X U → X to Z n Z_n Z n to obtain U n : = Z n × X U \mathcal{U}_n := Z_n \times_X \mathcal{U} U n := Z n × X U , these are affine schemes U n = Spec A n \mathcal{U}_n = \operatorname{Spec} A_n U n = Spec A n as in the proof of Lemma 3.34 . Moreover, U n \mathcal{U}_n U n is laft because U \mathcal{U} U , X X X , and Z n Z_n Z n are. Thus, U n \mathcal{U}_n U n is almost of finite type, so A n A_n A n is Noetherian. The presentation
X ^ Z ≃ colim n , m Z n , m \widehat{X}_{Z} \simeq\underset{n,m}{\operatorname{colim}} Z_{n,m} X Z ≃ n , m colim Z n , m
then supplies the desired coherent ind-geometric presentation of X ^ Z \widehat{X}_{Z} X Z .
Lastly, we show that U U U is a coherent geometric stack. Pulling back the smooth atlas U → X \mathcal{U} \to X U → X we obtain a smooth atlas of U U U . Since U U U is quasi-compact,27 {}^{27} 27 this induces a smooth affine atlas V → U \mathcal{V} \to U V → U . Hence, U U U is geometric. Moreover, V → U \mathcal{V} \to\mathcal{U} V → U is locally of finite presentation, since it is in particular an open immersion. This implies that V \mathcal{V} V is laft and affine, hence almost of finite type. Thus, it is the spectrum of a Noetherian k k k -algebra, as required. □
Corollary 3.36. In the setup of Lemma 3.35, L ( X ^ Z ) \mathcal{L}(\widehat{X}_{Z}) L ( X Z ) is a laft coherent ind-geometric stack, and L X \mathcal{L}X L X , L Z \mathcal{L}Z L Z , and L U \mathcal{L}U L U are coherent geometric stacks.
Proof. If X X X is a quasi-compact algebraic stack with affine diagonal, then so is L X \mathcal{L}X L X . Similarly, the map L Z → L X \mathcal{L}Z \to\mathcal{L}X L Z → L X is a closed immersion by Lemma 3.51. Since Z Z Z and X X X are almost of finite type, so are L Z \mathcal{L}Z L Z and L X \mathcal{L}X L X . Thus, L Z → L X \mathcal{L}Z \to\mathcal{L}X L Z → L X is almost finitely presented, which allows us to apply Lemma 3.35. The argument for L U \mathcal{L}U L U is as in the proof of Lemma 3.35. □
Lemma 3.37. Let X X X be a coherent ind-geometric stack in the sense of [CW23, Definitions 4.1 and 4.2]. Let A A A be a Noetherian k k k -algebra, and let X A : = X × Spec A X_A := X \times\operatorname{Spec} A X A := X × Spec A denote the base change. If the coherent presentation X ≃ colim n X n X \simeq\operatorname{colim}_n X_n X ≃ colim n X n has all X n X_n X n equipped with a faithfully flat atlas Spec A n → X n \operatorname{Spec} A_n \to X_n Spec A n → X n such that A n A_n A n are almost of finite type, then X A X_A X A is coherent ind-geometric.
Proof. Since the colimit colim n X n \operatorname{colim}_{n} X_n colim n X n is a colimit in the category S t k ^ k \widehat{\mathrm{Stk}}_k Stk k of convergent stacks, finite limits commute with this colimit by [HTT, Example 7.3.4.7] (cf. also the proof of [CW23, Proposition 4.21]). In particular, this implies that
X A ≃ colim n ( X n × Spec A ) . X_A \simeq\operatorname{colim}_{n}(X_n \times\operatorname{Spec} A). X A ≃ colim n ( X n × Spec A ) .
Moreover, if U n = Spec A n → X n U_n = \operatorname{Spec} A_n \to X_n U n = Spec A n → X n is a faithfully flat atlas of X n X_n X n with U n U_n U n almost of finite type, then U n × Spec A → X n × Spec A U_n \times\operatorname{Spec} A \to X_n \times\operatorname{Spec} A U n × Spec A → X n × Spec A is also an atlas exhibiting X n × Spec A X_n \times\operatorname{Spec} A X n × Spec A as locally Noetherian in the sense of [CW23, Definition 3.3]. Indeed, the product U n × Spec A U_n \times\operatorname{Spec} A U n × Spec A is Noetherian, because the U n U_n U n are almost of finite type and A A A is Noetherian. Thus, we conclude that X A X_A X A has a reasonable presentation by locally Noetherian geometric stacks; in particular, this is a coherent presentation by the remark after loc. cit.
Corollary 3.38. Let A A A be a Noetherian k k k -algebra. In the setup of Corollary 3.36, the base change along Spec A → Spec k \operatorname{Spec} A \to\operatorname{Spec} k Spec A → Spec k of each of the stacks L U \mathcal{L}U L U , L Z \mathcal{L}Z L Z , L X \mathcal{L}X L X , and L ( X d Z ) \mathcal{L}(X_dZ) L ( X d Z ) is coherent ind-geometric.
Proof. All of L U \mathcal{L}U L U , L Z \mathcal{L}Z L Z and L X \mathcal{L}X L X are locally Noetherian (in fact, Noetherian) geometric stacks. Hence, Lemma 3.37 applies to them. For L ( X d Z ) \mathcal{L}(X_dZ) L ( X d Z ) , we can again apply loc. cit. Indeed, observe that all stacks in the presentation constructed in Corollary 3.36 and Lemma 3.35 are locally Noetherian.
Properties of !-pullback and *-pushforward functors
Lemma 3.39. Let f : X → Y f : X \to Y f : X → Y be an étale morphism of laft Artin stacks. Then f ! : IndCoh ( Y ) → IndCoh ( X ) f^! : \operatorname{IndCoh}(Y) \to\operatorname{IndCoh}(X) f ! : IndCoh ( Y ) → IndCoh ( X ) is t t t -exact.
Proof. Since f f f is étale, f ! ≃ f ∗ , IndCoh f^! \simeq f^{*,\operatorname{IndCoh}} f ! ≃ f ∗ , IndCoh , so it suffices to show that f ∗ , IndCoh f^{*,\operatorname{IndCoh}} f ∗ , IndCoh is t t t -exact. By [Gai13, Proposition 11.7.5(b)], the functors Ψ Y : IndCoh ( Y ) → QCoh ( Y ) \Psi_Y : \operatorname{IndCoh}(Y) \to\operatorname{QCoh}(Y) Ψ Y : IndCoh ( Y ) → QCoh ( Y ) and Ψ X : IndCoh ( X ) → QCoh ( X ) \Psi_X : \operatorname{IndCoh}(X) \to\operatorname{QCoh}(X) Ψ X : IndCoh ( X ) → QCoh ( X ) are t t t -exact. Moreover, they intertwine f ∗ f^* f ∗ and f ∗ , IndCoh f^{*,\operatorname{IndCoh}} f ∗ , IndCoh in the sense that Ψ X ∘ f ∗ , IndCoh ≃ f ∗ ∘ Ψ Y \Psi_X \circ f^{*,\operatorname{IndCoh}} \simeq f^* \circ\Psi_Y Ψ X ∘ f ∗ , IndCoh ≃ f ∗ ∘ Ψ Y .
We claim that, in order to show that f ∗ , IndCoh f^{*,\operatorname{IndCoh}} f ∗ , IndCoh is t t t -exact, it suffices to show that f ∗ : QCoh ( Y ) → QCoh ( X ) f^* : \operatorname{QCoh}(Y) \to\operatorname{QCoh}(X) f ∗ : QCoh ( Y ) → QCoh ( X ) is t t t -exact.
Indeed, suppose that f ∗ f^* f ∗ is t t t -exact, let F ∈ IndCoh ( Y ) ≤ 0 \mathcal{F} \in\operatorname{IndCoh}(Y)^{\leq0} F ∈ IndCoh ( Y ) ≤ 0 . Then f ∗ Ψ Y ( F ) ≃ Ψ X f ∗ , IndCoh ( F ) ∈ QCoh ( X ) ≤ 0 f^*\Psi_Y(\mathcal{F}) \simeq\Psi_Xf^{*,\operatorname{IndCoh}}(\mathcal{F}) \in\operatorname{QCoh}(X)^{\leq0} f ∗ Ψ Y ( F ) ≃ Ψ X f ∗ , IndCoh ( F ) ∈ QCoh ( X ) ≤ 0 . Because Ψ X \Psi_X Ψ X is t t t -exact, we have
Ψ X ( τ ≥ 1 f ∗ , IndCoh F ) ≃ τ ≥ 1 Ψ X ( f ∗ , IndCoh ( F ) . \Psi_X(\tau^{\geq1}f^{*,\operatorname{IndCoh}}\mathcal{F}) \simeq\tau^{\geq1}\Psi_X(f^{*,\operatorname{IndCoh}}(\mathcal{F}). Ψ X ( τ ≥ 1 f ∗ , IndCoh F ) ≃ τ ≥ 1 Ψ X ( f ∗ , IndCoh ( F ) .
Since Ψ X ( f ∗ , IndCoh ( F ) \Psi_X(f^{*,\operatorname{IndCoh}}(\mathcal{F}) Ψ X ( f ∗ , IndCoh ( F ) is coconnective, τ ≥ 1 Ψ X ( f ∗ , IndCoh ( F ) ≃ 0 \tau^{\geq1}\Psi_X(f^{*,\operatorname{IndCoh}}(\mathcal{F}) \simeq0 τ ≥ 1 Ψ X ( f ∗ , IndCoh ( F ) ≃ 0 , meaning Ψ X ( τ ≥ 1 f ∗ , IndCoh F ) ≃ 0 \Psi_X(\tau^{\geq1}f^{*,\operatorname{IndCoh}}\mathcal{F}) \simeq0 Ψ X ( τ ≥ 1 f ∗ , IndCoh F ) ≃ 0 . On the other hand, by [Gai13, Proposition 11.7.5(b)] again, Ψ X : IndCoh ( X ) + → ∼ QCoh ( X ) + \Psi_X : \operatorname{IndCoh}(X)^+ \xrightarrow{\sim} \operatorname{QCoh}(X)^+ Ψ X : IndCoh ( X ) + ∼ QCoh ( X ) + is an equivalence. Thus, τ ≥ 1 f ∗ , IndCoh F ≃ 0 \tau^{\geq1}f^{*,\operatorname{IndCoh}}\mathcal{F} \simeq0 τ ≥ 1 f ∗ , IndCoh F ≃ 0 , implying that f ∗ , IndCoh F ∈ IndCoh ( X ) ≤ 0 f^{*,\operatorname{IndCoh}}\mathcal{F} \in\operatorname{IndCoh}(X)^{\leq0} f ∗ , IndCoh F ∈ IndCoh ( X ) ≤ 0 .
On the other hand, because f f f is eventually coconnective, f ∗ , IndCoh f^{*,\operatorname{IndCoh}} f ∗ , IndCoh preserves the subcategory of eventually coconnective, or bounded below, objects. Thus, if F ∈ IndCoh ( Y ) ≥ 0 ⊂ IndCoh ( Y ) + \mathcal{F} \in\operatorname{IndCoh}(Y)^{\geq0} \subset\operatorname{IndCoh}(Y)^+ F ∈ IndCoh ( Y ) ≥ 0 ⊂ IndCoh ( Y ) + , we see f ∗ , IndCoh F ∈ IndCoh ( X ) + f^{*,\operatorname{IndCoh}}\mathcal{F} \in\operatorname{IndCoh}(X)^+ f ∗ , IndCoh F ∈ IndCoh ( X ) + . Now note that
Ψ X f ∗ , IndCoh F ≃ f ∗ Ψ Y ( F ) ∈ QCoh ( X ) ≥ 0 \Psi_X f^{*,\operatorname{IndCoh}}\mathcal{F} \simeq f^*\Psi_Y(\mathcal{F}) \in\operatorname{QCoh}(X)^{\geq0} Ψ X f ∗ , IndCoh F ≃ f ∗ Ψ Y ( F ) ∈ QCoh ( X ) ≥ 0
by the t t t -exactness of Ψ Y \Psi_Y Ψ Y and f ∗ f^* f ∗ . Because Ψ X \Psi_X Ψ X restricts to a t t t -exact equivalence IndCoh ( X ) + → ∼ QCoh ( X ) + \operatorname{IndCoh}(X)^+ \xrightarrow{\sim} \operatorname{QCoh}(X)^+ IndCoh ( X ) + ∼ QCoh ( X ) + , we therefore have that f ∗ , IndCoh F ∈ IndCoh ( X ) ≥ 0 f^{*,\operatorname{IndCoh}}\mathcal{F} \in\operatorname{IndCoh}(X)^{\geq0} f ∗ , IndCoh F ∈ IndCoh ( X ) ≥ 0 . We conclude simply by noting that f ∗ : QCoh ( Y ) → QCoh ( X ) f^* : \operatorname{QCoh}(Y) \to\operatorname{QCoh}(X) f ∗ : QCoh ( Y ) → QCoh ( X ) is t t t -exact because f f f is étale ([GR17a, Chapter 3 Proposition 1.5.4(b)]).
Lemma 3.40. Let Y Y Y be a laft, quasi-compact algebraic stack with affine diagonal. Then Y Y Y is a QCA stack in the sense of [DG12, Definition 1.1.8].
Proof. Geometric stabilizers are affine as they are base changes of the affine diagonal. The classical inertia stack is finitely presented because Y c l Y^{\mathrm{cl}} Y cl is locally of finite type with affine diagonal. Thus, ( L Y ) c l → Y c l (\mathcal{L}Y)^{\mathrm{cl}} \to Y^{\mathrm{cl}} ( L Y ) cl → Y cl is a base change of the diagonal of Y c l Y^{\mathrm{cl}} Y cl . This diagonal is qcqs; hence, it is of finite presentation.
Lemma 3.41. Let f : X → Y f : X \to Y f : X → Y be a finitely presented morphism in S t k l a f t \mathrm{Stk}_{\mathrm{laft}} Stk laft which is representable by algebraic spaces, and suppose that Y Y Y is a laft, quasi-compact algebraic stack with affine diagonal. If X X X has affine diagonal, then it is QCA in the sense of [DG12, Definition 1.1.8].
Proof. We will show that X X X is a laft quasi-compact algebraic stack and then use Lemma 3.40 to conclude. Using a smooth affine atlas U → Y U \to Y U → Y , we obtain an algebraic space V = X × Y U V = X \mathbin{\times_Y} U V = X × Y U with V → U V \to U V → U finitely presented. In particular, it is quasi-compact; hence, V V V is quasi-compact. Choosing an étale surjection W → V W \to V W → V with W W W affine, we obtain a smooth affine atlas of X X X . □ \square □
Corollary 3.42. In the setup of Lemma 3.41 , suppose that f f f is furthermore smooth. Then
f ! : IndCoh ( Y ) ⟶ IndCoh ( X ) f^! : \operatorname{IndCoh}(Y) \longrightarrow\operatorname{IndCoh}(X) f ! : IndCoh ( Y ) ⟶ IndCoh ( X ) restricts to a functor
f ! ∣ Coh ( Y ) : Coh ( Y ) ⟶ Coh ( X ) . f^!|_{\operatorname{Coh}(Y)} : \operatorname{Coh}(Y) \longrightarrow\operatorname{Coh}(X). f ! ∣ Coh ( Y ) : Coh ( Y ) ⟶ Coh ( X ) . In particular, f ! f^! f ! preserves compact objects.
Proof. By Lemma 3.41 , we can use the characterization of coherent sheaves on QCA stacks given in [DG12, Section 3.3.1]. In particular, loc. cit. characterizes Coh ( Y ) ⊂ IndCoh ( Y ) \operatorname{Coh}(Y) \subset\operatorname{IndCoh}(Y) Coh ( Y ) ⊂ IndCoh ( Y ) as those sheaves F \mathcal{F} F such that for any smooth y : S → Y y : S \to Y y : S → Y with S S S affine, y ! F ∈ Coh ( S ) y^!\mathcal{F} \in\operatorname{Coh}(S) y ! F ∈ Coh ( S ) . Thus, suppose given a smooth map x : T → X x : T \to X x : T → X with T T T an affine scheme, and let F ∈ Coh ( Y ) \mathcal{F} \in\operatorname{Coh}(Y) F ∈ Coh ( Y ) . Since f ∘ x f \circ x f ∘ x is smooth, we find ( f ∘ x ) ! F ≃ x ! f ! F ∈ Coh ( T ) (f \circ x)^!\mathcal{F} \simeq x^!f^!\mathcal{F} \in\operatorname{Coh}(T) ( f ∘ x ) ! F ≃ x ! f ! F ∈ Coh ( T ) as desired. □ \square □
Lemma 3.43. Let X X X be a laft algebraic stack, and let j : U ↪ X j : U \hookrightarrow X j : U ↪ X be an almost finitely presented open immersion. Then the pushforward
j ∗ : IndCoh ( U ) ⟶ IndCoh ( X ) j_* : \operatorname{IndCoh}(U) \longrightarrow\operatorname{IndCoh}(X) j ∗ : IndCoh ( U ) ⟶ IndCoh ( X ) is fully faithful.
Proof. For derived schemes, this is [Gai13, Lemma 4.1.1]. Let S → X S \to X S → X be a smooth atlas. Consider the Cartesian diagram
U × X S → j ′ S p ′ ↓ ↓ p U → j X \begin{CD}
U \mathbin{\times_X} S @>{j'}>> S \\
@V{p'}VV @VV{p}V \\
U @>{j}>> X
\end{CD} U × X S p ′ ↓ ⏐ U j ′ j S ↓ ⏐ p X
Note that j ′ j' j ′ is also smooth and surjective; hence, U U U is an algebraic stack as well. We would like to show that the counit
j ! j ∗ ⟶ id j^!j_* \longrightarrow\operatorname{id} j ! j ∗ ⟶ id
is an isomorphism. We can pull it back along p ′ p' p ′ to obtain
p ′ ! j ! j ∗ ≃ j ′ ! p ! j ∗ ≃ j ′ ! j ∗ ′ p ′ ! p'^!j^!j_* \simeq j'^!p^!j_* \simeq j'^!j'_*p'^! p ′ ! j ! j ∗ ≃ j ′ ! p ! j ∗ ≃ j ′ ! j ∗ ′ p ′ !
Thus, the pulled-back counit becomes
j ′ ! j ∗ ′ p ′ ! → ∼ p ′ ! . j'^!j'_*p'^! \xrightarrow{\sim} p'^!. j ′ ! j ∗ ′ p ′ ! ∼ p ′ ! .
Moreover, tracing through the definitions, one finds that this map is precisely the counit of the ( j ′ ! , j ∗ ′ ) (j'^!,j'_*) ( j ′ ! , j ∗ ′ ) -adjunction precomposed with p ′ ! p'^! p ′ ! . Since p ′ p' p ′ is conservative, we conclude. Indeed, conservativity follows from descent along p ′ p' p ′ . □ \square □
Lemma 3.44. Let X X X be an algebraic stack almost of finite type and with affine diagonal, and let j : U ↪ X j : U \hookrightarrow X j : U ↪ X be an open substack. Then the pushforward
j ∗ : IndCoh ( U ) ⟶ IndCoh ( X ) j_* : \operatorname{IndCoh}(U) \longrightarrow\operatorname{IndCoh}(X) j ∗ : IndCoh ( U ) ⟶ IndCoh ( X ) is t t t -bounded.
Proof. Consider the Cartesian diagram
U × X S → j ′ S p ′ ↓ ↓ p U → j X \begin{CD}
U \times_X S @>{j'}>> S \\
@V{p'}VV @VV{p}V \\
U @>{j}>> X
\end{CD} U × X S p ′ ↓ ⏐ U j ′ j S ↓ ⏐ p X
where S → X S \to X S → X is a smooth affine atlas of X X X . By [Gai13, Proposition 11.7.5] and Corollary 7.5.4 of op. cit., the functors p ! p_! p ! and p ! ′ p'_! p ! ′ are t t t -exact up to a shift; hence, they are t t t -bounded. Thus, base change along the diagram implies that it is enough to check that j ∗ j_* j ∗ is t t t -bounded. Since open immersions are schematic, U × X S → S U \times_X S \to S U × X S → S is an open immersion of schemes; thus, we have reduced to the case when j : U ↪ X j : U \hookrightarrow X j : U ↪ X is an open immersion of schemes with X = Spec A X = \operatorname{Spec} A X = Spec A affine. Note that j j j is left t t t -exact by [GR17a, Chapter 4 Lemma 4.1.4], as it is right adjoint to a t t t -exact functor. To obtain the upper bound, we will first show that quasi-coherent pushforward j ∗ Q C o h j_*^{\mathrm{QCoh}} j ∗ QCoh is t t t -bounded. Write the scheme
U = ⋃ i = 1 r D ( f i ) ⊂ Spec A , U = \bigcup_{i=1}^{r} D(f_i) \subset\operatorname{Spec} A, U = i = 1 ⋃ r D ( f i ) ⊂ Spec A ,
where f i ∈ H 0 ( A ) f_i \in H^0(A) f i ∈ H 0 ( A ) . 28 We will induct on r r r . In the case r = 1 r = 1 r = 1 , U U U is affine and quasi-coherent pushforward j ∗ Q C o h j_*^{\mathrm{QCoh}} j ∗ QCoh is just restriction of modules along a localization, which is t t t -bounded. For r > 1 r > 1 r > 1 , write
U = T 1 ∪ T 2 , T 1 = ⋃ i = 1 r − 1 D ( f i ) , T 2 = D ( f r ) . U = T_1 \cup T_2,\qquad T_1 = \bigcup_{i=1}^{r-1} D(f_i),\qquad T_2 = D(f_r). U = T 1 ∪ T 2 , T 1 = i = 1 ⋃ r − 1 D ( f i ) , T 2 = D ( f r ) .
Write the inclusions j i : T i → X j_i : T_i \to X j i : T i → X and j 12 : T 1 ∩ T 2 : = T 1 × X T 2 → X j_{12} : T_1 \cap T_2 := T_1 \times_X T_2 \to X j 12 : T 1 ∩ T 2 := T 1 × X T 2 → X . 29 Quasi-coherent pushforward can be evaluated as the fiber in the Mayer–Vietoris triangle
j ∗ Q C o h ( F ) → j 1 ∗ Q C o h ( F ∣ T 1 ) ⊕ j 2 ∗ Q C o h ( F ∣ T 2 ) → j 12 ∗ Q C o h ( F ∣ T 1 ∩ T 2 ) . j_{*}^{\mathrm{QCoh}}(\mathcal{F}) \to j_{1*}^{\mathrm{QCoh}}(\mathcal{F}|_{T_1}) \oplus j_{2*}^{\mathrm{QCoh}}(\mathcal{F}|_{T_2}) \to j_{12*}^{\mathrm{QCoh}}(\mathcal{F}|_{T_1 \cap T_2}). j ∗ QCoh ( F ) → j 1 ∗ QCoh ( F ∣ T 1 ) ⊕ j 2 ∗ QCoh ( F ∣ T 2 ) → j 12 ∗ QCoh ( F ∣ T 1 ∩ T 2 ) .
Induction gives that the middle term is bounded. For the final term, we have
T 1 ∩ T 2 ≃ ⋃ i = 1 r − 1 D ( f i f r ) , T_1 \cap T_2 \simeq\bigcup_{i=1}^{r-1} D(f_i f_r), T 1 ∩ T 2 ≃ i = 1 ⋃ r − 1 D ( f i f r ) ,
so the final term is bounded as well. We conclude by rotating the Mayer–Vietoris triangle to obtain
j 12 ∗ Q C o h ( F ∣ T 1 ∩ T 2 ) [ − 1 ] → j ∗ Q C o h ( F ) → j 1 ∗ Q C o h ( F ∣ T 1 ) ⊕ j 2 ∗ Q C o h ( F ∣ T 2 ) . j_{12*}^{\mathrm{QCoh}}(\mathcal{F}|_{T_1 \cap T_2})[-1] \to j_*^{\mathrm{QCoh}}(\mathcal{F}) \to j_{1*}^{\mathrm{QCoh}}(\mathcal{F}|_{T_1}) \oplus j_{2*}^{\mathrm{QCoh}}(\mathcal{F}|_{T_2}). j 12 ∗ QCoh ( F ∣ T 1 ∩ T 2 ) [ − 1 ] → j ∗ QCoh ( F ) → j 1 ∗ QCoh ( F ∣ T 1 ) ⊕ j 2 ∗ QCoh ( F ∣ T 2 ) .
Thus, we see that the quasi-coherent pushforward j ∗ Q C o h j_*^{\mathrm{QCoh}} j ∗ QCoh is t t t -bounded on QCoh \operatorname{QCoh} QCoh . To extend to IndCoh \operatorname{IndCoh} IndCoh , we use that both Ψ X \Psi_X Ψ X and Ψ U \Psi_U Ψ U are t t t -exact equivalences on the bounded below categories by [Gai13, Proposition 11.7.5]. Hence,
Ψ X j ∗ ≃ j ∗ Q C o h Ψ U . \Psi_X j_* \simeq j_*^{\mathrm{QCoh}} \Psi_U. Ψ X j ∗ ≃ j ∗ QCoh Ψ U .
Since both j ∗ j_* j ∗ and j ∗ Q C o h j_*^{\mathrm{QCoh}} j ∗ QCoh are left t t t -bounded, we conclude.
Lemma 3.45. Let X X X be an algebraic stack almost of finite type and with affine diagonal. Let Z ↪ X Z \hookrightarrow X Z ↪ X be a closed substack. Then the ! ! ! -pullback
i ^ ! : IndCoh ( X ) → IndCoh ( X ^ Z ) \widehat{i}^{!} : \operatorname{IndCoh}(X) \to\operatorname{IndCoh}(\widehat{X}_Z) i ! : IndCoh ( X ) → IndCoh ( X Z ) along the inclusion i ^ : X ^ Z → X \widehat{i} : \widehat{X}_Z \to X i : X Z → X is t t t -bounded.
Proof. As in Lemma 3.44 , we reduce to the case when X X X is a scheme by pulling back along the atlas S → X S \to X S → X . Then, we apply Proposition 7.4.5 and Lemma 7.4.8 of [Gai13] to conclude, noting that the inclusion
IndCoh Z ( X ) ⊂ IndCoh ( X ) \operatorname{IndCoh}_{\mathcal{Z}}(X) \subset\operatorname{IndCoh}(X) IndCoh Z ( X ) ⊂ IndCoh ( X )
is t t t -exact. □ \square □
Quotient stacks
Here we collect various results on the structure of global quotient stacks (i.e. quotients of derived schemes by actions of group schemes), starting with the facts that open and closed immersions of quotient stacks are themselves quotient stacks.
Lemma 3.46. Let Z \mathcal{Z} Z be a closed substack of a stack X / G X/G X / G . Then there is an isomorphism Z / G → ∼ Z \mathcal{Z}/G \xrightarrow{\sim} \mathcal{Z} Z / G ∼ Z for a G G G -invariant closed subscheme Z ↪ X \mathcal{Z} \hookrightarrow X Z ↪ X .
Proof. Since closed immersions are stable under base change, the map Z : = Z × X / G X → X \mathcal{Z} := \mathcal{Z} \times_{X/G} X \to X Z := Z × X / G X → X is a closed immersion. Thus, Z \mathcal{Z} Z is a scheme. By construction, Z → Z \mathcal{Z} \to\mathcal{Z} Z → Z is a G G G -torsor. Thus, the natural G G G -equivariant map Z → Z \mathcal{Z} \to\mathcal{Z} Z → Z factors through the quotient Z / G \mathcal{Z}/G Z / G , giving Z → Z / G → ∼ Z \mathcal{Z} \to\mathcal{Z}/G \xrightarrow{\sim} \mathcal{Z} Z → Z / G ∼ Z . □ \square □
Lemma 3.47. Let U \mathcal{U} U be an open substack of a stack X / G X/G X / G . Then there is an isomorphism U / G → ∼ U \mathcal{U}/G \xrightarrow{\sim} \mathcal{U} U / G ∼ U for a G G G -invariant open subscheme U ↪ X \mathcal{U} \hookrightarrow X U ↪ X .
Proof. The proof is the same as the one above for Lemma 3.46 . □ \square □
We also have the following lemma describing the formal completion along a G G G -equivariant map of derived schemes.
Lemma 3.48. Let Y Y Y and U U U be prestacks equipped with an action of a smooth group prestack G G G , and let f : Y → U f : Y \to U f : Y → U be a G G G -equivariant map. Then the formal completion U ^ Y \widehat{U}_Y U Y has a canonical G G G -action, and there is a canonical equivalence U ^ Y / G → ∼ ( U / G ) ^ Y / G \widehat{U}_Y/G \xrightarrow{\sim} \widehat{(U/G)}_{Y/G} U Y / G ∼ ( U / G ) Y / G .
Proof. The canonical G G G -action on U ^ Y \widehat{U}_Y U Y is given as follows. Since ( − ) d R (-)_{\mathrm{dR}} ( − ) dR preserves limits, G d R G_{\mathrm{dR}} G dR has a group structure coming from that of G G G , and U d R U_{\mathrm{dR}} U dR and Y d R Y_{\mathrm{dR}} Y dR have natural G d R G_{\mathrm{dR}} G dR -actions coming from the G G G -actions on U U U and Y Y Y . These G d R G_{\mathrm{dR}} G dR -actions restrict along the group map G → G d R G \to G_{\mathrm{dR}} G → G dR to give G G G -actions on U d R U_{\mathrm{dR}} U dR and Y d R Y_{\mathrm{dR}} Y dR . Taking the fiber product U × U d R Y d R U \times_{U_{\mathrm{dR}}} Y_{\mathrm{dR}} U × U dR Y dR in the category of G G G -prestacks gives the desired G G G -action on U ^ Y \widehat{U}_Y U Y .
In order to show the desired equivalence, we note that there is a natural morphism U ^ Y / G → ( U / G ) ^ Y / G \widehat{U}_Y/G \to\widehat{(U/G)}_{Y/G} U Y / G → ( U / G ) Y / G induced by the canonical maps U → U / G U \to U/G U → U / G and Y → Y / G Y \to Y/G Y → Y / G . We proceed by showing that this natural map induces an isomorphism between the spaces of S S S -points of each stack, for an arbitrary affine derived scheme S S S . By definition, ( U / G ) ^ Y / G ( S ) = Map ( S , ( U / G ) ^ Y / G ) \widehat{(U/G)}_{Y/G}(S) = \operatorname{Map}(S,\widehat{(U/G)}_{Y/G}) ( U / G ) Y / G ( S ) = Map ( S , ( U / G ) Y / G ) is the ∞ \infty ∞ -groupoid of compatible diagrams,
{ S r e d → Y / G ↓ ↓ f S → U / G } . \left\{
\begin{CD}
S_{\mathrm{red}} @>>> Y/G \\
@VVV @VVfV \\
S @>>> U/G
\end{CD}
\right\}. ⎩ ⎨ ⎧ S red ↓ ⏐ S Y / G ↓ ⏐ f U / G ⎭ ⎬ ⎫ . where by abuse of notation we use f f f also to denote the induced map Y / G → U / G Y/G \to U/G Y / G → U / G . That is, an S S S -point of ( U / G ) ^ Y / G \widehat{(U/G)}_{Y/G} ( U / G ) Y / G is an S S S -point of U / G U/G U / G together with a chosen factorization through f f f of its restriction to S r e d S_{\mathrm{red}} S red . At the same time, an S S S -point of ( U / G ) ^ Y / G \widehat{(U/G)}_{Y/G} ( U / G ) Y / G is the data of a G G G -torsor P → S P \to S P → S together with a G G G -equivariant map θ : P → U \theta: P \to U θ : P → U . Thus, the S S S -points of ( U / G ) ^ Y / G \widehat{(U/G)}_{Y/G} ( U / G ) Y / G are equivalently compatible
diagrams,
{ P × S S r e d → Y ↓ ↓ f P → θ U \left\{
\begin{CD}
P \times_S S_{\mathrm{red}} @>>> Y \\
@VVV @VV f V \\
P @>{\theta}>> U
\end{CD}
\right. ⎩ ⎨ ⎧ P × S S red ↓ ⏐ P θ Y ↓ ⏐ f U
i.e. the data of a G G G -torsor P → S P \to S P → S together with a chosen factorization of P × S S r e d P \times_S S_{\mathrm{red}} P × S S red through f f f .
On the other hand, the S S S -points of ( U Y / G ^ ) ( S ) (\widehat{U_Y/G})(S) ( U Y / G ) ( S ) is the ∞ \infty ∞ -groupoid whose points consist of a chosen G G G -torsor P → S P \to S P → S together with a G G G -equivariant map p : P → U Y ^ p:P \to\widehat{U_Y} p : P → U Y , and a map p : P → U Y ^ p:P \to\widehat{U_Y} p : P → U Y is a point in the ∞ \infty ∞ -groupoid of compatible diagrams,
{ P r e d → Y ↓ ↓ f P → U \left\{
\begin{CD}
P_{\mathrm{red}} @>>> Y \\
@VVV @VV f V \\
P @>>> U
\end{CD}
\right. ⎩ ⎨ ⎧ P red ↓ ⏐ P Y ↓ ⏐ f U
The map of points
P r e d → Y ↓ ↓ P × S S r e d → Y ↓ ↓ P → U \begin{CD}
P_{\mathrm{red}} @>>> Y \\
@VVV @VVV \\
P \times_S S_{\mathrm{red}} @>>> Y \\
@VVV @VVV \\
P @>>> U
\end{CD} P red ↓ ⏐ P × S S red ↓ ⏐ P Y ↓ ⏐ Y ↓ ⏐ U
corresponding to the induced map of ∞ \infty ∞ -groupoids ( U Y / G ^ ) ( S ) → ( ( U / G ^ ) Y / G ) ( S ) (\widehat{U_Y/G})(S) \to((\widehat{U/G})_{Y/G})(S) ( U Y / G ) ( S ) → (( U / G ) Y / G ) ( S ) is the map of diagrams in which Y → Y Y \to Y Y → Y , U → U U \to U U → U , P → P P \to P P → P are the identity morphisms, and P r e d → P × S S r e d P_{\mathrm{red}} \to P \times_S S_{\mathrm{red}} P red → P × S S red is the universal map to the fiber product. Since P r e d → P × S S r e d P_{\mathrm{red}} \to P \times_S S_{\mathrm{red}} P red → P × S S red is natural in P P P , in order to show that ( U Y / G ^ ) ( S ) → ( ( U / G ^ ) Y / G ) ( S ) (\widehat{U_Y/G})(S) \to((\widehat{U/G})_{Y/G})(S) ( U Y / G ) ( S ) → (( U / G ) Y / G ) ( S ) is an equivalence, it suffices to show that P r e d → P × S S r e d P_{\mathrm{red}} \to P \times_S S_{\mathrm{red}} P red → P × S S red is an equivalence. This is clearly true étale locally on S r e d S_{\mathrm{red}} S red since G G G is smooth (and P P P is a G G G -torsor), and since the property of being reduced is étale local, we are done. □ \square □
Miscellanea
Lemma 3.49. Suppose that X → Y X \to Y X → Y and Z → Y Z \to Y Z → Y are maps of prestacks. Let W : = Z × Y X W := Z \times_Y X W := Z × Y X denote the fiber product. Then Z W ^ ≃ X Y ^ × Y Z \widehat{Z_W} \simeq\widehat{X_Y} \times_Y Z Z W ≃ X Y × Y Z .
Proof. By definition, Z W ^ = Z × Z d R W d R \widehat{Z_W} = Z \times_{Z_{\mathrm{dR}}} W_{\mathrm{dR}} Z W = Z × Z dR W dR . On the other hand, we have a natural equivalence W d R = ( Z × Y X ) d R ≃ Z d R × Y d R X d R W_{\mathrm{dR}} = (Z \times_Y X)_{\mathrm{dR}} \simeq Z_{\mathrm{dR}} \times_{Y_{\mathrm{dR}}} X_{\mathrm{dR}} W dR = ( Z × Y X ) dR ≃ Z dR × Y dR X dR , giving an equivalence,
Z W ^ ≃ Z × Z d R ( Z d R × Y d R X d R ) ≃ Z × Y d R X d R ≃ Z × Y Y × Y d R X d R = Z × Y Y X ^ . \begin{aligned}
\widehat{Z_W} \simeq Z \times_{Z_{\mathrm{dR}}} (Z_{\mathrm{dR}} \times_{Y_{\mathrm{dR}}} X_{\mathrm{dR}}) \\
&\simeq Z \times_{Y_{\mathrm{dR}}} X_{\mathrm{dR}} \\
&\simeq Z \times_Y Y \times_{Y_{\mathrm{dR}}} X_{\mathrm{dR}} = Z \times_Y \widehat{Y_X}.
\end{aligned} Z W ≃ Z × Z dR ( Z dR × Y dR X dR ) ≃ Z × Y dR X dR ≃ Z × Y Y × Y dR X dR = Z × Y Y X .
□ \square □
Lemma 3.50. Let X → Y X \to Y X → Y and Z → W Z \to W Z → W be maps of prestacks, and suppose that we have a map y : Y X ^ → W Z ^ : = W y:\widehat{Y_X} \to\widehat{W_Z} := W y : Y X → W Z := W . Then there is a natural equivalence,
W y ^ ≃ W X ^ . \widehat{W_y} \simeq\widehat{W_X}. W y ≃ W X .
where the formal completion on the right-hand side it taken along the composition of the sequence of maps X → Y ^ X → W ^ X → W X \to\widehat{Y}_X \to\widehat{W}_X \to W X → Y X → W X → W .
Proof. Using the universl property of fiber products, we obtain the following sequence of maps:
W Y : = W × W d R Y d R = W ^ Z × W ^ Z d R Y ^ X d R = ( W × W d R Z d R ) × ( W × W d R Z d R ) d R ( Y × Y d R X d R ) d R ≃ ( W × W d R Z d R ) × W d R × W d R Z d R ( Y d R × Y d R X d R ) ≃ ( W × W d R Z d R ) × Z d R X d R ≃ W × W d R X d R = : W ^ X , \begin{aligned}
\mathcal{W}_Y &:= \mathcal{W} \times_{W_{\mathrm{dR}}} Y_{\mathrm{dR}} \\
&= \widehat{\mathcal{W}}_Z \times_{\widehat{W}_{Z_{\mathrm{dR}}}} \widehat{Y}_{X_{\mathrm{dR}}} \\
&= (\mathcal{W} \times_{W_{\mathrm{dR}}} Z_{\mathrm{dR}}) \times_{(\mathcal{W} \times_{W_{\mathrm{dR}}} Z_{\mathrm{dR}})_{\mathrm{dR}}} (Y \times_{Y_{\mathrm{dR}}} X_{\mathrm{dR}})_{\mathrm{dR}} \\
&\simeq(\mathcal{W} \times_{W_{\mathrm{dR}}} Z_{\mathrm{dR}}) \times_{W_{\mathrm{dR}} \times_{W_{\mathrm{dR}}} Z_{\mathrm{dR}}} (Y_{\mathrm{dR}} \times_{Y_{\mathrm{dR}}} X_{\mathrm{dR}}) \\
&\simeq(\mathcal{W} \times_{W_{\mathrm{dR}}} Z_{\mathrm{dR}}) \times_{Z_{\mathrm{dR}}} X_{\mathrm{dR}} \\
&\simeq\mathcal{W} \times_{W_{\mathrm{dR}}} X_{\mathrm{dR}} \\
&=: \widehat{\mathcal{W}}_X,
\end{aligned} W Y := W × W dR Y dR = W Z × W Z dR Y X dR = ( W × W dR Z dR ) × ( W × W dR Z dR ) dR ( Y × Y dR X dR ) dR ≃ ( W × W dR Z dR ) × W dR × W dR Z dR ( Y dR × Y dR X dR ) ≃ ( W × W dR Z dR ) × Z dR X dR ≃ W × W dR X dR =: W X ,
where the only non-trivial equivalence is the one on the fourth line coming from the fact that ( − ) d R (-)_{\mathrm{dR}} ( − ) dR preserves limits as a right adjoint.
Loop spaces
Lemma 3.51. Let i : Z ↪ X i : Z \hookrightarrow X i : Z ↪ X be a closed substack of a stack X X X , and let K ∈ A n i K \in\mathrm{Ani} K ∈ Ani be nonempty. Then the induced map i M a p : Map ( K B , Z ) → Map ( K B , X ) i_{\mathrm{Map}} : \operatorname{Map}(K_B, Z) \to\operatorname{Map}(K_B, X) i Map : Map ( K B , Z ) → Map ( K B , X ) is a closed immersion, where K B K_B K B denotes the Betti stack of K K K .
Proof. For every K ∈ A n i K \in\mathrm{Ani} K ∈ Ani , we tautologically have colim K ∗ K ≃ K \operatorname{colim}_{K} *_{K} \simeq K colim K ∗ K ≃ K , where the latter indicates the constant diagram ∗ K : K → A n i *_{K} : K \to\mathrm{Ani} ∗ K : K → Ani with value ∗ * ∗ . The functor ( − ) : A n i → P r e S t k (-) : \mathrm{Ani} \to\mathrm{PreStk} ( − ) : Ani → PreStk taking an anima to its constant prestack preserves colimits, as does the sheafification functor L : P r e S t k → S t k L : \mathrm{PreStk} \to\mathrm{Stk} L : PreStk → Stk . Thus, the Betti stack K B : = L ( K ) K_B := L(K) K B := L ( K ) is equivalent to colim K ∗ B \operatorname{colim}_{K} *_{B} colim K ∗ B . For any stack Y Y Y , Map ( − , Y ) \operatorname{Map}(-,Y) Map ( − , Y ) takes colimits in the first argument to limits, so we obtain an equivalence
Map ( K B , Y ) ≃ Map ( colim K ∗ B , Y ) ≃ lim K Map ( ∗ B , Y ) ≃ lim K Y . \operatorname{Map}(K_B,Y) \simeq\operatorname{Map}(\operatorname{colim}_{K} *_{B},Y) \simeq\lim_{K}\operatorname{Map}(*_{B},Y) \simeq\lim_{K}Y. Map ( K B , Y ) ≃ Map ( colim K ∗ B , Y ) ≃ K lim Map ( ∗ B , Y ) ≃ K lim Y .
where the latter object, explicitly, is the limit of the diagram
D Y : K ≃ K o p → ∗ K o p A n i o p → ( − ) B o p S t k o p → Map ( − , Y ) S t k . D_Y : K \simeq K^{\mathrm{op}} \xrightarrow{*_{K}^{\mathrm{op}}} \mathrm{Ani}^{\mathrm{op}} \xrightarrow{(-)_{B}^{\mathrm{op}}} \mathrm{Stk}^{\mathrm{op}} \xrightarrow{\operatorname{Map}(-,Y)} \mathrm{Stk}. D Y : K ≃ K op ∗ K op Ani op ( − ) B op Stk op Map ( − , Y ) Stk .
Moreover, this equivalence is natural in Y Y Y , so we obtain a commutative diagram,
Map ( K B , Z ) → ≃ lim K D Z i Map ↓ ↓ ι Map ( K B , X ) → ≃ lim K D X . \begin{CD}
\operatorname{Map}(K_B,Z) @>{\simeq}>> \lim_{K}D_Z \\
@V{i_{\operatorname{Map}}}VV @VV{\iota}V \\
\operatorname{Map}(K_B,X) @>{\simeq}>> \lim_{K}D_X.
\end{CD} Map ( K B , Z ) i Map ↓ ⏐ Map ( K B , X ) ≃ ≃ K lim D Z ↓ ⏐ ι K lim D X .
where ι \iota ι denotes the map of limits induced by the natural transformation D Z → D X D_Z \to D_X D Z → D X induced by i : Z ↪ X i : Z \hookrightarrow X i : Z ↪ X . It suffices, therefore, to show that ι \iota ι is a closed immersion.
Given a diagram in Fun ( K , P r e S t k ) / D Y ≃ Fun ( K , P r e S t k / Y ) \operatorname{Fun}(K,\mathrm{PreStk})/D_Y \simeq\operatorname{Fun}(K,\mathrm{PreStk}/Y) Fun ( K , PreStk ) / D Y ≃ Fun ( K , PreStk / Y ) , there are two limit operations one may perform. The first is to take the limit within the overcategory P r e S t k / Y \mathrm{PreStk}/Y PreStk / Y to obtain a functor,
lim K / Y : Fun ( K , P r e S t k ) / D Y → P r e S t k / Y , \lim_{K}^{/Y} : \operatorname{Fun}(K,\mathrm{PreStk})/D_Y \to\mathrm{PreStk}/Y, K lim / Y : Fun ( K , PreStk ) / D Y → PreStk / Y ,
which corresponds to the right adjoint of the constant diagram functor D ( − ) : P r e S t k / Y → Fun ( K , P r e S t k Y ) ≃ Fun ( K , P r e S t k ) / D Y D_{(-)} : \mathrm{PreStk}/Y \to\operatorname{Fun}(K,\mathrm{PreStk}_Y) \simeq\operatorname{Fun}(K,\mathrm{PreStk})/D_Y D ( − ) : PreStk / Y → Fun ( K , PreStk Y ) ≃ Fun ( K , PreStk ) / D Y . In particular, if Y = ∗ Y = * Y = ∗ , we get the functor lim K : Fun ( K , P r e S t k ) → P r e S t k \lim_{K} : \operatorname{Fun}(K,\mathrm{PreStk}) \to\mathrm{PreStk} lim K : Fun ( K , PreStk ) → PreStk . The functor lim K \lim_{K} lim K in turn induces a functor on overcategories,
lim K : Fun ( K , P r e S t k ) / D Y → P r e S t k / lim K D Y , \lim_{K} : \operatorname{Fun}(K,\mathrm{PreStk})/D_Y \to\mathrm{PreStk}/\lim_{K}D_Y, K lim : Fun ( K , PreStk ) / D Y → PreStk / K lim D Y ,
which furnishes the second limit operation.
Now let S S S be an affine derived scheme, and let f : S → lim K D X f:S \to\lim_{K} DX f : S → lim K D X be an arbitrary map. In order to show that ι \iota ι is a closed immersion, we must, by definition ([CW23, §3.7]), show that the map of 0 0 0 -truncated stacks,
τ ≤ 0 ( lim K Z × lim K X S ) → τ ≤ 0 S \tau_{\leq0}\left(\lim_{K} Z \mathbin{\times_{\lim_{K} X}} S\right) \to\tau_{\leq0}S τ ≤ 0 ( K lim Z × l i m K X S ) → τ ≤ 0 S
is a closed immersion. For any T ∈ PreStk / lim K D X \mathcal{T} \in\operatorname{PreStk}_{/\lim_{K} DX} T ∈ PreStk / l i m K D X , and any S ∈ PreStk / S \mathcal{S} \in\operatorname{PreStk}_{/S} S ∈ PreStk / S , we have the natural equivalence,
Map PreStk / S ( S , T × lim K D X S ) ≃ Map PreStk ( S , T ) × Map PreStk ( S , lim K D X ) Map PreStk / S ( S , S ) , (11) \operatorname{Map}_{\operatorname{PreStk}_{/S}}\left(\mathcal{S},\mathcal{T} \mathbin{\times_{\lim_{K} DX}} S\right) \simeq\operatorname{Map}_{\operatorname{PreStk}}(\mathcal{S},\mathcal{T}) \mathbin{\times_{\operatorname{Map}_{\operatorname{PreStk}}(\mathcal{S},\lim_{K} DX)}} \operatorname{Map}_{\operatorname{PreStk}_{/S}}(\mathcal{S},S), \tag*{(11)} Map PreStk / S ( S , T × l i m K D X S ) ≃ Map PreStk ( S , T ) × Map PreStk ( S , l i m K D X ) Map PreStk / S ( S , S ) , (11)
coming from the fact that the overcategory PreStk / S \operatorname{PreStk}_{/S} PreStk / S is equivalent to the oriented product PreStk × PreStk × ⃗ PreStk { S } \operatorname{PreStk}\mathbin{\times_{\operatorname{PreStk}}\mkern-6mu\vec{\mkern6mu\times}\operatorname{PreStk}\{S\}} PreStk × PreStk × PreStk { S } ([Lur26, Tag 02VT]), and morphisms in the latter are precisely what appears above. Now let T = lim K D Z \mathcal{T}=\lim_{K} DZ T = lim K D Z , where the structure map lim K D Z → lim K D X \lim_{K} DZ \to\lim_{K} DX lim K D Z → lim K D X comes from the map D Z → D X DZ \to DX D Z → D X induced by i : Z ↪ X i:Z \hookrightarrow X i : Z ↪ X . Then the right-hand side of (11) becomes
Map ( S , lim K D Z ) × Map ( S , lim K D X ) Map / S ( S , S ) ≃ Map ( S , lim K D Z ) × Map ( S , lim K D X ) ∗ ≃ lim K Map ( S , D Z ) × lim K Map ( S , D X ) lim K Δ ∗ ≃ lim K ( Map ( S , D Z ) × Map ( S , D X ) Δ ∗ ) ≃ lim K ( Map / D S ( S , D Z × D X D S ) ) ≃ Map ( S , lim K / X ( D Z × D X D S ) ) . \begin{aligned}
\operatorname{Map}\left(\mathcal{S},\lim_{K} DZ\right) \mathbin{\times_{\operatorname{Map}(\mathcal{S},\lim_{K} DX)}} \operatorname{Map}_{/S}(\mathcal{S},S)
&\simeq\operatorname{Map}\left(\mathcal{S},\lim_{K} DZ\right) \mathbin{\times_{\operatorname{Map}(\mathcal{S},\lim_{K} DX)}} * \\
&\simeq\lim_{K}\operatorname{Map}(\mathcal{S},DZ) \mathbin{\times_{\lim_{K}\operatorname{Map}(\mathcal{S},DX)}} \lim_{K}\Delta_* \\
&\simeq\lim_{K}\left(\operatorname{Map}(\mathcal{S},DZ) \mathbin{\times_{\operatorname{Map}(\mathcal{S},DX)}} \Delta_*\right) \\
&\simeq\lim_{K}\left(\operatorname{Map}_{/D_S}(\mathcal{S},DZ \mathbin{\times_{DX}} D_S)\right) \\
&\simeq\operatorname{Map}\left(\mathcal{S},\lim_{K}^{/X}(DZ \mathbin{\times_{DX}} D_S)\right).
\end{aligned} Map ( S , K lim D Z ) × Map ( S , l i m K D X ) Map / S ( S , S ) ≃ Map ( S , K lim D Z ) × Map ( S , l i m K D X ) ∗ ≃ K lim Map ( S , D Z ) × l i m K Map ( S , D X ) K lim Δ ∗ ≃ K lim ( Map ( S , D Z ) × Map ( S , D X ) Δ ∗ ) ≃ K lim ( Map / D S ( S , D Z × D X D S ) ) ≃ Map ( S , K lim / X ( D Z × D X D S ) ) .
where Δ ∗ \Delta_* Δ ∗ denotes the constant K K K -shaped diagram K → Ani K \to\operatorname{Ani} K → Ani with value ∗ * ∗ . Here, the first equivalence uses that Map PreStk / S ( S , S ) ≃ { S → S } ≃ ∗ \operatorname{Map}_{\operatorname{PreStk}_{/S}}(\mathcal{S},S)\simeq\{S\to S\}\simeq* Map PreStk / S ( S , S ) ≃ { S → S } ≃ ∗ ; the second equivalence uses that Map ( S , − ) \operatorname{Map}(\mathcal{S},-) Map ( S , − ) commutes with limits and ∗ ≃ lim K Δ ∗ *\simeq\lim_{K}\Delta_* ∗ ≃ lim K Δ ∗ ; the third equivalence use that limits commute with one another; the fourth equivalence uses the definition of mapping spaces in the overcategory; and the final equivalence uses that limits commute with Map / D X ( S , − ) \operatorname{Map}_{/DX}(\mathcal{S},-) Map / D X ( S , − ) once again. Since S ∈ PreStk / S \mathcal{S}\in\operatorname{PreStk}_{/S} S ∈ PreStk / S was arbitrary, this shows that,
lim K / S ( D Z × D X D S ) → ∼ lim K D Z × lim K D X S . \lim_{K}^{/S}(DZ \mathbin{\times_{DX}} D_S) \xrightarrow{\sim} \lim_{K} DZ \mathbin{\times_{\lim_{K} DX}} S. K lim / S ( D Z × D X D S ) ∼ K lim D Z × l i m K D X S .
Thus, we are reduced to showing that the canonical map τ ≤ 0 ( lim K / S ( D Z × D X D S ) ) → τ ≤ 0 S \tau_{\leq0}\left(\lim_{K}^{/S}(DZ \mathbin{\times_{DX}} D_S)\right)\to\tau_{\leq0}S τ ≤ 0 ( lim K / S ( D Z × D X D S ) ) → τ ≤ 0 S is the image of a closed immersion of ordinary affine schemes under the inclusion i ≤ 0 : PreStk ≤ 0 → PreStk i_{\leq0}:\operatorname{PreStk}_{\leq0}\to\operatorname{PreStk} i ≤ 0 : PreStk ≤ 0 → PreStk . Since r ≤ 0 ∘ i ≤ 0 ≃ id PreStk ≤ 0 r_{\leq0}\circ i_{\leq0}\simeq\operatorname{id}_{\operatorname{PreStk}_{\leq0}} r ≤ 0 ∘ i ≤ 0 ≃ id PreStk ≤ 0 , 30 it therefore suffices to show that r ≤ 0 ( lim K / S ( D Z × D X D S ) ) → r ≤ 0 S r_{\leq0}\left(\lim_{K}^{/S}(DZ \mathbin{\times_{DX}} D_S)\right)\to r_{\leq0}S r ≤ 0 ( lim K / S ( D Z × D X D S ) ) → r ≤ 0 S is a closed immersion of ordinary affine schemes. For each point k ∈ K k\in K k ∈ K , we note that r ≤ 0 ( D Z ( k ) × D X ( k ) D S ( k ) ) = r ≤ 0 ( Z × X S ) → r ≤ 0 S r_{\leq0}(DZ(k)\mathbin{\times_{DX(k)}}D_S(k))=r_{\leq0}(Z\mathbin{\times_X}S)\to r_{\leq0}S r ≤ 0 ( D Z ( k ) × D X ( k ) D S ( k )) = r ≤ 0 ( Z × X S ) → r ≤ 0 S is a closed immersion because Z ↪ X Z\hookrightarrow X Z ↪ X is a closed immersion. In particular, since r ≤ 0 S ≃ Spec π 0 O ( S ) r_{\leq0}S\simeq\operatorname{Spec}\pi_0\mathcal{O}(S) r ≤ 0 S ≃ Spec π 0 O ( S ) is a classical affine scheme, this implies that r ≤ 0 ( D Z ( k ) × D X ( k ) D S ( k ) ) r_{\leq0}(DZ(k)\mathbin{\times_{DX(k)}}D_S(k)) r ≤ 0 ( D Z ( k ) × D X ( k ) D S ( k )) is an affine classical scheme, and the induced map π 0 O ( S ) → π 0 O ( D Z ( k ) × D X ( k ) D S ( k ) ) \pi_0\mathcal{O}(S)\to\pi_0\mathcal{O}(DZ(k)\mathbin{\times_{DX(k)}}D_S(k)) π 0 O ( S ) → π 0 O ( D Z ( k ) × D X ( k ) D S ( k )) is a surjective map of commutative rings. On the other hand, since r ≤ 0 r_{\leq0} r ≤ 0 is a right adjoint (to the inclusion i ≤ 0 i_{\leq0} i ≤ 0 ), it preserves limits. In particular, we have
r ≤ 0 ( lim K / S ( D Z × D X D S ) ) ≃ lim K / ( r ≤ 0 S ) r ≤ 0 ( D Z × D X D S ) . r_{\leq0}\left(\lim_{K}^{/S}(DZ\mathbin{\times_{DX}}D_S)\right)\simeq\lim_{K}^{/(r_{\leq0}S)}r_{\leq0}(DZ\mathbin{\times_{DX}}D_S). r ≤ 0 ( K lim / S ( D Z × D X D S ) ) ≃ K lim / ( r ≤ 0 S ) r ≤ 0 ( D Z × D X D S ) .
The latter is a limit of classical affine schemes. Since r ≤ 0 S r_{\leq0}S r ≤ 0 S is affine, the canonical map lim K / S r ≤ 0 ( D Z × D X D S ) → r ≤ 0 S \lim_{K}^{/S}r_{\leq0}(DZ\mathbin{\times_{DX}}D_S)\to r_{\leq0}S lim K / S r ≤ 0 ( D Z × D X D S ) → r ≤ 0 S is determined by its induced map on global sections,
π 0 O ( S ) → O ( lim K / S r ≤ 0 ( D Z × D X D S ) ) ≃ colim K π 0 O ( S ) / O ( r ≤ 0 ( D Z ( k ) × D X ( k ) D S ( k ) ) ) . (12) \pi_0\mathcal{O}(S)\to\mathcal{O}\left(\lim_{K}^{/S}r_{\leq0}(DZ\mathbin{\times_{DX}}D_S)\right)\simeq\operatorname*{colim}_{K}^{\pi_0\mathcal{O}(S)/\mathcal{O}}\left(r_{\leq0}(DZ(k)\mathbin{\times_{DX(k)}}D_S(k))\right). \tag*{(12)} π 0 O ( S ) → O ( K lim / S r ≤ 0 ( D Z × D X D S ) ) ≃ K colim π 0 O ( S ) / O ( r ≤ 0 ( D Z ( k ) × D X ( k ) D S ( k )) ) . (12)
where the equivalence comes from the fact that lim K / S r ≤ 0 ( D Z × D X D S ) \lim_{K}/S_{r\leq0}(D_Z \times_{D_X} D_S) lim K / S r ≤ 0 ( D Z × D X D S ) is a limit of affine schemes. If we can show that this induced map of global sections is surjective, then we are done. Since every morphism in K K K is invertible, however, all the structure morphisms of the diagram O ( r ≤ 0 ( D Z ( k ) × D X ( k ) D S ( k ) ) ) \mathcal{O}\left(r_{\leq0}\left(D_Z(k) \times_{D_X(k)} D_S(k)\right)\right) O ( r ≤ 0 ( D Z ( k ) × D X ( k ) D S ( k ) ) ) are isomorphisms. Thus, it is obvious that the map (12) is surjective, as desired. □
In light of the above, it makes sense to formulate the following lemma.
Lemma 3.54. Let i : Z ↪ X i : Z \hookrightarrow X i : Z ↪ X be a closed substack of a stack X X X with open complement U U U , and let K ∈ A n i K \in\mathrm{Ani} K ∈ Ani be a nonempty connected anima, whose Betti stack is denoted K B K_B K B . Then Map ( K B , U ) \operatorname{Map}(K_B,U) Map ( K B , U ) is the open complement to Map ( K B , Z ) \operatorname{Map}(K_B,Z) Map ( K B , Z ) in Map ( K B , X ) \operatorname{Map}(K_B,X) Map ( K B , X ) .
Proof. Let S S S be an affine derived scheme, and consider a point f : S → Map ( K B , X ) f : S \to\operatorname{Map}(K_B,X) f : S → Map ( K B , X ) . By definition, f f f factors through the complement Map ( K B , X ) ∖ Map ( K B , Z ) \operatorname{Map}(K_B,X) \setminus\operatorname{Map}(K_B,Z) Map ( K B , X ) ∖ Map ( K B , Z ) if and only if S × Map ( K B , X ) Map ( K B , Z ) ≃ ∅ S \times_{\operatorname{Map}(K_B,X)} \operatorname{Map}(K_B,Z) \simeq\varnothing S × Map ( K B , X ) Map ( K B , Z ) ≃ ∅ . On the other hand, f f f factors through Map ( K B , U ) \operatorname{Map}(K_B,U) Map ( K B , U ) if and only if, for the map f ~ : S × K B → X \widetilde{f} : S \times K_B \to X f : S × K B → X corresponding to f f f under adjunction, ( S × K B ) × X Z ≃ ∅ (S \times K_B) \times_X Z \simeq\varnothing ( S × K B ) × X Z ≃ ∅ . It thus suffices to show that
S × Map ( K B , X ) Map ( K B , Z ) ≃ ∅ ⟺ ( S × K B ) × X Z ≃ ∅ . S \times_{\operatorname{Map}(K_B,X)} \operatorname{Map}(K_B,Z) \simeq\varnothing\quad\Longleftrightarrow\quad(S \times K_B) \times_X Z \simeq\varnothing. S × Map ( K B , X ) Map ( K B , Z ) ≃ ∅ ⟺ ( S × K B ) × X Z ≃ ∅ .
Suppose ( S × K B ) × X Z ≃ ∅ (S \times K_B) \times_X Z \simeq\varnothing ( S × K B ) × X Z ≃ ∅ . For any T T T , we have the following chain of equivalences:
Map ( T , S × Map ( K B , X ) Map ( K B , Z ) ) ≃ Map ( T , S ) × Map ( T , Map ( K B , X ) ) Map ( T , Map ( K B , Z ) ) ≃ Map ( T , S ) × Map ( T , Map ( K B , S × K B ) ) Map ( T , Map ( K B , S × K B ) ) × Map ( T , Map ( K B , X ) ) Map ( T , Map ( K B , Z ) ) ≃ Map ( T , S ) × Map ( T × K B , S × K B ) Map ( T × K B , ( S × K B ) × X Z ) ≃ ∅ , \begin{aligned}
\operatorname{Map}\left(T,S \times_{\operatorname{Map}(K_B,X)} \operatorname{Map}(K_B,Z)\right)
&\simeq\operatorname{Map}(T,S) \times_{\operatorname{Map}(T,\operatorname{Map}(K_B,X))} \operatorname{Map}(T,\operatorname{Map}(K_B,Z)) \\
&\simeq\operatorname{Map}(T,S) \times_{\operatorname{Map}(T,\operatorname{Map}(K_B,S \times K_B))} \operatorname{Map}(T,\operatorname{Map}(K_B,S \times K_B)) \times_{\operatorname{Map}(T,\operatorname{Map}(K_B,X))} \operatorname{Map}(T,\operatorname{Map}(K_B,Z)) \\
&\simeq\operatorname{Map}(T,S) \times_{\operatorname{Map}(T \times K_B,S \times K_B)} \operatorname{Map}\left(T \times K_B,(S \times K_B) \times_X Z\right) \\
&\simeq\varnothing,
\end{aligned} Map ( T , S × Map ( K B , X ) Map ( K B , Z ) ) ≃ Map ( T , S ) × Map ( T , Map ( K B , X )) Map ( T , Map ( K B , Z )) ≃ Map ( T , S ) × Map ( T , Map ( K B , S × K B )) Map ( T , Map ( K B , S × K B )) × Map ( T , Map ( K B , X )) Map ( T , Map ( K B , Z )) ≃ Map ( T , S ) × Map ( T × K B , S × K B ) Map ( T × K B , ( S × K B ) × X Z ) ≃ ∅ ,
where in the second line the map Map ( T , S ) → Map ( T , Map ( K B , S × K B ) ) \operatorname{Map}(T,S) \to\operatorname{Map}(T,\operatorname{Map}(K_B,S \times K_B)) Map ( T , S ) → Map ( T , Map ( K B , S × K B )) is given by postcomposition with the map S → Map ( K B , S × K B ) S \to\operatorname{Map}(K_B,S \times K_B) S → Map ( K B , S × K B ) corresponding to the identity id S × K B \operatorname{id}_{S \times K_B} id S × K B under adjunction, and the map Map ( T , Map ( K B , S × K B ) ) → Map ( T , Map ( K B , X ) ) \operatorname{Map}(T,\operatorname{Map}(K_B,S \times K_B)) \to\operatorname{Map}(T,\operatorname{Map}(K_B,X)) Map ( T , Map ( K B , S × K B )) → Map ( T , Map ( K B , X )) is given by postcomposition with the map Map ( K B , S × K B ) → Map ( K B , X ) \operatorname{Map}(K_B,S \times K_B) \to\operatorname{Map}(K_B,X) Map ( K B , S × K B ) → Map ( K B , X ) given itself by postcomposition with the map S × K B → X S \times K_B \to X S × K B → X corresponding under adjunction to the map S → Map ( K B , X ) S \to\operatorname{Map}(K_B,X) S → Map ( K B , X ) coming from the first coordinate of the chosen T T T -point in S × Map ( K B , X ) Map ( K B , Z ) S \times_{\operatorname{Map}(K_B,X)} \operatorname{Map}(K_B,Z) S × Map ( K B , X ) Map ( K B , Z ) .
It remains to show the converse. Suppose that ( S × K B ) × X Z ≄ ∅ (S \times K_B) \times_X Z \not\simeq\varnothing ( S × K B ) × X Z ≃ ∅ . Since emptiness is witnessed by field-valued points, there exists T = Spec κ T = \operatorname{Spec}\kappa T = Spec κ for some field κ \kappa κ such that Map ( T , ( S × K B ) × X Z ) \operatorname{Map}(T,(S \times K_B) \times_X Z) Map ( T , ( S × K B ) × X Z ) is nonempty. Choose a point in Map ( T , ( S × K B ) × X Z ) \operatorname{Map}(T,(S \times K_B) \times_X Z) Map ( T , ( S × K B ) × X Z ) . The data of such a point is, by the universal property, the data of maps s ~ : T → S × K B \widetilde{s} : T \to S \times K_B s : T → S × K B , and z : T → Z z : T \to Z z : T → Z together with a homotopy of the compositions,
( T → s ~ S × K B → f ~ X ) ⇒ η ( T → z Z → i X ) . \left(T \xrightarrow{\widetilde{s}} S \times K_B \xrightarrow{\widetilde{f}} X\right) \xRightarrow{\eta} \left(T \xrightarrow{z} Z \xrightarrow{i} X\right). ( T s S × K B f X ) η ( T z Z i X ) .
The task at hand is to produce a T T T -point of S × Map ( K B , X ) Map ( K B , Z ) S \times_{\operatorname{Map}(K_B,X)} \operatorname{Map}(K_B,Z) S × Map ( K B , X ) Map ( K B , Z ) from this data.
Let s : T → S s : T \to S s : T → S denote the map obtained by the composition T → s ~ S × K B → pr 1 S T \xrightarrow{\widetilde{s}} S \times K_B \xrightarrow{\operatorname{pr}_1} S T s S × K B pr 1 S , and k : T → K B k : T \to K_B k : T → K B the map obtained by T → s ~ S × K B → pr 2 K B T \xrightarrow{\widetilde{s}} S \times K_B \xrightarrow{\operatorname{pr}_2} K_B T s S × K B pr 2 K B . We seek a point ζ : T → Map ( K B , Z ) \zeta: T \to\operatorname{Map}(K_B,Z) ζ : T → Map ( K B , Z ) together with a homotopy H : f ∘ s ⇒ i ∗ ∘ ζ H : f \circ s \Rightarrow i_* \circ\zeta H : f ∘ s ⇒ i ∗ ∘ ζ , where i ∗ : Map ( K B , Z ) → Map ( K B , X ) i_* : \operatorname{Map}(K_B,Z) \to\operatorname{Map}(K_B,X) i ∗ : Map ( K B , Z ) → Map ( K B , X ) denotes the map given by composition with i i i . There is a canonical map,
Map ( T , Map ( K B , X ) ) → Map S p c ( K B ( T ) , X ( T ) ) , \operatorname{Map}(T,\operatorname{Map}(K_B,X)) \to\operatorname{Map}_{\mathrm{Spc}}(K_B(T),X(T)), Map ( T , Map ( K B , X )) → Map Spc ( K B ( T ) , X ( T )) ,
under which the map f ∘ s : T → Map ( K B , X ) f \circ s : T \to\operatorname{Map}(K_B,X) f ∘ s : T → Map ( K B , X ) is sent to a map f ~ s : K B ( T ) → X ( T ) \widetilde{f}_s : K_B(T) \to X(T) f s : K B ( T ) → X ( T ) . The homotopy η \eta η tells us precisely that the image of k ∈ K B ( T ) k \in K_B(T) k ∈ K B ( T ) lies in the image of the map z ( T ) → X ( T ) z(T) \to X(T) z ( T ) → X ( T ) . Consider the Cartesian diagram
K B ( T ) × X ( T ) Z ( T ) → p K K B ( T ) p Z ↓ ↓ f ~ s Z ( T ) → i T X ( T ) \begin{CD}
K_B(T) \times_{X(T)} Z(T) @>{p_K}>> K_B(T) \\
@V{p_Z}VV @VV{\widetilde f_s}V \\
Z(T) @>{i_T}>> X(T)
\end{CD} K B ( T ) × X ( T ) Z ( T ) p Z ↓ ⏐ Z ( T ) p K i T K B ( T ) ↓ ⏐ f s X ( T )
We claim that p K p_K p K is an equivalence. Indeed, it has one nonempty fiber by assumption, over k ∈ K B ( T ) k \in K_B(T) k ∈ K B ( T ) . Given another fiber, over k ′ ∈ K B ( T ) k' \in K_B(T) k ′ ∈ K B ( T ) , there is a path k → k ′ k \to k' k → k ′ . This path defines a map of fibers by concatenation: A point in the fiber over k k k is a point a ∈ K B ( T ) × X ( T ) Z ( T ) a \in K_B(T) \times_{X(T)} Z(T) a ∈ K B ( T ) × X ( T ) Z ( T ) together with a path p K ( a ) → k p_K(a) \to k p K ( a ) → k . Thus, the fiber over k ′ k' k ′ cannot be empty. Since T T T is classical, i T : Z ( T ) ↪ X ( T ) i_T : Z(T) \hookrightarrow X(T) i T : Z ( T ) ↪ X ( T ) is a monomorphism of spaces,31 hence its base change p K p_K p K is a monomorphism, so its fibers are all nonempty and contractible. Thus, p K p_K p K is an equivalence. Thus, we obtain a map ζ ~ : = p Z ∘ ( p K ) − 1 : K B ( T ) → Z ( T ) \widetilde{\zeta} := p_Z \circ(p_K)^{-1} : K_B(T) \to Z(T) ζ := p Z ∘ ( p K ) − 1 : K B ( T ) → Z ( T ) . Since T = Spec κ T = \operatorname{Spec}\kappa T = Spec κ , it is a connected affine scheme, so K B ( T ) ≃ K K_B(T) \simeq K K B ( T ) ≃ K . Because Stk κ \operatorname{Stk}_{\kappa} Stk κ is an ∞ \infty ∞ -topos, it is cotensored over Spc \operatorname{Spc} Spc ([HTT, Remark 5.5.2.6]), meaning there is a natural equivalence,
Map ( T , Map ( K B , Z ) ) ≃ Map ( K , Z ( T ) ) . (13) \operatorname{Map}(T,\operatorname{Map}(K_B,Z)) \simeq\operatorname{Map}(K,Z(T)). \tag*{(13)} Map ( T , Map ( K B , Z )) ≃ Map ( K , Z ( T )) . (13)
Let ζ : T → Map ( K B , Z ) \zeta: T \to\operatorname{Map}(K_B,Z) ζ : T → Map ( K B , Z ) denote the point that corresponds to ζ ~ \widetilde{\zeta} ζ under this equivalence.
It remains to specify a homotopy H : f ∘ s ⟹ i ∗ ∘ ζ H : f \circ s \Longrightarrow i_* \circ\zeta H : f ∘ s ⟹ i ∗ ∘ ζ . We have an explicit homotopy
H ~ : f ~ s ≃ f ~ s ∘ p K ∘ ( p K ) − 1 ⇒ σ ∘ ( p K ) − 1 i T ∘ p Z ∘ ( p K ) − 1 ≃ i T ∘ ζ ~ , \widetilde{H} : \widetilde{f}_s \simeq\widetilde{f}_s \circ p_K \circ(p_K)^{-1} \xRightarrow{\sigma\circ(p_K)^{-1}} i_T \circ p_Z \circ(p_K)^{-1} \simeq i_T \circ\widetilde{\zeta}, H : f s ≃ f s ∘ p K ∘ ( p K ) − 1 σ ∘ ( p K ) − 1 i T ∘ p Z ∘ ( p K ) − 1 ≃ i T ∘ ζ ,
of points in Map ( K , X ( T ) ) \operatorname{Map}(K,X(T)) Map ( K , X ( T )) . Under the natural equivalence (13) , H ~ \widetilde{H} H gives a path H : f ∘ s ⟹ i ∗ ∘ ζ H : f \circ s \Longrightarrow i_* \circ\zeta H : f ∘ s ⟹ i ∗ ∘ ζ in Map ( T , Map ( K B , X ) ) \operatorname{Map}(T,\operatorname{Map}(K_B,X)) Map ( T , Map ( K B , X )) . Thus, the tuple ( s , ζ , H ) (s,\zeta,H) ( s , ζ , H ) specifies T T T -point of S × Map ( K B , X ) Map ( K B , Z ) S \times_{\operatorname{Map}(K_B,X)} \operatorname{Map}(K_B,Z) S × Map ( K B , X ) Map ( K B , Z ) , so we are done. □
In particular, taking K = S 1 K = S^1 K = S 1 in Lemma 3.51 and Lemma 3.54, we obtain the following results about the loop stack of a complementary substack.
Corollary 3.56. Let X X X be a derived stack, and let i : Z ↪ X i : Z \hookrightarrow X i : Z ↪ X be a closed immersion. Let U ↪ X U \hookrightarrow X U ↪ X denote its complement. Then L Z ↪ L X LZ \hookrightarrow LX L Z ↪ L X is a closed immersion, and L U ↪ L X LU \hookrightarrow LX LU ↪ L X is its complementary open substack.
It will be important later on to identify the loop stack of a formal completion with the formal completion along the corresponding induced map of loop stacks.
Lemma 3.57. Let Y → X Y \to X Y → X be any map of derived stacks. Then there is an S 1 S^1 S 1 -equivariant isomorphism,
L ( X Y ) ^ ≃ L X Y ^ . \widehat{\mathcal{L}(X_Y)} \simeq\widehat{\mathcal{L}X_{Y}}. L ( X Y ) ≃ L X Y . Proof. By definition, X Y ^ : = X × X d R Y d R \widehat{X_Y} := X \times_{X_{\mathrm{dR}}} Y_{\mathrm{dR}} X Y := X × X dR Y dR . As a representable functor, L ( − ) \mathcal{L}(-) L ( − ) commutes with limits, so the natural morphism L ( X Y ^ ) → ≃ L X × L ( X d R ) L ( Y d R ) \mathcal{L}(\widehat{X_Y}) \xrightarrow{\simeq} \mathcal{L}X \times_{\mathcal{L}(X_{\mathrm{dR}})} \mathcal{L}(Y_{\mathrm{dR}}) L ( X Y ) ≃ L X × L ( X dR ) L ( Y dR ) is an isomorphism. Thus, the lemma follows if we show that the de Rham stack and loop space constructions commute, i.e. if we show the natural map ( L Y ) d R → L ( Y d R ) (\mathcal{L}Y)_{\mathrm{dR}} \to\mathcal{L}(Y_{\mathrm{dR}}) ( L Y ) dR → L ( Y dR ) is an isomorphism for any stack Y Y Y . This follows immediately from the fact that ( − ) d R (-)_{\mathrm{dR}} ( − ) dR preserves limits ([GR14, Lemma 1.1.4]), and the loop stack can be written as L Y ≃ Y × Y × Y Y \mathcal{L}Y \simeq Y \times_{Y \times Y} Y L Y ≃ Y × Y × Y Y . It is clear that all maps admit canonical S 1 S^1 S 1 -equivariant structures. □
Lemma 3.58. Let Y → X Y \to X Y → X be any map of derived stacks. Then the natural S 1 S^{1} S 1 -equivariant map
L u ( X ^ Y ) → L u X ^ L u Y \mathcal{L}^{u}\left(\widehat{X}_{Y}\right) \to\widehat{\mathcal{L}^{u}X}_{\mathcal{L}^{u}Y} L u ( X Y ) → L u X L u Y is an isomorphism.
Proof. Let L = ( − ) r e d L=(-)_{\mathrm{red}} L = ( − ) red be the left adjoint to R = ( − ) d R R=(-)_{\mathrm{dR}} R = ( − ) dR , and let F = B G a × − F=B\mathbb{G}_{a}\times- F = B G a × − be the left adjoint to L u = Map ( B G a , − ) . \mathcal{L}^{u}=\operatorname{Map}(B\mathbb{G}_{a},-). L u = Map ( B G a , − ) . We claim that there is a canonical equivalence
L ∘ F ≃ F ∘ L . L\circ F \simeq F\circ L. L ∘ F ≃ F ∘ L . This suffices to prove the claim. Indeed, passing to right adjoints gives an equivalence
L u ∘ R ≃ R ∘ L u . \mathcal{L}^{u}\circ R \simeq R\circ\mathcal{L}^{u}. L u ∘ R ≃ R ∘ L u . Since L u \mathcal{L}^{u} L u preserves limits, we have
L u ( X ^ Y ) ≃ L u X × L u ( X d R ) L u ( Y d R ) \mathcal{L}^{u}\left(\widehat{X}_{Y}\right) \simeq\mathcal{L}^{u}X \times_{\mathcal{L}^{u}(X_{\mathrm{dR}})} \mathcal{L}^{u}(Y_{\mathrm{dR}}) L u ( X Y ) ≃ L u X × L u ( X dR ) L u ( Y dR ) ≃ L u X × ( L u X ) d R ( L u Y ) d R \simeq\mathcal{L}^{u}X \times_{(\mathcal{L}^{u}X)_{\mathrm{dR}}}(\mathcal{L}^{u}Y)_{\mathrm{dR}} ≃ L u X × ( L u X ) dR ( L u Y ) dR ≃ L u X ^ L u Y . \simeq\widehat{\mathcal{L}^{u}X}_{\mathcal{L}^{u}Y}. ≃ L u X L u Y . Moreover, this composition of equivalences is precisely the natural map
L u ( X ^ Y ) → L u X ^ L u Y . \mathcal{L}^{u}\left(\widehat{X}_{Y}\right) \to\widehat{\mathcal{L}^{u}X}_{\mathcal{L}^{u}Y}. L u ( X Y ) → L u X L u Y . We will now prove the equivalence L ∘ F ≃ F ∘ L L\circ F \simeq F\circ L L ∘ F ≃ F ∘ L . To see this, observe first that reduction commutes with products with A n \mathbb{A}^{n} A n . Indeed, for all n n n , the functors
Z ↦ L ( Z × A n ) and Z ↦ L ( Z ) × A n Z \mapsto L(Z\times\mathbb{A}^{n}) \qquad\text{and} \qquad Z \mapsto L(Z)\times\mathbb{A}^{n} Z ↦ L ( Z × A n ) and Z ↦ L ( Z ) × A n preserve colimits. Thus, the canonical comparison map between them is an isomorphism if and only if it is an isomorphism on affine schemes. There, it is simply the identity
( H 0 ( A ) [ x 1 , … , x n ] ) r e d = H 0 ( A ) r e d [ x 1 , … , x n ] . \left(H^{0}(A)[x_{1},\ldots,x_{n}]\right)_{\mathrm{red}}=H^{0}(A)_{\mathrm{red}}[x_{1},\ldots,x_{n}]. ( H 0 ( A ) [ x 1 , … , x n ] ) red = H 0 ( A ) red [ x 1 , … , x n ] . Now, apply this in each degree to the bar presentation of B G a B\mathbb{G}_{a} B G a :
L ( Z × B G a ) ≃ colim [ n ] ∈ Δ o p L ( Z × G a n ) L(Z\times B\mathbb{G}_{a}) \simeq\underset{[n]\in\Delta^{\mathrm{op}}}{\operatorname{colim}}\,L(Z\times\mathbb{G}_{a}^{n}) L ( Z × B G a ) ≃ [ n ] ∈ Δ op colim L ( Z × G a n ) ≃ colim [ n ] ∈ Δ o p L ( Z ) × G a n \simeq\underset{[n]\in\Delta^{\mathrm{op}}}{\operatorname{colim}}\,L(Z)\times\mathbb{G}_{a}^{n} ≃ [ n ] ∈ Δ op colim L ( Z ) × G a n ≃ L ( Z ) × B G a . \simeq L(Z)\times B\mathbb{G}_{a}. ≃ L ( Z ) × B G a .
Lemma 3.59. Let i : Z → X i:Z\to X i : Z → X be a closed substack of a stack X X X . Then L ^ i : L ^ Z → L ^ X \widehat{\mathcal{L}}i:\widehat{\mathcal{L}}Z\to\widehat{\mathcal{L}}X L i : L Z → L X is a closed immersion. Moreover, if U U U is the open complement to Z Z Z in X X X , then L ^ U \widehat{\mathcal{L}}U L U is the open complement to L ^ Z \widehat{\mathcal{L}}Z L Z in L ^ X \widehat{\mathcal{L}}X L X , and the natural S 1 S^{1} S 1 -equivariant map
L ^ ( X ^ Z ) → ( L ^ X ) ^ L ^ Z \widehat{\mathcal{L}}\left(\widehat{X}_{Z}\right)\to\widehat{\left(\widehat{\mathcal{L}}X\right)}_{\widehat{\mathcal{L}}Z} L ( X Z ) → ( L X ) L Z is an isomorphism.
Proof. Note that Z d R → X d R Z_{\mathrm{dR}}\to X_{\mathrm{dR}} Z dR → X dR is a monomorphism, since the de Rham prestack only evaluates on underlying reduced, classical stacks, where this holds. Hence, by Corollary 3.56, ( L Z ) d R → ( L X ) d R (\mathcal{L}Z)_{\mathrm{dR}}\to(\mathcal{L}X)_{\mathrm{dR}} ( L Z ) dR → ( L X ) dR is a monomorphism as well; thus,
Z d R ≃ ( L Z ) d R × ( L X ) d R X d R . Z_{\mathrm{dR}}\simeq(\mathcal{L}Z)_{\mathrm{dR}}\times_{(\mathcal{L}X)_{\mathrm{dR}}}X_{\mathrm{dR}}. Z dR ≃ ( L Z ) dR × ( L X ) dR X dR . In turn, this implies that
L Z ^ ≃ L Z × L X L X ^ . \widehat{\mathcal{L}Z}\simeq\mathcal{L}Z\times_{\mathcal{L}X}\widehat{\mathcal{L}X}. L Z ≃ L Z × L X L X .
so the map Z ^ → X ^ \widehat{\mathcal Z} \to\widehat{\mathcal X} Z → X is the base change of a closed immersion hence also a closed immersion. We first show that the natural map
U ^ → X ^ × X U (14) \widehat{\mathcal U} \to\widehat{\mathcal X} \mathop{\times}_{\mathcal X} \mathcal U \tag*{(14)} U → X × X U (14)
is an isomorphism. Indeed, we compute
X ^ × X U : = X d R × ( X ) d R X ^ × X U ≃ U × ( X ) d R X d R ≃ U × ( U ) d R ( U ) d R × ( X ) d R X d R ≃ U × ( U ) d R ( U × X X ) d R , \begin{aligned}
\widehat{\mathcal X} \mathop{\times}_{\mathcal X} \mathcal U &:= X_{\mathrm{dR}} \mathop{\times}_{(\mathcal X)_{\mathrm{dR}}} \widehat{\mathcal X} \mathop{\times}_{\mathcal X} \mathcal U \\
&\simeq\mathcal U \mathop{\times}_{(\mathcal X)_{\mathrm{dR}}} X_{\mathrm{dR}} \\
&\simeq\mathcal U \mathop{\times}_{(\mathcal U)_{\mathrm{dR}}} (\mathcal U)_{\mathrm{dR}} \mathop{\times}_{(\mathcal X)_{\mathrm{dR}}} X_{\mathrm{dR}} \\
&\simeq\mathcal U \mathop{\times}_{(\mathcal U)_{\mathrm{dR}}} (\mathcal U \mathop{\times}_{\mathcal X} X)_{\mathrm{dR}},
\end{aligned} X × X U := X dR × ( X ) dR X × X U ≃ U × ( X ) dR X dR ≃ U × ( U ) dR ( U ) dR × ( X ) dR X dR ≃ U × ( U ) dR ( U × X X ) dR ,
where in the last line we use the fact that ( − ) d R (-)_{\mathrm{dR}} ( − ) dR preserves limits, since it is a right adjoint. Thus, we see that it suffices to show that the canonical map
u : U → U × X X u : U \to\mathcal U \mathop{\times}_{\mathcal X} X u : U → U × X X
is an isomorphism. To see this, note that there is also a natural map
e : U × X X → U → ev U U , e : \mathcal U \mathop{\times}_{\mathcal X} X \to\mathcal U \xrightarrow{\operatorname{ev}_{U}} U, e : U × X X → U ev U U ,
where the first map is the projection to U \mathcal U U and the second is evaluation at the base point of S 1 S^1 S 1 . We claim that the composition e ∘ u ≃ id U e \circ u \simeq\operatorname{id}_{U} e ∘ u ≃ id U . Indeed, the composition e ∘ u e \circ u e ∘ u is naturally isomorphic to the composition ev U ∘ c U \operatorname{ev}_{U} \circ c_{U} ev U ∘ c U , where c U : U → U c_{U} : U \to\mathcal U c U : U → U is the constant loop map. By definition, ev U ∘ c U ≃ id U \operatorname{ev}_{U} \circ c_{U} \simeq\operatorname{id}_{U} ev U ∘ c U ≃ id U . We next claim that u ∘ e ≃ id P u \circ e \simeq\operatorname{id}_{P} u ∘ e ≃ id P , where P : = U × X X P := \mathcal U \mathop{\times}_{\mathcal X} X P := U × X X . To see this, note that the maps u , e u, e u , e are relative to X X X . Moreover, the projection P → X P \to X P → X is an open immersion by Corollary 3.56, so it is a monomorphism. Thus, the composition u ∘ e u \circ e u ∘ e is an endomorphism of P P P relative to X X X ; i.e., it is an object of
Map / X ( P , P ) . \operatorname{Map}_{/X}(P,P). Map / X ( P , P ) .
Now, since P → X P \to X P → X is a monomorphism, for every S S S , the map of spaces
Map ( S , P ) → Map ( S , X ) \operatorname{Map}(S,P) \to\operatorname{Map}(S,X) Map ( S , P ) → Map ( S , X )
has fibers which are either empty or contractible. For S = P S = P S = P , we see that the fiber over P → X P \to X P → X is precisely
Map / X ( P , P ) . \operatorname{Map}_{/X}(P,P). Map / X ( P , P ) .
Since this fiber contains in particular the identity endomorphism of P P P , we conclude that id P ≃ u ∘ e \operatorname{id}_{P} \simeq u \circ e id P ≃ u ∘ e . Thus, we conclude that the natural map (14) is an isomorphism as desired.
Let now S → Z ^ S \to\widehat{\mathcal Z} S → Z be an S S S -point. We will show that
S × X ^ U ^ ≃ ∅ , S \mathop{\times}_{\widehat{\mathcal X}} \widehat{\mathcal U} \simeq\varnothing, S × X U ≃ ∅ ,
which, together with the isomorphism (14) will prove that U ^ \widehat{\mathcal U} U is the open complement to Z ^ \widehat{\mathcal Z} Z in X ^ \widehat{\mathcal X} X . For any stack Y Y Y , Y ≃ ∅ Y \simeq\varnothing Y ≃ ∅ if and only if Y d R ≃ ∅ Y_{\mathrm{dR}} \simeq\varnothing Y dR ≃ ∅ . Applying this to the stack S × X ^ U ^ S \mathop{\times}_{\widehat{\mathcal X}} \widehat{\mathcal U} S × X U , we find
( S × X ^ U ^ ) d R ≃ S d R × ( X ^ ) d R ( U ^ ) d R ≃ S d R × X d R U d R ≃ S d R × Z d R Z d R × X d R U d R ≃ S d R × Z d R ( Z × X U ) d R ≃ S d R × Z d R ∅ d R ≃ S d R × Z d R ∅ ≃ ∅ , \begin{aligned}
\left(S \mathop{\times}_{\widehat{\mathcal X}} \widehat{\mathcal U}\right)_{\mathrm{dR}} &\simeq S_{\mathrm{dR}} \mathop{\times}_{(\widehat{\mathcal X})_{\mathrm{dR}}} (\widehat{\mathcal U})_{\mathrm{dR}} \\
&\simeq S_{\mathrm{dR}} \mathop{\times}_{X_{\mathrm{dR}}} U_{\mathrm{dR}} \\
&\simeq S_{\mathrm{dR}} \mathop{\times}_{Z_{\mathrm{dR}}} Z_{\mathrm{dR}} \mathop{\times}_{X_{\mathrm{dR}}} U_{\mathrm{dR}} \\
&\simeq S_{\mathrm{dR}} \mathop{\times}_{Z_{\mathrm{dR}}} (Z \mathop{\times}_{X} U)_{\mathrm{dR}} \\
&\simeq S_{\mathrm{dR}} \mathop{\times}_{Z_{\mathrm{dR}}} \varnothing_{\mathrm{dR}} \\
&\simeq S_{\mathrm{dR}} \mathop{\times}_{Z_{\mathrm{dR}}} \varnothing\\
&\simeq\varnothing,
\end{aligned} ( S × X U ) dR ≃ S dR × ( X ) dR ( U ) dR ≃ S dR × X dR U dR ≃ S dR × Z dR Z dR × X dR U dR ≃ S dR × Z dR ( Z × X U ) dR ≃ S dR × Z dR ∅ dR ≃ S dR × Z dR ∅ ≃ ∅ ,
where in the first and fourth lines we use that ( − ) d R (-)_{\mathrm{dR}} ( − ) dR preserves limits, as it is a right adjoint, in the second line, we use that the de Rham prestack is nil-invariant, and in the fifth line we use that U U U is the open complement to Z Z Z in X X X . We now show that the natural map
L ( X Z ) ^ ⟶ L ( X ) ^ Z ^ \widehat{\mathcal{L}(X_Z)} \longrightarrow\widehat{\mathcal{L}(X)}_{\widehat{Z}} L ( X Z ) ⟶ L ( X ) Z
is an isomorphism. Indeed, we have the canonical isomorphisms
L ( X Z ) ^ : = L ( X Z ) ^ X Z ^ ≃ L ( X Z ) × ( X Z ) d R Z d R ≃ L X × ( L X ) d R ( L Z ) d R × ( L Z ) d R Z d R → ≃ L X × ( L X ) d R Z d R ≃ L X × ( L X ) d R ( L Z ^ ) d R = : ( L X ) ( L Z ^ ) . \begin{aligned}
\widehat{\mathcal{L}(X_Z)} &:= \widehat{\mathcal{L}(X_Z)}_{\widehat{X_Z}} \\
&\simeq\mathcal{L}(X_Z) \times_{(X_Z)_{\mathrm{dR}}} Z_{\mathrm{dR}} \\
&\simeq\mathcal{L}X \times_{(\mathcal{L}X)_{\mathrm{dR}}} (\mathcal{L}Z)_{\mathrm{dR}} \times_{(\mathcal{L}Z)_{\mathrm{dR}}} Z_{\mathrm{dR}} \\
&\xrightarrow{\simeq} \mathcal{L}X \times_{(\mathcal{L}X)_{\mathrm{dR}}} Z_{\mathrm{dR}} \\
&\simeq\mathcal{L}X \times_{(\mathcal{L}X)_{\mathrm{dR}}} (\widehat{\mathcal{L}Z})_{\mathrm{dR}} \\
&=: (\mathcal{L}X)_{(\widehat{\mathcal{L}Z})}.
\end{aligned} L ( X Z ) := L ( X Z ) X Z ≃ L ( X Z ) × ( X Z ) dR Z dR ≃ L X × ( L X ) dR ( L Z ) dR × ( L Z ) dR Z dR ≃ L X × ( L X ) dR Z dR ≃ L X × ( L X ) dR ( L Z ) dR =: ( L X ) ( L Z ) .
where in the second and fifth lines we use that the functor ( − ) d R (-)_{\mathrm{dR}} ( − ) dR is nil-invariant. On the other hand, again by nil-invariance of ( − ) d R (-)_{\mathrm{dR}} ( − ) dR , we have canonical isomorphisms
( L X ^ ) Z ^ ≃ L X ^ × X d R Z d R ≃ ( L X ^ ) Z . \begin{aligned}
(\widehat{\mathcal{L}X})_{\widehat{Z}} \simeq\widehat{\mathcal{L}X} \times_{X_{\mathrm{dR}}} Z_{\mathrm{dR}} \\
&\simeq(\widehat{\mathcal{L}X})_Z.
\end{aligned} ( L X ) Z ≃ L X × X dR Z dR ≃ ( L X ) Z .
Thus, it suffices to show that the natural map
( L X ^ ) Z ⟶ ( L X ) Z ^ (\widehat{\mathcal{L}X})_Z \longrightarrow(\mathcal{L}X)_{\widehat{Z}} ( L X ) Z ⟶ ( L X ) Z
is an isomorphism. But this follows from Lemma 3.50 . □
Potent D-modules
Following [BZN25], we make the following definition.
Definition 4.1. Let X X X satisfy the conditions of Corollary 3.36. The category of potent D \mathcal{D} D -modules on X X X is defined to be
D p o t ( X ) : = IndCoh ( L X ) t T a t e = Ind ( Coh ( L X ) t S 1 ) . \mathcal{D}^{\mathrm{pot}}(X) := \operatorname{IndCoh}(\mathcal{L}X)^{t\mathrm{Tate}} = \operatorname{Ind}(\operatorname{Coh}(\mathcal{L}X)^{tS^1}). D pot ( X ) := IndCoh ( L X ) t Tate = Ind ( Coh ( L X ) t S 1 ) .
Remark 4.2. Note that, for X X X a coherent stack, the subcategory of coherent objects of both QCoh ( X ) \operatorname{QCoh}(X) QCoh ( X ) and IndCoh ( X ) \operatorname{IndCoh}(X) IndCoh ( X ) is Coh ( X ) \operatorname{Coh}(X) Coh ( X ) . Thus, in defining the category of potent D \mathcal{D} D -modules, it makes no difference whether we take QCoh ( L X ) t T a t e \operatorname{QCoh}(\mathcal{L}X)^{t\mathrm{Tate}} QCoh ( L X ) t Tate or IndCoh ( L X ) t T a t e \operatorname{IndCoh}(\mathcal{L}X)^{t\mathrm{Tate}} IndCoh ( L X ) t Tate .
Remark 4.3. The name in the above definition takes inspiration from the equivalence under Koszul duality between cyclic sheaves on the unipotent loop space of X X X and ordinary D \mathcal{D} D -modules on X X X . Potent D \mathcal{D} D -modules correspond to cyclic sheaves on the full—or potent —loop space rather than just the unipotent loop space of X X X .
Lemma 4.4. Suppose that C , D ∈ Coh t \mathcal{C}, \mathcal{D} \in\operatorname{Coh}_{t} C , D ∈ Coh t , and that L : C ⇄ D : R L : \mathcal{C} \rightleftarrows\mathcal{D} : R L : C ⇄ D : R are adjoint functors which each are left t t t -exact up to a shift, exact, and which preserve uniformly left t t t -bounded filtered colimits. Then the induced functors R ( L ) : R ( C ) ⇄ R ( D ) : R ( R ) \mathcal{R}(L) : \mathcal{R}(\mathcal{C}) \rightleftarrows\mathcal{R}(\mathcal{D}) : \mathcal{R}(R) R ( L ) : R ( C ) ⇄ R ( D ) : R ( R ) are adjoint to one another.
Proof. The existence of the induced functors R ( L ) \mathcal{R}(L) R ( L ) and R ( R ) \mathcal{R}(R) R ( R ) is established by [3 Lemma 4.6.1] , wherein it is also stated that they preserve colimits. We first note that L L L preserves coherent objects. Indeed, let c ∈ Coh ( C ) c \in\operatorname{Coh}(\mathcal{C}) c ∈ Coh ( C ) , and let { d i } \{d_i\} { d i } be a uniformly left t-bounded filtered diagram in D \mathcal{D} D . Then
Map D ( L c , colim i d i ) ≃ Map C ( c , R ( colim i d i ) ) (15) \operatorname{Map}_{\mathcal{D}}(Lc,\operatorname*{colim}_{i}d_i) \simeq\operatorname{Map}_{\mathcal{C}}(c,R(\operatorname*{colim}_{i}d_i)) \tag*{(15)} Map D ( L c , i colim d i ) ≃ Map C ( c , R ( i colim d i )) (15)
≃ Map C ( c , colim i R ( d i ) ) (16) \simeq\operatorname{Map}_{\mathcal{C}}(c,\operatorname*{colim}_{i}R(d_i)) \tag*{(16)} ≃ Map C ( c , i colim R ( d i )) (16)
≃ colim i Map C ( c , R ( d i ) ) (17) \simeq\operatorname*{colim}_{i}\operatorname{Map}_{\mathcal{C}}(c,R(d_i)) \tag*{(17)} ≃ i colim Map C ( c , R ( d i )) (17)
≃ colim i Map D ( L c , d i ) , (18) \simeq\operatorname*{colim}_{i}\operatorname{Map}_{\mathcal{D}}(Lc,d_i), \tag*{(18)} ≃ i colim Map D ( L c , d i ) , (18)
where (16) follows from the hypothesis that R R R preserves uniformly left t-bounded filtered colimits, and (17) follows from c c c being almost compact and the diagram R ( d i ) R(d_i) R ( d i ) being uniformly left t-bounded. Thus L c Lc L c is almost compact. At the same time, c c c , being coherent, is left t-bounded (i.e. in C < ∞ \mathcal{C}_{<\infty} C < ∞ ). The functor L L L being left t-exact up to a shift means precisely that L ( C < ∞ ) ⊂ D < ∞ L(\mathcal{C}_{<\infty}) \subset\mathcal{D}_{<\infty} L ( C < ∞ ) ⊂ D < ∞ . Thus L c ∈ D < ∞ Lc \in\mathcal{D}_{<\infty} L c ∈ D < ∞ , hence coherent in the sense of [3 Definition 4.2.2] .
Since L L L preserves coherent objects, the regularized functor R ( L ) \mathcal{R}(L) R ( L ) preserves compact objects. By [3 Proposition 4.3.2] , the canonical functors R ( C ) → C \mathcal{R}(\mathcal{C}) \to\mathcal{C} R ( C ) → C and R ( D ) → D \mathcal{R}(\mathcal{D}) \to\mathcal{D} R ( D ) → D restrict to equivalences R ( C ) < ∞ → ∼ C ∞ \mathcal{R}(\mathcal{C})_{<\infty} \xrightarrow{\sim} \mathcal{C}_{\infty} R ( C ) < ∞ ∼ C ∞ and R ( D ) < ∞ → ∼ D < ∞ \mathcal{R}(\mathcal{D})_{<\infty} \xrightarrow{\sim} \mathcal{D}_{<\infty} R ( D ) < ∞ ∼ D < ∞ . Thus, for any c ∈ Coh ( C ) c \in\operatorname{Coh}(\mathcal{C}) c ∈ Coh ( C ) and d ∈ Coh ( D ) d \in\operatorname{Coh}(\mathcal{D}) d ∈ Coh ( D ) , we have
Map R ( D ) ( R ( L ) c , d ) ≃ Map R ( D ) < ∞ ( R ( L ) c , d ) ≃ Map D < ∞ ( R ( L ) c , d ) ≃ Map D < ∞ ( L c , d ) ≃ Map C < ∞ ( c , R d ) ≃ Map C < ∞ ( c , R ( R ) d ) ≃ Map R ( C ) < ∞ ( c , R ( R ) d ) ≃ Map R ( C ) ( c , R ( R ) d ) . \begin{aligned}
\operatorname{Map}_{\mathcal{R}(\mathcal{D})}(\mathcal{R}(L)c,d)
&\simeq\operatorname{Map}_{\mathcal{R}(\mathcal{D})_{<\infty}}(\mathcal{R}(L)c,d) \\
&\simeq\operatorname{Map}_{\mathcal{D}_{<\infty}}(\mathcal{R}(L)c,d) \\
&\simeq\operatorname{Map}_{\mathcal{D}_{<\infty}}(Lc,d) \\
&\simeq\operatorname{Map}_{\mathcal{C}_{<\infty}}(c,Rd) \\
&\simeq\operatorname{Map}_{\mathcal{C}_{<\infty}}(c,\mathcal{R}(R)d) \\
&\simeq\operatorname{Map}_{\mathcal{R}(\mathcal{C})_{<\infty}}(c,\mathcal{R}(R)d) \\
&\simeq\operatorname{Map}_{\mathcal{R}(\mathcal{C})}(c,\mathcal{R}(R)d).
\end{aligned} Map R ( D ) ( R ( L ) c , d ) ≃ Map R ( D ) < ∞ ( R ( L ) c , d ) ≃ Map D < ∞ ( R ( L ) c , d ) ≃ Map D < ∞ ( L c , d ) ≃ Map C < ∞ ( c , R d ) ≃ Map C < ∞ ( c , R ( R ) d ) ≃ Map R ( C ) < ∞ ( c , R ( R ) d ) ≃ Map R ( C ) ( c , R ( R ) d ) .
Since R ( L ) ( c ) \mathcal{R}(L)(c) R ( L ) ( c ) is compact, and since R ( R ) \mathcal{R}(R) R ( R ) preserves colimits by construction, both sides of the equivalence Map R ( D ) ( R ( L ) c , d ) ≃ Map R ( C ) ( c , R ( R ) d ) \operatorname{Map}_{\mathcal{R}(\mathcal{D})}(\mathcal{R}(L)c,d) \simeq\operatorname{Map}_{\mathcal{R}(\mathcal{C})}(c,\mathcal{R}(R)d) Map R ( D ) ( R ( L ) c , d ) ≃ Map R ( C ) ( c , R ( R ) d ) commute with filtered colimits in d d d . On the other hand, since Map ( − , − ) \operatorname{Map}(-,-) Map ( − , − ) commutes with colimits in the first argument, and since R ( L ) \mathcal{R}(L) R ( L ) commutes with colimits in c c c , we deduce that both sides of the same equivalence also commute with filtered colimits in c c c . Thus, we get that Map R ( C ) ( R ( L ) x , y ) ≃ Map R ( D ) ( x , R ( R ) y ) \operatorname{Map}_{\mathcal{R}(\mathcal{C})}(\mathcal{R}(L)x,y) \simeq\operatorname{Map}_{\mathcal{R}(\mathcal{D})}(x,\mathcal{R}(R)y) Map R ( C ) ( R ( L ) x , y ) ≃ Map R ( D ) ( x , R ( R ) y ) , for any x ∈ R ( C ) x \in\mathcal{R}(\mathcal{C}) x ∈ R ( C ) and y ∈ R ( D ) y \in\mathcal{R}(\mathcal{D}) y ∈ R ( D ) , as desired. It remains to observe that this equivalence is natural in both x x x and y y y , since all the equivalences in the chain above are natural in c c c and d d d . □
Lemma 4.9. Suppose that L : C ⇄ D : R L : \mathcal{C} \rightleftarrows\mathcal{D} : R L : C ⇄ D : R is an adjunction in S 1 S^1 S 1 -mod. Then the induced functors L t S 1 : C t S 1 ⇄ D t S 1 : R t S 1 L^{tS^1} : \mathcal{C}^{tS^1} \rightleftarrows\mathcal{D}^{tS^1} : R^{tS^1} L t S 1 : C t S 1 ⇄ D t S 1 : R t S 1 form an adjunction in k [ β ± 1 ] k[\beta^{\pm1}] k [ β ± 1 ] -mod.
Proof. It is known that S 1 S^1 S 1 -invariants are a functor ( − ) S 1 : S 1 (-)^{S^1} : S^1 ( − ) S 1 : S 1 -mod → k [ β ] \to k[\beta] → k [ β ] -mod, and basechange is a functor − ⊗ k [ β ] k [ β ± 1 ] : k [ β ] -\otimes_{k[\beta]} k[\beta^{\pm1}] : k[\beta] − ⊗ k [ β ] k [ β ± 1 ] : k [ β ] -mod → k [ β ± 1 ] \to k[\beta^{\pm1}] → k [ β ± 1 ] -mod. This implies that an adjunction in each of the source categories is taken to an adjunction in each of the target categories, which proves the claim. □
Corollary 4.10. Suppose that C , D ∈ Fun ( B S 1 , Coh t ) \mathcal{C}, \mathcal{D} \in\operatorname{Fun}(BS^1,\operatorname{Coh}_t) C , D ∈ Fun ( B S 1 , Coh t ) , and that L : C ⇄ D : R L : \mathcal{C} \rightleftarrows\mathcal{D} : R L : C ⇄ D : R are S 1 S^1 S 1 -equivariant adjoint functors which each are left t-exact up to a shift, exact, and which preserve uniformly left t-bounded filtered colimits. Then the induced functors L t Tate : C t Tate ⇄ D t Tate : R t Tate L^{t\operatorname{Tate}} : \mathcal{C}^{t\operatorname{Tate}} \rightleftarrows\mathcal{D}^{t\operatorname{Tate}} : R^{t\operatorname{Tate}} L t Tate : C t Tate ⇄ D t Tate : R t Tate are adjoint to one another.
Proof. This follows immediately from Lemma 4.4 and Lemma 4.9 .
Potent D-modules with support
Definition 4.11. Let X X X be a derived stack. Given a closed embedding Z → i X Z \xrightarrow{i} X Z i X , let U → j X U \xrightarrow{j} X U j X be the complementary open embedding. We define
D Z p o t ( X ) : = fib ( D p o t ( X ) → j ′ ! D p o t ( U ) ) \mathcal{D}^{\mathrm{pot}}_Z(X) := \operatorname{fib}\left(\mathcal{D}^{\mathrm{pot}}(X) \xrightarrow{j'!} \mathcal{D}^{\mathrm{pot}}(U)\right) D Z pot ( X ) := fib ( D pot ( X ) j ′ ! D pot ( U ) )
to be the subcategory of potent D \mathcal{D} D -modules on X X X supported on Z Z Z .
Lemma 4.12. Let X X X be an algebraic stack almost of finite type and with affine diagonal, and let Z ↪ X Z \hookrightarrow X Z ↪ X be a closed substack such that Z ↪ X Z \hookrightarrow X Z ↪ X is almost of finite presentation. Let i ^ : X ^ Z → X \widehat{i}: \widehat{X}_Z \to X i : X Z → X denote induced map from the formal completion to X X X . Then ! ! ! -pullback along i ^ \widehat{i} i induces an equivalence,
i ^ ! : D Z p o t ( X ) → ∼ D p o t ( X ^ Z ) , \widehat{i}^{!}:\mathcal{D}^{\mathrm{pot}}_Z(X) \xrightarrow{\sim} \mathcal{D}^{\mathrm{pot}}(\widehat{X}_Z), i ! : D Z pot ( X ) ∼ D pot ( X Z ) ,
with adjoint inverse i ^ ∗ \widehat{i}_* i ∗ .
Proof. We will deduce the result from the corresponding statement for ind-coherent sheaves. By [Pre11, Theorem 5.2.8], it is know that ! ! ! -pullback along the induced map L i ^ : L X X Z ↪ L X \mathcal{L}\widehat{i}:\mathcal{L}X_{XZ} \hookrightarrow\mathcal{L}X L i : L X X Z ↪ L X induces an equivalence,
( L i ^ ) ! : IndCoh ( L X ) L Z → ∼ IndCoh ( L X ^ Z ) . (\mathcal{L}\widehat{i})^!:\operatorname{IndCoh}(\mathcal{L}X)_{\mathcal{L}Z} \xrightarrow{\sim} \operatorname{IndCoh}(\mathcal{L}\widehat{X}_Z). ( L i ) ! : IndCoh ( L X ) L Z ∼ IndCoh ( L X Z ) .
By Lemma 3.57, there is an S 1 S^1 S 1 -equivariant isomorphism, L ( X ^ Z ) ≃ L X X Z \mathcal{L}(\widehat{X}_Z) \simeq\mathcal{L}X_{XZ} L ( X Z ) ≃ L X X Z , implying that
IndCoh ( L X ^ Z ) t T a t e ≃ D p o t ( X ^ Z ) . \operatorname{IndCoh}\left(\mathcal{L}\widehat{X}_Z\right)^{t\mathrm{Tate}} \simeq\mathcal{D}^{\mathrm{pot}}(\widehat{X}_Z). IndCoh ( L X Z ) t Tate ≃ D pot ( X Z ) .
It therefore suffices to show that ( IndCoh ( L X ) L Z ) t T a t e ≃ fib ( j p o t ′ ) (\operatorname{IndCoh}(\mathcal{L}X)_{\mathcal{L}Z})^{t\mathrm{Tate}} \simeq\operatorname{fib}(j'_{\mathrm{pot}}) ( IndCoh ( L X ) L Z ) t Tate ≃ fib ( j pot ′ ) , where we use j p o t ′ j'_{\mathrm{pot}} j pot ′ to distinguish the potent D \mathcal{D} D -module pullback from the IndCoh pullback.
Let j : U ↪ X j:U \hookrightarrow X j : U ↪ X denote the complement to i : Z ↪ X i:Z \hookrightarrow X i : Z ↪ X . By Corollary 3.56, the inclusions
L i : L Z ↪ L X ← L j L U \mathcal{L}i:\mathcal{L}Z \hookrightarrow\mathcal{L}X \xleftarrow{\mathcal{L}j} \mathcal{L}U L i : L Z ↪ L X L j L U
give a decomposition of L X \mathcal{L}X L X into a complementary pair of open and closed substacks. The inclusion L j \mathcal{L}j L j is canonically S 1 S^1 S 1 -equivariant, so L j ! \mathcal{L}j^! L j ! is an S 1 S^1 S 1 -equivariant functor, meaning we have an S 1 S^1 S 1 -equivariant fiber sequence
fib ( L j ! ) ⟶ IndCoh ( L X ) → L j ! IndCoh ( L U ) . \operatorname{fib}(\mathcal{L}j^!) \longrightarrow\operatorname{IndCoh}(\mathcal{L}X) \xrightarrow{\mathcal{L}j^!} \operatorname{IndCoh}(\mathcal{L}U). fib ( L j ! ) ⟶ IndCoh ( L X ) L j ! IndCoh ( L U ) .
By [Ber21, Proposition 2.3.3], the compact objects of IndCoh ( L X ) \operatorname{IndCoh}(\mathcal{L}X) IndCoh ( L X ) (resp. IndCoh ( L U ) \operatorname{IndCoh}(\mathcal{L}U) IndCoh ( L U ) ) are given by Coh ( L X ) \operatorname{Coh}(\mathcal{L}X) Coh ( L X ) (resp. Coh ( L U ) \operatorname{Coh}(\mathcal{L}U) Coh ( L U ) ). Since L j \mathcal{L}j L j is an open immersion, j ! ≃ j ∗ j^! \simeq j^* j ! ≃ j ∗ . Thus, L j ! \mathcal{L}j^! L j ! has right adjoint L j ∗ \mathcal{L}j_* L j ∗ , which is automatically continuous, implying that L j ! \mathcal{L}j^! L j ! preserves compact objects. We therefore obtain a fiber sequence of small categories:
fib ( L j ! ∣ Coh ( L X ) ) ⟶ Coh ( L X ) → L j ! ∣ Coh ( L X ) Coh ( L U ) . \operatorname{fib}\left(\mathcal{L}j^!|_{\operatorname{Coh}(\mathcal{L}X)}\right) \longrightarrow\operatorname{Coh}(\mathcal{L}X) \xrightarrow{\mathcal{L}j^!|_{\operatorname{Coh}(\mathcal{L}X)}} \operatorname{Coh}(\mathcal{L}U). fib ( L j ! ∣ Coh ( L X ) ) ⟶ Coh ( L X ) L j ! ∣ Coh ( L X ) Coh ( L U ) .
By [Ber21, Proposition 2.5.2], for any closed immersion of laft quasi-compact stacks Y ↪ X \mathcal{Y} \hookrightarrow\mathcal{X} Y ↪ X , the image of Coh ( Y ) \operatorname{Coh}(\mathcal{Y}) Coh ( Y ) under pushforward along the induced map i : Y → X ^ Y i:\mathcal{Y} \to\widehat{\mathcal{X}}_{\mathcal{Y}} i : Y → X Y generates IndCoh ( X ^ Y ) \operatorname{IndCoh}(\widehat{\mathcal{X}}_{\mathcal{Y}}) IndCoh ( X Y ) . At the same time, we know from [Pre11, Theorem 5.2.8] that IndCoh ( X ^ Y ) ≃ IndCoh Y ( X ) \operatorname{IndCoh}(\widehat{\mathcal{X}}_{\mathcal{Y}}) \simeq\operatorname{IndCoh}_{\mathcal{Y}}(\mathcal{X}) IndCoh ( X Y ) ≃ IndCoh Y ( X ) . Hence, the proof of [GR17a, Chapter II.1 Proposition 6.1.3(d)] goes through in our setting, showing that
IndCoh L Z ( L X ) : = fib ( L j ! ) ≃ Ind ( fib ( L j ! ∣ Coh ( L X ) ) ) = : Ind ( Coh L Z ( L X ) ) . \operatorname{IndCoh}_{\mathcal{L}Z}(\mathcal{L}X) := \operatorname{fib}(\mathcal{L}j^!) \simeq\operatorname{Ind}\left(\operatorname{fib}\left(\mathcal{L}j^!|_{\operatorname{Coh}(\mathcal{L}X)}\right)\right) =: \operatorname{Ind}(\operatorname{Coh}_{\mathcal{L}Z}(\mathcal{L}X)). IndCoh L Z ( L X ) := fib ( L j ! ) ≃ Ind ( fib ( L j ! ∣ Coh ( L X ) ) ) =: Ind ( Coh L Z ( L X )) .
Thus, we have a localization sequence of objects of Fun ( B S 1 , Coh t ) \operatorname{Fun}(BS^1,\operatorname{Coh}_t) Fun ( B S 1 , Coh t ) ,
IndCoh ( L X ^ Z ) ← ≃ IndCoh L Z ( L X ) ⟶ IndCoh ( L X ) ⟶ IndCoh ( L U ) . \operatorname{IndCoh}(\mathcal{L}\widehat{X}_Z) \xleftarrow{\simeq} \operatorname{IndCoh}_{\mathcal{L}Z}(\mathcal{L}X) \longrightarrow\operatorname{IndCoh}(\mathcal{L}X) \longrightarrow\operatorname{IndCoh}(\mathcal{L}U). IndCoh ( L X Z ) ≃ IndCoh L Z ( L X ) ⟶ IndCoh ( L X ) ⟶ IndCoh ( L U ) .
Moreover, the right adjoints in this sequence are t t t -bounded by Lemma 3.51, Lemma 3.44 and Lemma 3.45 . Thus, by Theorem 3.12 , we obtain a localization sequence after applying ( − ) t T a t e (-)^{t\mathrm{Tate}} ( − ) t Tate ; hence,
D Z p o t ( X ) → ∼ D p o t ( X ^ Z ) \mathcal{D}^{\mathrm{pot}}_Z(X) \xrightarrow{\sim} \mathcal{D}^{\mathrm{pot}}(\widehat{X}_Z) D Z pot ( X ) ∼ D pot ( X Z )
as desired. □
Corollary 4.13. In the setting of Lemma 4.12, ! ! ! -pullback along ı ^ \widehat{\imath} induces an equivalence
IndCoh L ^ Z ( L ^ ( X ) ) t T a t e → ∼ IndCoh ( L ^ ( X ^ Z ) ) t T a t e \operatorname{IndCoh}_{\widehat{\mathcal{L}}_Z}(\widehat{\mathcal{L}}(X))^{t\mathrm{Tate}} \xrightarrow{\sim} \operatorname{IndCoh}(\widehat{\mathcal{L}}(\widehat{X}_Z))^{t\mathrm{Tate}} IndCoh L Z ( L ( X ) ) t Tate ∼ IndCoh ( L ( X Z ) ) t Tate Proof. We will show that the proof of Lemma 4.12 applies to the map L ^ ı ^ \widehat{\mathcal{L}}\widehat{\imath} L . Indeed, to obtain the desired localization sequence, it is sufficient, by Lemma 3.59 , to show that the natural map
L ^ ( X ^ Z ) ⟶ ( L ^ ( X ) ) Z ^ (19) \widehat{\mathcal{L}}(\widehat{X}_Z) \longrightarrow\bigl(\widehat{\mathcal{L}}(X)\bigr)_{\widehat{Z}} \tag*{(19)} L ( X Z ) ⟶ ( L ( X ) ) Z (19) is an isomorphism and that the ind-coherent sheaf categories in question are compactly generated. We have that (19) is an isomorphism by Lemma 3.59 . We now seek to apply [CW23, Proposition 5.30] to show that all categories in question are compactly generated by Coh. For this, note that by nil-invariance of ( − ) d R (-)_{\mathrm{dR}} ( − ) dR , we have canonical isomorphisms
L ^ X L ^ Z ≃ L X × ( L X ) d R X d R × X d R Z d R \widehat{\mathcal{L}}_X\widehat{\mathcal{L}}_Z \simeq\mathcal{L}X \mathop{\times}_{(\mathcal{L}X)_{\mathrm{dR}}} X_{\mathrm{dR}} \mathop{\times}_{X_{\mathrm{dR}}} Z_{\mathrm{dR}} L X L Z ≃ L X × ( L X ) dR X dR × X dR Z dR ≃ L X × ( L X ) d R Z d R \simeq\mathcal{L}X \mathop{\times}_{(\mathcal{L}X)_{\mathrm{dR}}} Z_{\mathrm{dR}} ≃ L X × ( L X ) dR Z dR = : ( L X ) Z . =: (\mathcal{L}X)_Z. =: ( L X ) Z . Since L X \mathcal{L}X L X is an algebraic stack almost of finite type and with affine diagonal by assumption, we see that by Lemma 3.35 ( L X ) Z (\mathcal{L}X)_Z ( L X ) Z is coherent ind-geometric, so loc. cit. applies to it. For the stacks L ^ X \widehat{\mathcal{L}}X L X and L ^ U \widehat{\mathcal{L}}U L U , we again use Lemma 3.35. Thus, we obtain a localization sequence of objects of Fun ( B S 1 , Coh t ) \operatorname{Fun}(BS^1,\operatorname{Coh}_t) Fun ( B S 1 , Coh t )
IndCoh ( L ^ ( X ^ Z ) ) ← ∼ IndCoh L ^ Z ( L ^ X ) ⟶ IndCoh ( L ^ X ) ⟶ IndCoh ( L ^ U ) . \operatorname{IndCoh}(\widehat{\mathcal{L}}(\widehat{X}_Z)) \xleftarrow{\sim} \operatorname{IndCoh}_{\widehat{\mathcal{L}}_Z}(\widehat{\mathcal{L}}X) \longrightarrow\operatorname{IndCoh}(\widehat{\mathcal{L}}X) \longrightarrow\operatorname{IndCoh}(\widehat{\mathcal{L}}U). IndCoh ( L ( X Z )) ∼ IndCoh L Z ( L X ) ⟶ IndCoh ( L X ) ⟶ IndCoh ( L U ) . It remains to check that the right adjoints in this localization sequence are t t t -bounded. For the map L ^ ı ^ : L ^ ( X ^ Z ) → L ^ X \widehat{\mathcal{L}}\widehat{\imath}:\widehat{\mathcal{L}}(\widehat{X}_Z)\to\widehat{\mathcal{L}}X L : L ( X Z ) → L X , we use the natural equivalence
L ^ ( X ^ Z ) ≃ ( L X ) ^ Z , \widehat{\mathcal{L}}(\widehat{X}_Z) \simeq\widehat{(\mathcal{L}X)}_Z, L ( X Z ) ≃ ( L X ) Z , so that by Lemma 3.45 , the ! ! ! -pullback along L ^ ( X ^ Z ) → L X \widehat{\mathcal{L}}(\widehat{X}_Z)\to\mathcal{L}X L ( X Z ) → L X is t t t -bounded. Now, the map L ^ ( X ^ Z ) → L X \widehat{\mathcal{L}}(\widehat{X}_Z)\to\mathcal{L}X L ( X Z ) → L X factors as
L ^ ( X ^ Z ) → L ^ ı ^ L ^ X → ι X L X , \widehat{\mathcal{L}}(\widehat{X}_Z) \xrightarrow{\widehat{\mathcal{L}}\widehat{\imath}} \widehat{\mathcal{L}}X \xrightarrow{\iota_X} \mathcal{L}X, L ( X Z ) L L X ι X L X , and the ! ! ! -pullback along L ^ X → L X \widehat{\mathcal{L}}X\to\mathcal{L}X L X → L X is t t t -bounded by loc. cit. . Moreover, by [Pre11, Theorem 5.2.8], the ∗ * ∗ -pushforward along ι X : L ^ X → L X \iota_X:\widehat{\mathcal{L}}X\to\mathcal{L}X ι X : L X → L X is fully faithful and t t t -exact. Thus, the unit id → ∼ ı ^ ! ı ^ ∗ \operatorname{id}\xrightarrow{\sim}\widehat{\imath}_!\widehat{\imath}^* id ∼ ! ∗ is an isomorphism. We can thus write
( L ^ ı ^ ) ! ≃ ( L ^ ı ^ ) ! ι X ∗ ι X ∗ ≃ ( ι X ∘ L ^ ı ^ ) ! ι X ∗ . (\widehat{\mathcal{L}}\widehat{\imath})^! \simeq(\widehat{\mathcal{L}}\widehat{\imath})^! \iota_X^*\iota_{X*} \simeq(\iota_X\circ\widehat{\mathcal{L}}\widehat{\imath})^! \iota_{X*}. ( L ) ! ≃ ( L ) ! ι X ∗ ι X ∗ ≃ ( ι X ∘ L ) ! ι X ∗ . The right-hand side is t t t -bounded; hence, we conclude that ( L ^ ı ^ ) ! (\widehat{\mathcal{L}}\widehat{\imath})^! ( L ) ! is t t t -bounded as well. Now, we will show that ∗ * ∗ -pushforward along L ^ ȷ ^ : L ^ U → L ^ X \widehat{\mathcal{L}}\widehat{\jmath}:\widehat{\mathcal{L}}U\to\widehat{\mathcal{L}}X L : L U → L X is also t t t -bounded. Consider the Cartesian square
L ^ U → L ^ ȷ ^ L ^ X ι U ↓ ↓ ι X L U → L ȷ ^ L X \begin{CD}
\widehat{\mathcal{L}}U @>{\widehat{\mathcal{L}}\widehat{\jmath}}>> \widehat{\mathcal{L}}X \\
@V{\iota_U}VV @VV{\iota_X}V \\
\mathcal{L}U @>{\mathcal{L}\widehat{\jmath}}>> \mathcal{L}X
\end{CD} L U ι U ↓ ⏐ L U L L L X ↓ ⏐ ι X L X
Base change along this square shows that
ι X ! ( L j ) ∗ ≃ ( L j ^ ) ∗ ι U ! . \iota_X^!(\mathcal{L}j)_* \simeq(\mathcal{L}\widehat{j})_*\iota_U^!. ι X ! ( L j ) ∗ ≃ ( L j ) ∗ ι U ! .
Again, by [Pre11, Theorem 5.2.8], ι U ∗ \iota_{U*} ι U ∗ is fully faithful and t t t -exact, so ι U ! ι U ∗ ≃ id \iota_U^!\iota_{U*} \simeq\operatorname{id} ι U ! ι U ∗ ≃ id . Thus,
( L j ^ ) ∗ ≃ ι X ! ( L j ) ∗ ι U ∗ . (\mathcal{L}\widehat{j})_* \simeq\iota_X^!(\mathcal{L}j)_*\iota_{U*}. ( L j ) ∗ ≃ ι X ! ( L j ) ∗ ι U ∗ .
Now, all of the functors on the right-hand side are t t t -bounded. Indeed, ( L j ) ∗ (\mathcal{L}j)_* ( L j ) ∗ is t t t -bounded by Lemma 3.44 , and ι X ! \iota_X^! ι X ! is t t t -bounded by Lemma 3.45 . This allows the same method of proof as in Lemma 4.12.
Nil-invariance of cyclic sheaves on the α \alpha α -potent loop space
α \alpha α -potent loop space
The free loop space of a quotient by a reductive group admits a kind of “Jordan decomposition,” described in [Che20, §3], that allows one to work over equivalence classes of semisimple elements in G / G G/G G / G one at a time. Namely, given a global quotient X / G X/G X / G , with G G G a reductive group, there is a canonical map,
μ : L ( X / G ) → L ( B G ) ≃ G / G → G / / G . \mu:\mathcal{L}(X/G) \to\mathcal{L}(BG) \simeq G/G \to G/\!/G. μ : L ( X / G ) → L ( B G ) ≃ G / G → G / / G .
where G / / G G/\!/G G / / G is the affinization of G / G G/G G / G —a variety parametrizing semisimple conjugacy classes of G G G . For each semisimple conjugacy class α ∈ G / / G \alpha\in G/\!/G α ∈ G / / G ,32 one defines the α \alpha α -potent loop space to be the formal completion of L ( X / G ) \mathcal{L}(X/G) L ( X / G ) along the fiber of μ : L ( X / G ) → G / / G \mu:\mathcal{L}(X/G) \to G/\!/G μ : L ( X / G ) → G / / G over α \alpha α :
L α ( X / G ) : = L ( X / G ) ^ μ − 1 ( α ) . \mathcal{L}^{\alpha}(X/G) := \widehat{\mathcal{L}(X/G)}_{\mu^{-1}(\alpha)}. L α ( X / G ) := L ( X / G ) μ − 1 ( α ) .
From definition of the formal completion, there is a canonical map ι α : L α ( X / G ) → L ( X / G ) \iota_\alpha:\mathcal{L}^{\alpha}(X/G) \to\mathcal{L}(X/G) ι α : L α ( X / G ) → L ( X / G ) . We have the following analog of Lemma 3.57.
Lemma 5.1. Let X ≃ X / G \mathcal{X} \simeq X/G X ≃ X / G be a global quotient stack with G G G a linear group with reductive neutral component acting on a prestack X X X . Let Z ↪ X Z \hookrightarrow X Z ↪ X be a closed immersion with complement U ↪ X U \hookrightarrow X U ↪ X . Then the induced map L α ( Z ) → L α ( X ) \mathcal{L}^{\alpha}(Z) \to\mathcal{L}^{\alpha}(X) L α ( Z ) → L α ( X ) is a closed immersion whose open complement is L α ( U ) → L α ( X ) \mathcal{L}^{\alpha}(U) \to\mathcal{L}^{\alpha}(X) L α ( U ) → L α ( X ) . Moreover, we have an equivalence,
L α ( X Z ^ ) ≃ ( L α X ) L α Z , \mathcal{L}^{\alpha}(\widehat{X_Z}) \simeq\left(\mathcal{L}^{\alpha}X\right)_{\mathcal{L}^{\alpha}Z}, L α ( X Z ) ≃ ( L α X ) L α Z , for each α ∈ G / / G \alpha\in G/\!/G α ∈ G / / G .
Proof. By Lemma 3.48 , X Z ^ / G ≃ X Z ^ \widehat{X_Z}/G \simeq\widehat{\mathcal{X}_Z} X Z / G ≃ X Z . Denote by O α ( X Z ^ ) \mathcal{O}_{\alpha}(\widehat{\mathcal{X}_Z}) O α ( X Z ) (resp. O α ( Z ) \mathcal{O}_{\alpha}(Z) O α ( Z ) , O α ( X ) \mathcal{O}_{\alpha}(X) O α ( X ) , O α ( U ) \mathcal{O}_{\alpha}(U) O α ( U ) ) the derived fiber of μ X Z ^ : L ( X Z ^ ) ≃ L ( X Z ^ / G ) → G / / G \mu_{\widehat{\mathcal{X}_Z}}:\mathcal{L}(\widehat{\mathcal{X}_Z}) \simeq\mathcal{L}(\widehat{X_Z}/G) \to G/\!/G μ X Z : L ( X Z ) ≃ L ( X Z / G ) → G / / G over α ∈ G / / G \alpha\in G/\!/G α ∈ G / / G (resp. the derived fibers of μ X \mu_X μ X , μ Z \mu_Z μ Z , and μ U \mu_U μ U over α ∈ G / / G \alpha\in G/\!/G α ∈ G / / G ).
Let S → L α X S \to\mathcal{L}^{\alpha}X S → L α X be an S S S -point. We will first show that
L α U ≃ L α X × L X L U . \mathcal{L}^{\alpha}U \simeq\mathcal{L}^{\alpha}X \times_{\mathcal{L}X} \mathcal{L}U. L α U ≃ L α X × L X L U .
Indeed,
L α U ≃ L U × ( L U ) d R ( L U × L X O α ( X ) ) d R ≃ L U × ( L X ) d R O α ( X ) ≃ L U × L X L X × ( L X ) d R O α ( X ) = : L U × L X L α X , (20) \begin{aligned}
\mathcal{L}^{\alpha}U &\simeq\mathcal{L}U \mathop{\times}_{(\mathcal{L}U)_{\mathrm{dR}}} \left(\mathcal{L}U \mathop{\times}_{\mathcal{L}X} \mathcal{O}_{\alpha}(X)\right)_{\mathrm{dR}} \\
&\simeq\mathcal{L}U \mathop{\times}_{(\mathcal{L}X)_{\mathrm{dR}}} \mathcal{O}_{\alpha}(X) \\
&\simeq\mathcal{L}U \mathop{\times}_{\mathcal{L}X} \mathcal{L}X \mathop{\times}_{(\mathcal{L}X)_{\mathrm{dR}}} \mathcal{O}_{\alpha}(X) \\
&=: \mathcal{L}U \mathop{\times}_{\mathcal{L}X} \mathcal{L}^{\alpha}X, \tag*{(20)}
\end{aligned} L α U ≃ L U × ( L U ) dR ( L U × L X O α ( X ) ) dR ≃ L U × ( L X ) dR O α ( X ) ≃ L U × L X L X × ( L X ) dR O α ( X ) =: L U × L X L α X , (20)
where in (20) , we use the Cartesian diagram
O α ( U ) → L U ↓ ↓ O α ( X ) → L X \begin{CD}
\mathcal{O}_{\alpha}(\mathcal{U}) @>>> \mathcal{L}\mathcal{U} \\
@VVV @VVV \\
\mathcal{O}_{\alpha}(\mathcal{X}) @>>> \mathcal{L}\mathcal{X}
\end{CD} O α ( U ) ↓ ⏐ O α ( X ) L U ↓ ⏐ LX
and in (5.3) we use that ( − ) d R (-)_{\mathrm{dR}} ( − ) dR preserves limits by [GR14, Lemma 1.1.4]. Thus, to see that L α U \mathcal{L}^{\alpha}\mathcal{U} L α U is the open complement to L α Z \mathcal{L}^{\alpha}\mathcal{Z} L α Z , it is sufficient to show that
S × L α X L α Z ≃ ∅ ⟺ S × L X L Z ≃ ∅ . S \mathop{\times}_{\mathcal{L}^{\alpha}\mathcal{X}} \mathcal{L}^{\alpha}\mathcal{Z} \simeq\varnothing\Longleftrightarrow S \mathop{\times}_{\mathcal{L}\mathcal{X}} \mathcal{L}\mathcal{Z} \simeq\varnothing. S × L α X L α Z ≃ ∅ ⟺ S × LX LZ ≃ ∅ .
In fact, we claim that
S × L α X L α Z ≃ S × L X L Z . S \mathop{\times}_{\mathcal{L}^{\alpha}\mathcal{X}} \mathcal{L}^{\alpha}\mathcal{Z} \simeq S \mathop{\times}_{\mathcal{L}\mathcal{X}} \mathcal{L}\mathcal{Z}. S × L α X L α Z ≃ S × LX LZ .
Indeed, by using the Cartesian square
O α ( Z ) → L Z ↓ ↓ O α ( X ) → L X (21) \begin{CD}
\mathcal{O}_{\alpha}(\mathcal{Z}) @>>> \mathcal{L}\mathcal{Z} \\
@VVV @VVV \\
\mathcal{O}_{\alpha}(\mathcal{X}) @>>> \mathcal{L}\mathcal{X} \tag*{(21)}
\end{CD} O α ( Z ) ↓ ⏐ O α ( X ) LZ ↓ ⏐ LX (21)
we see that
S × L α X L α Z : = S × L α X L Z × ( L Z ) d R O α ( Z ) d R → ≃ S × L α X L Z × ( L Z ) d R ( L Z ) d R × ( L X ) d R O α ( X ) d R → ≃ S × L α X O α ( X ) d R × ( L X ) d R L Z ← ≃ S × L α X L α X × L X L Z → ≃ S × L X L Z . \begin{aligned}
S \mathop{\times}_{\mathcal{L}^{\alpha}\mathcal{X}} \mathcal{L}^{\alpha}\mathcal{Z}
&:= S \mathop{\times}_{\mathcal{L}^{\alpha}\mathcal{X}} \mathcal{L}\mathcal{Z} \mathop{\times}_{(\mathcal{L}\mathcal{Z})_{\mathrm{dR}}} \mathcal{O}_{\alpha}(\mathcal{Z})_{\mathrm{dR}} \\
&\xrightarrow{\simeq} S \mathop{\times}_{\mathcal{L}^{\alpha}\mathcal{X}} \mathcal{L}\mathcal{Z} \mathop{\times}_{(\mathcal{L}\mathcal{Z})_{\mathrm{dR}}} (\mathcal{L}\mathcal{Z})_{\mathrm{dR}} \mathop{\times}_{(\mathcal{L}\mathcal{X})_{\mathrm{dR}}} \mathcal{O}_{\alpha}(\mathcal{X})_{\mathrm{dR}} \\
&\xrightarrow{\simeq} S \mathop{\times}_{\mathcal{L}^{\alpha}\mathcal{X}} \mathcal{O}_{\alpha}(\mathcal{X})_{\mathrm{dR}} \mathop{\times}_{(\mathcal{L}\mathcal{X})_{\mathrm{dR}}} \mathcal{L}\mathcal{Z} \\
&\xleftarrow{\simeq} S \mathop{\times}_{\mathcal{L}^{\alpha}\mathcal{X}} \mathcal{L}^{\alpha}\mathcal{X} \mathop{\times}_{\mathcal{L}\mathcal{X}} \mathcal{L}\mathcal{Z} \\
&\xrightarrow{\simeq} S \mathop{\times}_{\mathcal{L}\mathcal{X}} \mathcal{L}\mathcal{Z}.
\end{aligned} S × L α X L α Z := S × L α X LZ × ( LZ ) dR O α ( Z ) dR ≃ S × L α X LZ × ( LZ ) dR ( LZ ) dR × ( LX ) dR O α ( X ) dR ≃ S × L α X O α ( X ) dR × ( LX ) dR LZ ≃ S × L α X L α X × LX LZ ≃ S × LX LZ .
Unraveling the definitions, we have,
L α X × ( L α X ) d R ( L α Z ) d R ≃ ( L X × ( L X ) d R O α ( X ) d R ) × ( L X × ( L X ) d R O α ( X ) d R ) d R ( L X × ( L Z ) d R O α ( Z ) d R ) d R ≃ ( L X × ( L X ) d R O α ( X ) d R ) × ( L X ) d R × ( L X ) d R O α ( X ) d R ( ( L Z ) d R × ( L Z ) d R O α ( Z ) d R ) ≃ ( L X × ( L X ) d R ( L Z ) d R ) × ( L X ) d R × ( L X ) d R ( L Z ) d R ( O α ( X ) d R × O α ( X ) d R O α ( Z ) d R ) ≃ ( L X × ( L X ) d R ( L Z ) d R ) × ( L X × ( L X ) d R ( L Z ) d R ) d R ( O α ( X ) d R × O α ( Z ) d R O α ( Z ) d R ) d R = : L X Z ‾ × ( L X Z ) d R O α ( X ) O α ( Z ) d R ‾ . \begin{aligned}
\mathcal{L}^{\alpha}\mathcal{X} \mathop{\times}_{(\mathcal{L}^{\alpha}\mathcal{X})_{\mathrm{dR}}}(\mathcal{L}^{\alpha}\mathcal{Z})_{\mathrm{dR}}
&\simeq(\mathcal{L}\mathcal{X} \mathop{\times}_{(\mathcal{L}\mathcal{X})_{\mathrm{dR}}}\mathcal{O}_{\alpha}(\mathcal{X})_{\mathrm{dR}}) \mathop{\times}_{(\mathcal{L}\mathcal{X}\mathop{\times}_{(\mathcal{L}\mathcal{X})_{\mathrm{dR}}}\mathcal{O}_{\alpha}(\mathcal{X})_{\mathrm{dR}})_{\mathrm{dR}}}(\mathcal{L}\mathcal{X}\mathop{\times}_{(\mathcal{L}\mathcal{Z})_{\mathrm{dR}}}\mathcal{O}_{\alpha}(\mathcal{Z})_{\mathrm{dR}})_{\mathrm{dR}} \\
&\simeq(\mathcal{L}\mathcal{X}\mathop{\times}_{(\mathcal{L}\mathcal{X})_{\mathrm{dR}}}\mathcal{O}_{\alpha}(\mathcal{X})_{\mathrm{dR}}) \mathop{\times}_{(\mathcal{L}\mathcal{X})_{\mathrm{dR}}\mathop{\times}_{(\mathcal{L}\mathcal{X})_{\mathrm{dR}}}\mathcal{O}_{\alpha}(\mathcal{X})_{\mathrm{dR}}} ((\mathcal{L}\mathcal{Z})_{\mathrm{dR}}\mathop{\times}_{(\mathcal{L}\mathcal{Z})_{\mathrm{dR}}}\mathcal{O}_{\alpha}(\mathcal{Z})_{\mathrm{dR}}) \\
&\simeq(\mathcal{L}\mathcal{X}\mathop{\times}_{(\mathcal{L}\mathcal{X})_{\mathrm{dR}}}(\mathcal{L}\mathcal{Z})_{\mathrm{dR}}) \mathop{\times}_{(\mathcal{L}\mathcal{X})_{\mathrm{dR}}\mathop{\times}_{(\mathcal{L}\mathcal{X})_{\mathrm{dR}}}(\mathcal{L}\mathcal{Z})_{\mathrm{dR}}} (\mathcal{O}_{\alpha}(\mathcal{X})_{\mathrm{dR}}\mathop{\times}_{\mathcal{O}_{\alpha}(\mathcal{X})_{\mathrm{dR}}}\mathcal{O}_{\alpha}(\mathcal{Z})_{\mathrm{dR}}) \\
&\simeq(\mathcal{L}\mathcal{X}\mathop{\times}_{(\mathcal{L}\mathcal{X})_{\mathrm{dR}}}(\mathcal{L}\mathcal{Z})_{\mathrm{dR}}) \mathop{\times}_{(\mathcal{L}\mathcal{X}\mathop{\times}_{(\mathcal{L}\mathcal{X})_{\mathrm{dR}}}(\mathcal{L}\mathcal{Z})_{\mathrm{dR}})_{\mathrm{dR}}}(\mathcal{O}_{\alpha}(\mathcal{X})_{\mathrm{dR}}\mathop{\times}_{\mathcal{O}_{\alpha}(\mathcal{Z})_{\mathrm{dR}}}\mathcal{O}_{\alpha}(\mathcal{Z})_{\mathrm{dR}})_{\mathrm{dR}} \\
&=: \overline{\mathcal{L}\mathcal{X}_{\mathcal{Z}}} \mathop{\times}_{(\mathcal{L}\mathcal{X}_{\mathcal{Z}})_{\mathrm{dR}}}\overline{\mathcal{O}_{\alpha}(\mathcal{X})_{\mathcal{O}_{\alpha}(\mathcal{Z})_{\mathrm{dR}}}}.
\end{aligned} L α X × ( L α X ) dR ( L α Z ) dR ≃ ( LX × ( LX ) dR O α ( X ) dR ) × ( LX × ( LX ) dR O α ( X ) dR ) dR ( LX × ( LZ ) dR O α ( Z ) dR ) dR ≃ ( LX × ( LX ) dR O α ( X ) dR ) × ( LX ) dR × ( LX ) dR O α ( X ) dR (( LZ ) dR × ( LZ ) dR O α ( Z ) dR ) ≃ ( LX × ( LX ) dR ( LZ ) dR ) × ( LX ) dR × ( LX ) dR ( LZ ) dR ( O α ( X ) dR × O α ( X ) dR O α ( Z ) dR ) ≃ ( LX × ( LX ) dR ( LZ ) dR ) × ( LX × ( LX ) dR ( LZ ) dR ) dR ( O α ( X ) dR × O α ( Z ) dR O α ( Z ) dR ) dR =: L X Z × ( L X Z ) dR O α ( X ) O α ( Z ) dR .
On the other hand, by Lemma 3.57, L ( X Z ~ ) ≃ L X Z ‾ \mathcal{L}(\widetilde{\mathcal{X}_{\mathcal{Z}}}) \simeq\overline{\mathcal{L}\mathcal{X}_{\mathcal{Z}}} L ( X Z ) ≃ L X Z , so we have
( L α X ) L α Z ‾ ≃ L X Z ‾ × ( L X Z ) d R O α ( X ) O α ( Z ) d R ‾ ≃ L ( X Z ~ ) × L ( X Z ~ ) d R O α ( X ) O α ( Z ) d R ‾ . \overline{(\mathcal{L}^{\alpha}\mathcal{X})_{\mathcal{L}^{\alpha}\mathcal{Z}}} \simeq\overline{\mathcal{L}\mathcal{X}_{\mathcal{Z}}} \mathop{\times}_{(\mathcal{L}\mathcal{X}_{\mathcal{Z}})_{\mathrm{dR}}} \overline{\mathcal{O}_{\alpha}(\mathcal{X})_{\mathcal{O}_{\alpha}(\mathcal{Z})_{\mathrm{dR}}}} \simeq\mathcal{L}(\widetilde{\mathcal{X}_{\mathcal{Z}}}) \mathop{\times}_{\mathcal{L}(\widetilde{\mathcal{X}_{\mathcal{Z}}})_{\mathrm{dR}}} \overline{\mathcal{O}_{\alpha}(\mathcal{X})_{\mathcal{O}_{\alpha}(\mathcal{Z})_{\mathrm{dR}}}}. ( L α X ) L α Z ≃ L X Z × ( L X Z ) dR O α ( X ) O α ( Z ) dR ≃ L ( X Z ) × L ( X Z ) dR O α ( X ) O α ( Z ) dR .
It therefore suffices to show that O α ( X ) O α ( Z ) ‾ ≃ O α ( X Z ~ ) \overline{\mathcal{O}_{\alpha}(\mathcal{X})_{\mathcal{O}_{\alpha}(\mathcal{Z})}} \simeq\mathcal{O}_{\alpha}(\widetilde{\mathcal{X}_{\mathcal{Z}}}) O α ( X ) O α ( Z ) ≃ O α ( X Z ) . This follows from applying Lemma 3.49 to the Cartesian square (21) to obtain O α ( X ) O α ( Z ) ‾ ≃ X Z ~ × X O α ( X ) \overline{\mathcal{O}_{\alpha}(\mathcal{X})_{\mathcal{O}_{\alpha}(\mathcal{Z})}} \simeq\widetilde{\mathcal{X}_{\mathcal{Z}}} \mathop{\times}_{\mathcal{X}} \mathcal{O}_{\alpha}(\mathcal{X}) O α ( X ) O α ( Z ) ≃ X Z × X O α ( X ) , and then observing that the latter objects is exactly O α ( X Z ~ ) \mathcal{O}_{\alpha}(\widetilde{\mathcal{X}_{\mathcal{Z}}}) O α ( X Z ) since μ X Z ~ : X Z ~ → G / / G \mu_{\widetilde{\mathcal{X}_{\mathcal{Z}}}}:\widetilde{\mathcal{X}_{\mathcal{Z}}}\to G/\!/G μ X Z : X Z → G / / G factors as X Z ~ → X → μ X G / / G \widetilde{\mathcal{X}_{\mathcal{Z}}}\to\mathcal{X}\xrightarrow{\mu_{\mathcal{X}}}G/\!/G X Z → X μ X G / / G . □
Corollary 5.5. Suppose in the setting of Lemma 4.12 that X = U / G X=U/G X = U / G is a global quotient stack where U U U is a derived scheme, G G G is an algebraic group with reductive neutral component, and we denote ι : X Z → X \iota:X_{\mathcal{Z}}\to X ι : X Z → X . Then !-pullback along L u ι \mathcal{L}^{u}\iota L u ι induces an equivalence
IndCoh L u Z ( L u X ) t T a t e → ∼ IndCoh ( L u ( X Z ) ) t T a t e . \operatorname{IndCoh}_{\mathcal{L}^{u}\mathcal{Z}}(\mathcal{L}^{u}X)^{t\mathrm{Tate}}\xrightarrow{\sim}\operatorname{IndCoh}(\mathcal{L}^{u}(X_{\mathcal{Z}}))^{t\mathrm{Tate}}. IndCoh L u Z ( L u X ) t Tate ∼ IndCoh ( L u ( X Z ) ) t Tate .
Proof. We have natural isomorphisms
L u ( X Z ) ≃ ( L u X ) L u Z ≃ L X × ( L X ) d R O [ 1 ] ( Z ) d R \begin{aligned}
\mathcal{L}^{u}(X_{\mathcal{Z}})\simeq(\mathcal{L}^{u}X)_{\mathcal{L}^{u}\mathcal{Z}} \\
&\simeq\mathcal{L}X\times_{(LX)_{\mathrm{dR}}}\mathcal{O}_{[1]}(Z)_{\mathrm{dR}}
\end{aligned} L u ( X Z ) ≃ ( L u X ) L u Z ≃ L X × ( L X ) dR O [ 1 ] ( Z ) dR
: = ( L X ) O [ 1 ] ( Z ) , := (\mathcal{L}X)_{\mathcal{O}_{[1]}(Z)}, := ( L X ) O [ 1 ] ( Z ) ,
where we use nil-invariance of ( − ) d R (\mathord{-})_{\mathrm{dR}} ( − ) dR and the equivalence L u ≃ L [ 1 ] u \mathcal{L}^{u}\simeq\mathcal{L}^{u}_{[1]} L u ≃ L [ 1 ] u in the second line. Thus, the proof of Corollary 4.13 applies. □
A-trivializations
We recall the notions of A A A -trivialization for prestacks and categories equipped with an action of a group prestack.
Definition 5.6. ([[BZCHN23, Definition A.16]) Let G G G be a linear algebraic group with reductive neutral component acting on a prestack X X X , and let A ⊂ G A\subset G A ⊂ G be a normal subgroup with G ′ = G / A G'=G/A G ′ = G / A . An A A A -trivialization (of the G G G -action on X X X ) is a G ′ G' G ′ -action on X X X and an equivalence
X / G ≃ X / G ′ × B G ′ B G . X/G\simeq X/G'\times_{BG'}BG. X / G ≃ X / G ′ × B G ′ B G .
When A = ⟨ α ⟩ A=\langle\alpha\rangle A = ⟨ α ⟩ , for α ∈ Z ( G ) \alpha\in Z(G) α ∈ Z ( G ) a central element, we call an A A A -trivialization an α \alpha α -trivialization.
Lemma 5.7. Let G G G be an affine algebraic group, and let A ⊂ G A\subset G A ⊂ G be a closed central subgroup.33 Denote the quotient G ′ = G / A G'=G/A G ′ = G / A . Suppose that X X X is a derived stack equipped with a G G G -action and an A A A -trivialization X / G ≃ X / G ′ × B G ′ B G X/G\simeq X/G'\times_{BG'}BG X / G ≃ X / G ′ × B G ′ B G . Then L ( X / G ) \mathcal{L}(X/G) L ( X / G ) has a canonical A A A -trivialization coming from that on X / G X/G X / G .
Proof. Since G ′ = G / A G'=G/A G ′ = G / A is the quotient of G G G by a central subgroup, there is an equivalence G / G ≃ G / G ′ × B G ′ B G G/G\simeq G/G'\times_{BG'}BG G / G ≃ G / G ′ × B G ′ B G , which gives an A A A -trivialization of L ( B G ) ≃ G / G \mathcal{L}(BG)\simeq G/G L ( B G ) ≃ G / G . This A A A -trivialization induces an A A A -trivialization of L ( X / G ) \mathcal{L}(X/G) L ( X / G ) as follows:
( X × X × X ( X × G ) ) / G → ∼ ( X / G ) × X / G × B G X / G ( X / G × B G G / G ) ≃ ( X / G ′ × B G ′ B G ) × ( X / G ′ × B G ′ X / G ′ ) × B G ′ B G ( X / G ′ × B G ′ G / G ′ ) × B G ′ B G \begin{aligned}
(X\times_{X\times X}(X\times G))/G\xrightarrow{\sim}(X/G)\times_{X/G\times_{BG}X/G}(X/G\times_{BG}G/G) \\
&\simeq\left(X/G'\times_{BG'}BG\right)\times_{(X/G'\times_{BG'}X/G')\times_{BG'}BG}\left(X/G'\times_{BG'}G/G'\right)\times_{BG'}BG
\end{aligned} ( X × X × X ( X × G )) / G ∼ ( X / G ) × X / G × B G X / G ( X / G × B G G / G ) ≃ ( X / G ′ × B G ′ B G ) × ( X / G ′ × B G ′ X / G ′ ) × B G ′ B G ( X / G ′ × B G ′ G / G ′ ) × B G ′ B G
→ ∼ ( X / G ′ × X / G ′ × B G ′ X / G ′ X / G ′ × B G ′ G / G ′ ) × B G ′ B G \xrightarrow{\sim}\left(X/G'\times_{X/G'\times_{BG'}X/G'}X/G'\times_{BG'}G/G'\right)\times_{BG'}BG ∼ ( X / G ′ × X / G ′ × B G ′ X / G ′ X / G ′ × B G ′ G / G ′ ) × B G ′ B G
← ∼ ( X × X × X X × G ) / G ′ × B G ′ B G . \xleftarrow{\sim}(X\times_{X\times X}X\times G)/G'\times_{BG'}BG. ∼ ( X × X × X X × G ) / G ′ × B G ′ B G .
Let U ⊂ G U\subset G U ⊂ G denote the unipotent cone of G G G . There is a similar A A A -trivialization on U / G U/G U / G from the equivalence U / G ≃ U / G ′ × B G ′ B G U/G\simeq U/G'\times_{BG'}BG U / G ≃ U / G ′ × B G ′ B G . Combining this with the A A A -trivialization of L ( B G ) \mathcal{L}(BG) L ( B G ) identified above also induces an A A A -trivialization of L u ( B G ) \mathcal{L}^{u}(BG) L u ( B G ) :
L ( B G ) × L ( B G ) d R ( U / G ) d R ≃ ( G / G ′ × B G ′ B G ) × ( G / G ′ × B G ′ B G ) d R ( U / G ′ × B G ′ B G ) d R ≃ ( G / G ′ × B G ′ B G ) × ( G / G ′ ) d R × ( B G ′ ) d R B G d R ( ( U / G ′ ) d R × B G d R ′ B G d R ) ≃ ( G / G ′ × ( G / G ′ ) d R ( U / G ′ ) d R ) × ( B G ′ × B G d R ′ B G d R ′ ) ( B G × B G d R B G d R ) ≃ ( G / G ′ × ( G / G ′ ) d R ( U / G ′ ) d R ) × B G ′ B G ≃ ( G / G ′ × G d R / G d R ′ U d R / G d R ′ ) × B G ′ B G ≃ ( G × G d R U d R ) / ( G ′ × G d R ′ G d R ′ ) × B G ′ B G ≃ G U ^ / G ′ × B G ′ B G . \begin{aligned}
\mathcal{L}(BG) \mathop{\times}_{\mathcal{L}(BG)_{\mathrm{dR}}} (U/G)_{\mathrm{dR}} \simeq(G/G' \mathop{\times}_{BG'} BG) \mathop{\times}_{(G/G' \times_{BG'} BG)_{\mathrm{dR}}} (U/G' \mathop{\times}_{BG'} BG)_{\mathrm{dR}} \\
&\simeq(G/G' \mathop{\times}_{BG'} BG) \mathop{\times}_{(G/G')_{\mathrm{dR}} \times(BG')_{\mathrm{dR}}} BG_{\mathrm{dR}} ((U/G')_{\mathrm{dR}} \mathop{\times}_{BG'_{\mathrm{dR}}} BG_{\mathrm{dR}}) \\
&\simeq(G/G' \mathop{\times}_{(G/G')_{\mathrm{dR}}} (U/G')_{\mathrm{dR}}) \mathop{\times}_{(BG' \times_{BG'_{\mathrm{dR}}} BG'_{\mathrm{dR}})} (BG \mathop{\times}_{BG_{\mathrm{dR}}} BG_{\mathrm{dR}}) \\
&\simeq(G/G' \mathop{\times}_{(G/G')_{\mathrm{dR}}} (U/G')_{\mathrm{dR}}) \mathop{\times}_{BG'} BG \\
&\simeq(G/G' \mathop{\times}_{G_{\mathrm{dR}}/G'_{\mathrm{dR}}} U_{\mathrm{dR}}/G'_{\mathrm{dR}}) \mathop{\times}_{BG'} BG \\
&\simeq(G \mathop{\times}_{G_{\mathrm{dR}}} U_{\mathrm{dR}})/(G' \mathop{\times}_{G'_{\mathrm{dR}}} G'_{\mathrm{dR}}) \mathop{\times}_{BG'} BG \\
&\simeq\widehat{G_U}/G' \mathop{\times}_{BG'} BG.
\end{aligned} L ( B G ) × L ( B G ) dR ( U / G ) dR ≃ ( G / G ′ × B G ′ B G ) × ( G / G ′ × B G ′ B G ) dR ( U / G ′ × B G ′ B G ) dR ≃ ( G / G ′ × B G ′ B G ) × ( G / G ′ ) dR × ( B G ′ ) dR B G dR (( U / G ′ ) dR × B G dR ′ B G dR ) ≃ ( G / G ′ × ( G / G ′ ) dR ( U / G ′ ) dR ) × ( B G ′ × B G dR ′ B G dR ′ ) ( B G × B G dR B G dR ) ≃ ( G / G ′ × ( G / G ′ ) dR ( U / G ′ ) dR ) × B G ′ B G ≃ ( G / G ′ × G dR / G dR ′ U dR / G dR ′ ) × B G ′ B G ≃ ( G × G dR U dR ) / ( G ′ × G dR ′ G dR ′ ) × B G ′ B G ≃ G U / G ′ × B G ′ B G .
Finally, using the A A A -trivializations on L u ( B G ) \mathcal{L}^{u}(BG) L u ( B G ) , L ( B G ) \mathcal{L}(BG) L ( B G ) , and L ( X / G ) \mathcal{L}(X/G) L ( X / G ) above, we obtain a canonical A A A -trivialization of L u ( X / G ) \mathcal{L}^{u}(X/G) L u ( X / G ) . Indeed, applying L u : = Map ( B G a , − ) \mathcal{L}^{u}:=\operatorname{Map}(BG_a,-) L u := Map ( B G a , − ) , which preserves limits, to the A A A -trivialization of X / G X/G X / G , we obtain
L u ( X / G ) ≃ L u ( X / G ′ ) × L u ( B G ′ ) L u ( B G ) ≃ ( L u ( X / G ′ ) × L u ( B G ′ ) G U ^ / G ′ ) × B G ′ B G . \begin{aligned}
\mathcal{L}^{u}(X/G) \simeq\mathcal{L}^{u}(X/G') \mathop{\times}_{\mathcal{L}^{u}(BG')} \mathcal{L}^{u}(BG) \\
&\simeq\left(\mathcal{L}^{u}(X/G') \mathop{\times}_{\mathcal{L}^{u}(BG')} \widehat{G_U}/G'\right) \mathop{\times}_{BG'} BG.
\end{aligned} L u ( X / G ) ≃ L u ( X / G ′ ) × L u ( B G ′ ) L u ( B G ) ≃ ( L u ( X / G ′ ) × L u ( B G ′ ) G U / G ′ ) × B G ′ B G .
We claim that this is an A A A -trivialization. Indeed, denoting
W : = L u ( X / G ) × L u ( B G ′ ) G U ^ / G ′ , W := \mathcal{L}^{u}(X/G) \mathop{\times}_{\mathcal{L}^{u}(BG')} \widehat{G_U}/G', W := L u ( X / G ) × L u ( B G ′ ) G U / G ′ ,
we need to show that W ≃ Y / G ′ W \simeq Y/G' W ≃ Y / G ′ and L u ( X / G ) ≃ Y / G \mathcal{L}^{u}(X/G) \simeq Y/G L u ( X / G ) ≃ Y / G for some stack Y Y Y . We claim this holds for Y : = W × B G ′ p t Y := W \mathop{\times}_{BG'} \mathrm{pt} Y := W × B G ′ pt . To see this, observe first that the base change of W × B G ′ B G W \mathop{\times}_{BG'} BG W × B G ′ B G along p t → B G \mathrm{pt} \to BG pt → B G gives a G G G -torsor
W × B G ′ p t ⟶ W × B G ′ B G ≃ L u ( X / G ) . W \mathop{\times}_{BG'} \mathrm{pt} \longrightarrow W \mathop{\times}_{BG'} BG \simeq\mathcal{L}^{u}(X/G). W × B G ′ pt ⟶ W × B G ′ B G ≃ L u ( X / G ) .
Thus, we see that
L u ( X / G ) ≃ ( W × B G ′ p t ) / G . \mathcal{L}^{u}(X/G) \simeq(W \mathop{\times}_{BG'} \mathrm{pt})/G. L u ( X / G ) ≃ ( W × B G ′ pt ) / G .
On the other hand, W × B G ′ p t W \mathop{\times}_{BG'} \mathrm{pt} W × B G ′ pt is the base change of W W W along the universal G ′ G' G ′ -torsor p t → B G ′ \mathrm{pt} \to BG' pt → B G ′ ; hence,
W ≃ ( W × B G ′ p t ) / G ′ W \simeq(W \mathop{\times}_{BG'} \mathrm{pt})/G' W ≃ ( W × B G ′ pt ) / G ′
as desired. □
5.2.1. Canonical A A A -trivialization on the fixed point locus. If A ⊂ G A \subset G A ⊂ G is a central subgroup, the centralizer G ( A ) G(A) G ( A ) clearly acts on the fixed point locus X A X^A X A of the induced A A A -action on X X X , and [BZCHN23, Proposition A.17]34 states that X A X^A X A possesses a canonical A A A -trivialization. In the sequel, it will be important to know that, for a G G G -equivariant map of prestacks f : X → Y f:X \to Y f : X → Y , the formal completion along the induced map on fixed points, f A : X A → Y A f^A:X^A \to Y^A f A : X A → Y A also possesses a canonical A A A -trivialization. This follows from [BZCHN23, Proposition A.17] and the following lemma.
Lemma 5.8. Let G G G be any group prestack, and let f : X → Y f:X \to Y f : X → Y be any G G G -equivariant map of prestacks. Let A ⊂ G A \subset G A ⊂ G be a reduced central subgroup. Then there is a natural equivalence ( Y X ^ ) A ≃ ( Y A ) X A ^ (\widehat{Y_X})^A \simeq\widehat{(Y^A)_{X^A}} ( Y X ) A ≃ ( Y A ) X A .35
Proof. First note that the G G G -equivariant structure on f : X → Y f : X \to Y f : X → Y restricts to a canonical A A A -equivariant, so we obtain map of fixed points, X A → U A X^{A} \to U^{A} X A → U A .
Recall the Cartesian diagram defined the fixed point locus of Y ^ X \widehat{Y}_{X} Y X ,
( Y ^ X ) A → Map ( B A , Y ^ X / A ) ↓ ↓ { id B A } → Map ( B A , B A ) . (22) \begin{CD}
(\widehat{Y}_{X})^{A} @>>> \operatorname{Map}(BA,\widehat{Y}_{X}/A) \\
@VVV @VVV \\
\{\operatorname{id}_{BA}\} @>>> \operatorname{Map}(BA,BA). \tag*{(22)}
\end{CD} ( Y X ) A ↓ ⏐ { id B A } Map ( B A , Y X / A ) ↓ ⏐ Map ( B A , B A ) . (22)
By definition, Y ^ X / A = ( Y × Y d R X d R ) / A \widehat{Y}_{X}/A = (Y \mathbin{\times_{Y_{\mathrm{dR}}}} X_{\mathrm{dR}})/A Y X / A = ( Y × Y dR X dR ) / A . Since the quotient by the group action of A A A is a sifted colimit, it commutes with finite products, so we obtain ( Y × Y d R X d R ) / A → ≃ Y / A × Y d R / A X d R / A (Y \mathbin{\times_{Y_{\mathrm{dR}}}} X_{\mathrm{dR}})/A \xrightarrow{\simeq} Y/A \mathbin{\times_{Y_{\mathrm{dR}}/A}} X_{\mathrm{dR}}/A ( Y × Y dR X dR ) / A ≃ Y / A × Y dR / A X dR / A . Mapping stacks preserve limits in the second argument, so we therefore have the Cartesian diagram
Map ( B A , Y ^ X / A ) → Map ( B A , Y / A ) ↓ ↓ Map ( B A , ( X / A ) d R ) → Map ( B A , ( Y / A ) d R ) . \begin{CD}
\operatorname{Map}(BA,\widehat{Y}_{X}/A) @>>> \operatorname{Map}(BA,Y/A) \\
@VVV @VVV \\
\operatorname{Map}(BA,(X/A)_{\mathrm{dR}}) @>>> \operatorname{Map}(BA,(Y/A)_{\mathrm{dR}}).
\end{CD} Map ( B A , Y X / A ) ↓ ⏐ Map ( B A , ( X / A ) dR ) Map ( B A , Y / A ) ↓ ⏐ Map ( B A , ( Y / A ) dR ) .
Limits commute with one another, so we may evaluate the top-left term of the fixed point diagram (22) as the limit of the cospan whose terms are the top-left terms of the pullback diagrams
S A → Map ( B A , S / A ) ↓ ↓ { id B A } → Map ( B A , B A ) . \begin{CD}
S^{A} @>>> \operatorname{Map}(BA,S/A) \\
@VVV @VVV \\
\{\operatorname{id}_{BA}\} @>>> \operatorname{Map}(BA,BA).
\end{CD} S A ↓ ⏐ { id B A } Map ( B A , S / A ) ↓ ⏐ Map ( B A , B A ) .
for S = Y S=Y S = Y , Y d R Y_{\mathrm{dR}} Y dR , and X d R X_{\mathrm{dR}} X dR . This gives precisely Y ^ X A → ≃ Y A × ( Y d R ) A ( X d R ) A \widehat{Y}_{X}^{A} \xrightarrow{\simeq} Y^{A} \mathbin{\times_{(Y_{\mathrm{dR}})^{A}}} (X_{\mathrm{dR}})^{A} Y X A ≃ Y A × ( Y dR ) A ( X dR ) A .
On the other hand, there is a natural map ( Y A ) ( X A ) = Y A × ( Y A ) d R ( X A ) d R → Y A × ( Y d R ) A ( X d R ) A (Y^{A})_{(X^{A})} = Y^{A} \mathbin{\times_{(Y^{A})_{\mathrm{dR}}}} (X^{A})_{\mathrm{dR}} \to Y^{A} \mathbin{\times_{(Y_{\mathrm{dR}})^{A}}} (X_{\mathrm{dR}})^{A} ( Y A ) ( X A ) = Y A × ( Y A ) dR ( X A ) dR → Y A × ( Y dR ) A ( X dR ) A induced by the universal morphism ( S A ) d R → ( S d R ) A (S^{A})_{\mathrm{dR}} \to(S_{\mathrm{dR}})^{A} ( S A ) dR → ( S dR ) A that exists for any stacks S S S with A A A -action, so it remains to show that ( S A ) d R → ≃ ( S d R ) A (S^{A})_{\mathrm{dR}} \xrightarrow{\simeq} (S_{\mathrm{dR}})^{A} ( S A ) dR ≃ ( S dR ) A . By definition, for a test affine T T T ,
S A ( T ) ≃ lim [ n ] ∈ Δ S ( T × A n ) . S^{A}(T) \simeq\lim_{[n]\in\Delta} S(T \times A^{n}). S A ( T ) ≃ [ n ] ∈ Δ lim S ( T × A n ) .
Now, A A A is reduced; hence, ( T × A n ) r e d ≃ T r e d × A n (T \times A^{n})_{\mathrm{red}} \simeq T_{\mathrm{red}} \times A^{n} ( T × A n ) red ≃ T red × A n . Thus, we compute
( S d R ) A ( T ) ≃ lim [ n ] ∈ Δ S d R ( T × A n ) ≃ lim [ n ] ∈ Δ S ( T r e d × A n ) ≃ S A ( T r e d ) ≃ ( S A ) d R ( T ) . \begin{aligned}
(S_{\mathrm{dR}})^{A}(T) &\simeq\lim_{[n]\in\Delta} S_{\mathrm{dR}}(T \times A^{n}) \\
&\simeq\lim_{[n]\in\Delta} S(T_{\mathrm{red}} \times A^{n}) \\
&\simeq S^{A}(T_{\mathrm{red}}) \\
&\simeq(S^{A})_{\mathrm{dR}}(T).
\end{aligned} ( S dR ) A ( T ) ≃ [ n ] ∈ Δ lim S dR ( T × A n ) ≃ [ n ] ∈ Δ lim S ( T red × A n ) ≃ S A ( T red ) ≃ ( S A ) dR ( T ) .
These isomorphisms are clearly natural in T T T and compatible with the group operations and induced action. Hence, we find
( S A ) d R ⟶ ( S d R ) A (S^{A})_{\mathrm{dR}} \longrightarrow(S_{\mathrm{dR}})^{A} ( S A ) dR ⟶ ( S dR ) A
is an isomorphism as desired. □
In particular, Lemma 5.8 and [BZCHN23, Proposition A.17] together imply that the formal completion along a map of fixed points has a canonical A A A -trivialization.
Corollary 5.10. Let G G G be a group prestack, and let X → Y X \to Y X → Y be a G G G -equivariant map of prestacks equipped with G G G -actions. Let A ⊂ G A \subset G A ⊂ G be a reduced central subgroup. Then Y A × X A X A Y^A \mathop{\times}_{X^A} X^A Y A × X A X A has a canonical A A A -trivialization.
α \alpha α -shifted action on the unipotent loop space
In addition to the standard S 1 S^1 S 1 -action on L α ( − ) \mathcal{L}^{\alpha}(-) L α ( − ) given by loop rotation, there is the so-called α \alpha α -shifted S 1 S^1 S 1 -action, which we describe below. In order to do so, we first recall the following definition.
If X X X is a prestack equipped with an action of a group prestack G G G , and α ∈ G \alpha\in G α ∈ G is a central element, then we obtain a canonical α \alpha α -trivialization,
X α / G ( α ) ≃ ( X α / G ′ ( α ) ) × B G ′ ( α ) B G ( α ) , (23) X^{\alpha}/G(\alpha) \simeq\left(X^{\alpha}/G'(\alpha)\right) \mathop{\times}_{BG'(\alpha)} BG(\alpha), \tag*{(23)} X α / G ( α ) ≃ ( X α / G ′ ( α ) ) × B G ′ ( α ) B G ( α ) , (23)
as a special case of [BZCHN23, Proposition A.17], where G ( α ) G(\alpha) G ( α ) here denotes the centralizer of α \alpha α , and G ′ ( α ) : = G ( α ) / ⟨ α ⟩ G'(\alpha) := G(\alpha)/\langle\alpha\rangle G ′ ( α ) := G ( α ) / ⟨ α ⟩ denotes the quotient. The α \alpha α -shifted loop rotation action on L ( X α / G ( α ) ) \mathcal{L}(X^{\alpha}/G(\alpha)) L ( X α / G ( α )) (resp. L α ( X α / G ( α ) ) \mathcal{L}^{\alpha}(X^{\alpha}/G(\alpha)) L α ( X α / G ( α )) and L α ( X α / G ( α ) ) ) \mathcal{L}^{\alpha}(X^{\alpha}/G(\alpha))) L α ( X α / G ( α ))) is defined in the following construction:
Construction 5.12. There is an S 1 S^1 S 1 -action on B G ( α ) BG(\alpha) B G ( α ) , denoted by B ( s h α ) B(sh_{\alpha}) B ( s h α ) , given by the delooping of the Z \mathbb{Z} Z -action s h α sh_{\alpha} s h α on G ( α ) G(\alpha) G ( α ) given by multiplication by α \alpha α . This induces an S 1 S^1 S 1 -action on X α / G ( α ) X^{\alpha}/G(\alpha) X α / G ( α ) , which we denote by σ α \sigma_{\alpha} σ α via its canonical α \alpha α -trivialization: σ α \sigma_{\alpha} σ α is the diagonal to the S 1 × S 1 S^1 \times S^1 S 1 × S 1 -action obtained from the trivial action on the first factor and the B ( s h α ) B(sh_{\alpha}) B ( s h α ) -action on the second factor in the product on the right-hand side of (23) . Composing this action σ α : B S 1 → PreStk \sigma_{\alpha} : BS^1 \to\operatorname{PreStk} σ α : B S 1 → PreStk with the loop space functor L ( − ) : PreStk → PreStk \mathcal{L}(-) : \operatorname{PreStk} \to\operatorname{PreStk} L ( − ) : PreStk → PreStk , we obtain an S 1 S^1 S 1 -action on L ( X α / G ( α ) ) \mathcal{L}(X^{\alpha}/G(\alpha)) L ( X α / G ( α )) called the α \alpha α -twisted action on L ( X α / G ( α ) ) \mathcal{L}(X^{\alpha}/G(\alpha)) L ( X α / G ( α )) . Since the ordinary S 1 S^1 S 1 -action ρ \rho ρ on L ( X α / G ( α ) ) \mathcal{L}(X^{\alpha}/G(\alpha)) L ( X α / G ( α )) given by loop rotation commutes with L ( σ α ) \mathcal{L}(\sigma_{\alpha}) L ( σ α ) , the α \alpha α -shifted loop rotation action is defined to be the diagonal to the S 1 × S 1 S^1 \times S^1 S 1 × S 1 -action ρ × L ( σ α ) \rho\times\mathcal{L}(\sigma_{\alpha}) ρ × L ( σ α ) . We denote this final S 1 S^1 S 1 -action by ρ ( α ) \rho(\alpha) ρ ( α ) .
One may also take the functor L α ( − ) : PreStk / B G → PreStk / B G \mathcal{L}^{\alpha}(-) : \operatorname{PreStk}/BG \to\operatorname{PreStk}/BG L α ( − ) : PreStk / B G → PreStk / B G in place of the usual loop space functor in the above construction to obtain the α \alpha α -shifted loop rotation action on L α ( X α / G ( α ) ) \mathcal{L}^{\alpha}(X^{\alpha}/G(\alpha)) L α ( X α / G ( α )) . This action is also denoted by ρ ( α ) \rho(\alpha) ρ ( α ) .
Fixed point and shifted localization
Let us now specialize to the case when X X X is a derived scheme equipped with an action of a reductive group G G G . For a chosen semisimple element α ∈ G \alpha\in G α ∈ G , let X α X^{\alpha} X α denote the homotopy fixed points of the action of ⟨ α ⟩ \langle\alpha\rangle ⟨ α ⟩ , and let G ( α ) ⊂ G G(\alpha) \subset G G ( α ) ⊂ G denote the centralizer of α \alpha α . Clearly G ( α ) G(\alpha) G ( α ) acts on X α X^{\alpha} X α , and there is a canonical map, X α / G ( α ) → X / G X^{\alpha}/G(\alpha) \to X/G X α / G ( α ) → X / G . By [BZCHN23, Theorem 2.9], if X X X is locally Noetherian, the above map induces an S 1 S^1 S 1 -equivariant equivalence on α \alpha α -potent loop spaces,
ρ α u : L α ( X α / G ( α ) ) → ≃ L α ( X / G ) \rho_{\alpha}^{u} : \mathcal{L}^{\alpha}(X^{\alpha}/G(\alpha)) \xrightarrow{\simeq} \mathcal{L}^{\alpha}(X/G) ρ α u : L α ( X α / G ( α )) ≃ L α ( X / G )
There is an automorphism s h α : L ( X α / G ( α ) ) → L ( X α / G ( α ) ) sh_{\alpha} : \mathcal{L}(X^{\alpha}/G(\alpha)) \to\mathcal{L}(X^{\alpha}/G(\alpha)) s h α : L ( X α / G ( α )) → L ( X α / G ( α )) defined as follows. From the canonical α \alpha α -trivialization on X α / G ( α ) X^{\alpha}/G(\alpha) X α / G ( α ) we obtain an induced product decomposition L ( X α / G ( α ) ) ≃ L ( X α / G ′ ( α ) ) × G ′ ( α ) / G ′ ( α ) G ( α ) / G ( α ) \mathcal{L}(X^{\alpha}/G(\alpha)) \simeq\mathcal{L}(X^{\alpha}/G'(\alpha)) \times_{G'(\alpha)/G'(\alpha)} G(\alpha)/G(\alpha) L ( X α / G ( α )) ≃ L ( X α / G ′ ( α )) × G ′ ( α ) / G ′ ( α ) G ( α ) / G ( α ) , where G ′ ( α ) G'(\alpha) G ′ ( α ) is the quotient G ( α ) / ⟨ α ⟩ G(\alpha)/\langle\alpha\rangle G ( α ) / ⟨ α ⟩ . Let s h α sh_{\alpha} s h α be the automorphism defined on this product by multiplication by α \alpha α on the G ( α ) / G ( α ) G(\alpha)/G(\alpha) G ( α ) / G ( α ) factor, and the identity on the other factors. This automorphism restricts to a map,
s h α : L u ( X α / G ( α ) ) → L α ( X α / G ( α ) ) sh_{\alpha} : \mathcal{L}^{u}(X^{\alpha}/G(\alpha)) \to\mathcal{L}^{\alpha}(X^{\alpha}/G(\alpha)) s h α : L u ( X α / G ( α )) → L α ( X α / G ( α ))
which is S 1 S^1 S 1 -equivariant for the α \alpha α -shifted loop rotation action (defined in Construction 5.12) on the source and the ordinary loop rotation action on the right. The shifted localization map,
s ℓ α u : L u ( X α / G ( α ) ) → L α ( X / G ) , s\ell_{\alpha}^{u} : \mathcal{L}^{u}(X^{\alpha}/G(\alpha)) \to\mathcal{L}^{\alpha}(X/G), s ℓ α u : L u ( X α / G ( α )) → L α ( X / G ) ,
defined to be the composition ρ α u ∘ s h α \rho_{\alpha}^{u} \circ sh_{\alpha} ρ α u ∘ s h α , is an S 1 S^1 S 1 -equivariant equivalence for the ρ ( α ) \rho(\alpha) ρ ( α ) action on the source and the ρ \rho ρ action on the target by [BZCHN23, Corollary A.18].
In what follows, we will require an extension of the shifted localization results of [BZCHN23] to the case when X X X is the formal completion of a map of derived schemes, which is not a derived scheme itself in general.
Lemma 5.13. Let G G G be a linear algebraic group with reductive neutral component acting on locally Noetherian derived schemes X X X and Y Y Y , and let α ∈ G \alpha\in G α ∈ G be a semisimple element. If f : X → Y f:X \to Y f : X → Y is a G G G -equivariant map, then the shifted localization map
s ℓ α α : L u ( ( Y X α ) ^ / G ( α ) ) ⟶ L α ( Y X ^ / G ) s\ell^\alpha_\alpha:\mathcal{L}^{u}\left(\widehat{(Y_X^\alpha)}/G(\alpha)\right)\longrightarrow\mathcal{L}^{\alpha}\left(\widehat{Y_X}/G\right) s ℓ α α : L u ( ( Y X α ) / G ( α ) ) ⟶ L α ( Y X / G ) is an equivalence.
Proof. By construction, the shifted localization map s ℓ α α : L u ( U α / G ( α ) ) → L α ( U / G ) s\ell^\alpha_\alpha:\mathcal{L}^{u}(U^\alpha/G(\alpha))\to\mathcal{L}^{\alpha}(U/G) s ℓ α α : L u ( U α / G ( α )) → L α ( U / G ) is natural along maps of derived G G G -schemes U U U . Since X X X and Y Y Y themselves satisfy the hypotheses of [BZCHN23, Corollary A.18], this means we obtain a commutative diagram of S 1 S^1 S 1 -equivariant maps,
L u ( X α / G ( α ) ) → ∼ L α ( X / G ) s ℓ α α ( f α / G ( α ) ) ↓ ↓ s ℓ α α ( f / G ) L u ( Y α / G ( α ) ) → ∼ L α ( Y / G ) . \begin{CD}
\mathcal{L}^{u}(X^\alpha/G(\alpha)) @>{\sim}>> \mathcal{L}^{\alpha}(X/G) \\
@V{s\ell^\alpha_\alpha(f^\alpha/G(\alpha))}VV @VV{s\ell^\alpha_\alpha(f/G)}V \\
\mathcal{L}^{u}(Y^\alpha/G(\alpha)) @>{\sim}>> \mathcal{L}^{\alpha}(Y/G).
\end{CD} L u ( X α / G ( α )) s ℓ α α ( f α / G ( α )) ↓ ⏐ L u ( Y α / G ( α )) ∼ ∼ L α ( X / G ) ↓ ⏐ s ℓ α α ( f / G ) L α ( Y / G ) . where f α : X α → Y α f^\alpha:X^\alpha\to Y^\alpha f α : X α → Y α is the induced map on fixed points. From this, we deduce that the induced map
L u ( Y α / G ( α ) ) L u ( X α / G ( α ) ) ⟶ L α ( Y / G ) L α ( X / G ) (24) \mathcal{L}^{u}(Y^\alpha/G(\alpha))_{\mathcal{L}^{u}(X^\alpha/G(\alpha))}\longrightarrow\mathcal{L}^{\alpha}(Y/G)_{\mathcal{L}^{\alpha}(X/G)} \tag*{(24)} L u ( Y α / G ( α ) ) L u ( X α / G ( α )) ⟶ L α ( Y / G ) L α ( X / G ) (24) is an equivalence. On the other hand,
by Lemma 5.8 , Lemma 5.1 , Lemma 3.48 , and the fact that L u = L [ 1 ] u \mathcal{L}^{u}=\mathcal{L}^{u}_{[1]} L u = L [ 1 ] u is a limit, hence commutes with limits, we see—by a direct fiber product calculation—that the left- and right-hand side of (24) are equivalent to
L u ( Y α / G ( α ) ^ X α / G ( α ) ) and L α ( Y / G ^ X / G ) , \mathcal{L}^{u}\left(\widehat{Y^\alpha/G(\alpha)}_{X^\alpha/G(\alpha)}\right)
\qquad\text{and}\qquad
\mathcal{L}^{\alpha}\left(\widehat{Y/G}_{X/G}\right), L u ( Y α / G ( α ) X α / G ( α ) ) and L α ( Y / G X / G ) , respectively;
by Lemma 3.48 , Y α / G ( α ) ^ X α / G ( α ) ≃ Y X α α / G ( α ) ^ \widehat{Y^\alpha/G(\alpha)}_{X^\alpha/G(\alpha)}\simeq\widehat{Y^\alpha_{X^\alpha}/G(\alpha)} Y α / G ( α ) X α / G ( α ) ≃ Y X α α / G ( α ) and Y / G ^ X / G ≃ Y X / G ^ \widehat{Y/G}_{X/G}\simeq\widehat{Y_X/G} Y / G X / G ≃ Y X / G ;
and by Lemma 5.8 , Y X α α ^ ≃ ( Y X ^ ) α \widehat{Y^\alpha_{X^\alpha}}\simeq\left(\widehat{Y_X}\right)^\alpha Y X α α ≃ ( Y X ) α .
All together, the equivalence (24) becomes the equivalence
s ℓ α α : L u ( Y X ^ α / G ( α ) ) → ∼ L α ( Y X ^ / G ) , s\ell^\alpha_\alpha:\mathcal{L}^{u}\left(\widehat{Y_X}^{\alpha}/G(\alpha)\right)\xrightarrow{\sim}\mathcal{L}^{\alpha}\left(\widehat{Y_X}/G\right), s ℓ α α : L u ( Y X α / G ( α ) ) ∼ L α ( Y X / G ) , as desired. □
There is a corresponding notion of A A A -trivialization for G G G -categories, which we recall below.
Definition 5.15. If G G G is an affine algebraic group and A ⊂ Z ( G ) A\subset Z(G) A ⊂ Z ( G ) is a central subgroup, then an A A A -trivialization of a QCoh ( B G ) \operatorname{QCoh}(BG) QCoh ( B G ) -module category C \mathcal{C} C is a category C ′ \mathcal{C}' C ′ and an equivalence
C ≃ C ′ ⊗ QCoh ( B ( G / A ) ) QCoh ( B G ) . \mathcal{C}\simeq\mathcal{C}'\otimes_{\operatorname{QCoh}(B(G/A))}\operatorname{QCoh}(BG). C ≃ C ′ ⊗ QCoh ( B ( G / A )) QCoh ( B G ) . Then the category of A A A -trivial objects of C \mathcal{C} C is denoted C A \mathcal{C}_A C A , which is defined to be the essential image in C \mathcal{C} C of the natural functor C ′ → C \mathcal{C}'\to\mathcal{C} C ′ → C . When A = ⟨ α ⟩ A=\langle\alpha\rangle A = ⟨ α ⟩ for α ∈ Z ( G ) \alpha\in Z(G) α ∈ Z ( G ) , we write C α : = C A \mathcal{C}_\alpha:=\mathcal{C}_A C α := C A .
Remark 5.16. If X X X is a prestack with G G G -action, an A A A -trivialization of X X X induces a A A A -trivialization of the G G G -category QCoh ( X / G ) \operatorname{QCoh}(X/G) QCoh ( X / G ) .
We now show that Tate invariants with respect to the α \alpha α -twisted loop rotation action on L u ( X α / G ( α ) ) \mathcal{L}^{u}(X^\alpha/G(\alpha)) L u ( X α / G ( α )) kills all objects which are not α \alpha α -trivial. In fact, this already happens at the level of S 1 S^1 S 1 -invariants, as is shown in the proof of the following result.
Proposition 5.17. Let G G G be a reductive affine algebraic group and X X X a laft prestack with G G G -action, and let α ∈ Z ( G ) \alpha\in Z(G) α ∈ Z ( G ) be a central semisimple element. Suppose further that X / G X/G X / G has an α \alpha α -trivialization and L u ( X / G ) \mathcal{L}^{u}(X/G) L u ( X / G ) is coherent ind-geometric in the sense of [CW23, Definitions 4.1 and 4.3]. Then the inclusion of the α \alpha α -trivial block of IndCoh ( L u ( X / G ) ) \operatorname{IndCoh}(\mathcal{L}^{u}(X/G)) IndCoh ( L u ( X / G )) induces an equivalence,
IndCoh ( L u ( X / G ) ) α t T a t e ( α ) → ∼ IndCoh ( L u ( X / G ) ) t T a t e ( α ) (25) \operatorname{IndCoh}(\mathcal{L}^{u}(X/G))^{\mathrm{tTate}(\alpha)}_{\alpha} \xrightarrow{\sim} \operatorname{IndCoh}(\mathcal{L}^{u}(X/G))^{\mathrm{tTate}(\alpha)} \tag*{(25)} IndCoh ( L u ( X / G ) ) α tTate ( α ) ∼ IndCoh ( L u ( X / G ) ) tTate ( α ) (25) where t T a t e ( α ) \mathrm{tTate}(\alpha) tTate ( α ) denotes the renormalized Tate invariants with respect to the α \alpha α -twisted loop rotation action.
Proof. We will investigate the S 1 S^1 S 1 -invariants of the category Coh ( L u ( X / G ) ) \operatorname{Coh}(\mathcal{L}^{u}(X/G)) Coh ( L u ( X / G )) . Let X ∗ ( ⟨ α ⟩ ) X^*(\langle\alpha\rangle) X ∗ (⟨ α ⟩) denote the character lattice of ⟨ α ⟩ \langle\alpha\rangle ⟨ α ⟩ . The category C = Coh ( L u ( X / G ) ) \mathcal{C}=\operatorname{Coh}(\mathcal{L}^{u}(X/G)) C = Coh ( L u ( X / G )) has a direct sum decomposition
C ≃ ⨁ χ ∈ X ∗ ( ⟨ α ⟩ ) C χ . \mathcal{C}\simeq\bigoplus_{\chi\in X^*(\langle\alpha\rangle)}\mathcal{C}_{\chi}. C ≃ χ ∈ X ∗ (⟨ α ⟩) ⨁ C χ .
arising from the α \alpha α -trivialization of the stack L u ( X / G ) \mathcal{L}^{u}(X/G) L u ( X / G ) coming from Lemma 5.7. The α \alpha α -shifted loop rotation action on L u ( X / G ) \mathcal{L}^{u}(X/G) L u ( X / G ) induces a corresponding S 1 S^1 S 1 -action on C \mathcal{C} C . Moreover, since any S 1 S^1 S 1 -action on C \mathcal{C} C preserves the decomposition C ≃ ⨁ χ C χ \mathcal{C}\simeq\bigoplus_{\chi}\mathcal{C}_{\chi} C ≃ ⨁ χ C χ ,36 we have
C S 1 ( α ) ≃ ⨁ χ ∈ X ∗ ( ⟨ α ⟩ ) C χ S 1 ( α ) . \mathcal{C}^{S^1(\alpha)}\simeq\bigoplus_{\chi\in X^*(\langle\alpha\rangle)}\mathcal{C}_{\chi}^{S^1(\alpha)}. C S 1 ( α ) ≃ χ ∈ X ∗ (⟨ α ⟩) ⨁ C χ S 1 ( α ) .
We claim that C χ S 1 ( α ) ≃ 0 \mathcal{C}_{\chi}^{S^1(\alpha)}\simeq0 C χ S 1 ( α ) ≃ 0 whenever χ ( α ) ≠ 1 \chi(\alpha)\ne1 χ ( α ) = 1 . Let χ ∈ X ∗ ( ⟨ α ⟩ ) \chi\in X^*(\langle\alpha\rangle) χ ∈ X ∗ (⟨ α ⟩) be any character such that χ ( α ) ≠ 1 \chi(\alpha)\ne1 χ ( α ) = 1 . Let c ~ ∈ C χ S 1 ( α ) \widetilde{c}\in\mathcal{C}_{\chi}^{S^1(\alpha)} c ∈ C χ S 1 ( α ) be an S 1 ( α ) S^1(\alpha) S 1 ( α ) -equivariant object, and denote by c ∈ C χ c\in\mathcal{C}_{\chi} c ∈ C χ its underlying object. Let ρ : Z → Aut Aut ( C ) ( id C ) \rho:\mathbb{Z}\to\operatorname{Aut}_{\operatorname{Aut}(\mathcal{C})}(\operatorname{id}_{\mathcal{C}}) ρ : Z → Aut Aut ( C ) ( id C ) denote ordinary loop rotation action, and ρ c : c → c \rho_c:c\to c ρ c : c → c denote the component at c c c of the natural automorphism ρ ( 1 ) \rho(1) ρ ( 1 ) . By Proposition 3.31 , ρ c \rho_c ρ c is a unipotent operator, hence has the form ρ c ≃ 1 + N c \rho_c\simeq1+N_c ρ c ≃ 1 + N c for some nilpotent operator N c ∈ Ω ∞ ( End C ( c ) ) N_c\in\Omega^\infty(\operatorname{End}_{\mathcal{C}}(c)) N c ∈ Ω ∞ ( End C ( c )) . Furthermore, the restriction to C χ \mathcal{C}_{\chi} C χ of the α \alpha α -shifted rotation action on C \mathcal{C} C is given by multiplication by the scalar χ ( α ) \chi(\alpha) χ ( α ) . Thus, the component at c c c of the natural automorphism of the identity endofunctor given by ρ ( α ) \rho(\alpha) ρ ( α ) is
ρ ( α ) c ≃ χ ( α ) ρ c ≃ χ ( α ) ( 1 + N c ) . \rho(\alpha)_c\simeq\chi(\alpha)\rho_c\simeq\chi(\alpha)(1+N_c). ρ ( α ) c ≃ χ ( α ) ρ c ≃ χ ( α ) ( 1 + N c ) .
The S 1 ( α ) S^1(\alpha) S 1 ( α ) -equivariant structure on c ~ \widetilde{c} c supplies a nullhomotopy 1 − ρ ( α ) c ≃ 0 1-\rho(\alpha)_c\simeq0 1 − ρ ( α ) c ≃ 0 . We claim, however, that 1 − ρ ( α ) c 1-\rho(\alpha)_c 1 − ρ ( α ) c is invertible. Indeed, let n ∈ N n\in\mathbb{N} n ∈ N be such that N c n ≃ 0 N_c^n\simeq0 N c n ≃ 0 . Then
( 1 − ρ ( α ) c ) − 1 ≃ ( ( 1 − χ ( α ) ( 1 + N c ) ) − 1 ≃ 1 1 − χ ( α ) ∑ i = 0 n − 1 ( χ ( α ) 1 − χ ( α ) N c ) i . \begin{aligned}
(1-\rho(\alpha)_c)^{-1}\simeq((1-\chi(\alpha)(1+N_c))^{-1} \\
&\simeq\frac{1}{1-\chi(\alpha)}\sum_{i=0}^{n-1}\left(\frac{\chi(\alpha)}{1-\chi(\alpha)}N_c\right)^i.
\end{aligned} ( 1 − ρ ( α ) c ) − 1 ≃ (( 1 − χ ( α ) ( 1 + N c ) ) − 1 ≃ 1 − χ ( α ) 1 i = 0 ∑ n − 1 ( 1 − χ ( α ) χ ( α ) N c ) i .
Thus, the identity automorphism must be nullhomotopic itself, id c ≃ 0 \operatorname{id}_c\simeq0 id c ≃ 0 , which implies that c ≃ 0 c\simeq0 c ≃ 0 . Since the forgetful functor C χ S 1 ( α ) → C χ \mathcal{C}_{\chi}^{S^1(\alpha)}\to\mathcal{C}_{\chi} C χ S 1 ( α ) → C χ is conservative,37 we conclude that c ~ ≃ 0 \widetilde{c}\simeq0 c ≃ 0 . Thus C χ S 1 ( α ) ≃ 0 \mathcal{C}_{\chi}^{S^1(\alpha)}\simeq0 C χ S 1 ( α ) ≃ 0 for χ ( α ) ≠ 1 \chi(\alpha)\ne1 χ ( α ) = 1 . To complete the proof of equivalence (25) , it suffices to observe that the α \alpha α -trivial block of C S 1 \mathcal{C}^{S^1} C S 1 is precisely the summand C χ 1 \mathcal{C}_{\chi_1} C χ 1 for the character χ 1 \chi_1 χ 1 such that χ 1 ( α ) = 1 \chi_1(\alpha)=1 χ 1 ( α ) = 1 .
Nil-invariance
We now prove the main result of this section: that the category of cyclic sheaves on the α \alpha α -potent loop space of a global quotient stack is invariant under nil-isomorphism. Such a result follows from a form of Koszul duality for cyclic sheaves appearing in the future work [BZCHN], although our proof does not rely on this, instead utilizing the fixed point localization results of [3 ] . We do however utilize the following theorem of [BZCHN].
Theorem 5.18. Let X / G X/G X / G be a quotient stack of a derived scheme of finite type by an affine algebraic group. Then the canonical inclusion ι X : L ^ ( X / G ) → L u ( X / G ) \iota_X : \widehat{\mathcal L}(X/G) \to\mathcal L^u(X/G) ι X : L ( X / G ) → L u ( X / G ) induces a pair of adjoint equivalences
IndCoh ( L ^ ( X / G ) ) t T a t e ⇄ ι X t T a t e ( ι X ! ) t T a t e IndCoh ( L u ( X / G ) ) t T a t e . \operatorname{IndCoh}(\widehat{\mathcal L}(X/G))^{t\mathrm{Tate}} \underset{(\iota_X^!)^{t\mathrm{Tate}}}{\overset{\iota_X^{t\mathrm{Tate}}}{\rightleftarrows}} \operatorname{IndCoh}(\mathcal L^u(X/G))^{t\mathrm{Tate}}. IndCoh ( L ( X / G ) ) t Tate ( ι X ! ) t Tate ⇄ ι X t Tate IndCoh ( L u ( X / G ) ) t Tate .
For use in the below proposition, we record the following corollary.
Corollary 5.19. In the setup of Theorem 5.18, suppose that Z → X Z \to X Z → X is a G G G -invariant closed subscheme and G G G has a reductive neutral component. If Z → X Z \to X Z → X is almost finitely presented and X X X has affine diagonal, then the canonical inclusion ι ^ : L ^ ( X Z / G ) → L u ( X Z / G ) \widehat{\iota} : \widehat{\mathcal L}(X_Z/G) \to\mathcal L^u(X_Z/G) ι : L ( X Z / G ) → L u ( X Z / G ) induces a pair of adjoint equivalences
IndCoh ( L ^ ( X Z / G ) ) t T a t e ⇄ ι ^ t T a t e ( ι ^ ! ) t T a t e IndCoh ( L u ( X Z / G ) ) t T a t e . \operatorname{IndCoh}(\widehat{\mathcal L}(X_Z/G))^{t\mathrm{Tate}} \underset{(\widehat{\iota}^!)^{t\mathrm{Tate}}}{\overset{\widehat{\iota}^{t\mathrm{Tate}}}{\rightleftarrows}} \operatorname{IndCoh}(\mathcal L^u(X_Z/G))^{t\mathrm{Tate}}. IndCoh ( L ( X Z / G ) ) t Tate ( ι ! ) t Tate ⇄ ι t Tate IndCoh ( L u ( X Z / G ) ) t Tate .
Proof. For ease of notation, we drop the ( − ) t T a t e (-)^{t\mathrm{Tate}} ( − ) t Tate when denoting the relevant pullbacks or pushforwards; for example, we write ι ^ ! \widehat{\iota}^{!} ι ! for ( ι ^ ! ) t T a t e (\widehat{\iota}^{!})^{t\mathrm{Tate}} ( ι ! ) t Tate . Denote the open complement to Z Z Z in X X X by U U U , and let j : U / G ↪ X / G j : U/G \hookrightarrow X/G j : U / G ↪ X / G and ι : Z / G ↪ X / G \iota: Z/G \hookrightarrow X/G ι : Z / G ↪ X / G denote the inclusions. We claim that we have a commutative diagram
L ^ ( U / G ) → j ^ L ^ ( X / G ) ι U ↓ ↓ ι X L u ( U / G ) → j u L u ( X / G ) (26) \begin{CD}
\widehat{\mathcal L}(U/G) @>{\widehat{j}}>> \widehat{\mathcal L}(X/G) \\
@V{\iota_U}VV @VV{\iota_X}V \\
\mathcal L^u(U/G) @>{j^u}>> \mathcal L^u(X/G) \tag*{(26)}
\end{CD} L ( U / G ) ι U ↓ ⏐ L u ( U / G ) j j u L ( X / G ) ↓ ⏐ ι X L u ( X / G ) (26)
where j ^ \widehat{j} j (resp. j u j^u j u ) denotes L ^ j \widehat{\mathcal L}j L j (resp. L u j \mathcal L^u j L u j ). Indeed, by Lemma 3.59 (resp. Lemma 5.1 with α = 1 \alpha= 1 α = 1 ), the open complement to L ^ ( Z / G ) \widehat{\mathcal L}(Z/G) L ( Z / G ) in L ^ ( X / G ) \widehat{\mathcal L}(X/G) L ( X / G ) (resp. to L u ( Z / G ) \mathcal L^u(Z/G) L u ( Z / G ) in L u ( X / G ) \mathcal L^u(X/G) L u ( X / G ) ) is L ^ ( U / G ) \widehat{\mathcal L}(U/G) L ( U / G ) (resp. L u ( U / G ) \mathcal L^u(U/G) L u ( U / G ) ). Applying IndCoh ( − ) t T a t e \operatorname{IndCoh}(-)^{t\mathrm{Tate}} IndCoh ( − ) t Tate to (26) , we obtain a commutative diagram
IndCoh ( L u ( X / G ) ) t T a t e → ι X ! IndCoh ( L ^ ( X / G ) ) t T a t e ( j u ) ! ↓ ↓ j ^ ! IndCoh ( L u ( U / G ) ) t T a t e → ι U ! IndCoh ( L ^ ( U / G ) ) t T a t e \begin{CD}
\operatorname{IndCoh}(\mathcal L^u(X/G))^{t\mathrm{Tate}} @>{\iota_X^!}>> \operatorname{IndCoh}(\widehat{\mathcal L}(X/G))^{t\mathrm{Tate}} \\
@V{(j^u)!}VV @VV{\widehat{j}^{!}}V \\
\operatorname{IndCoh}(\mathcal L^u(U/G))^{t\mathrm{Tate}} @>{\iota_U^!}>> \operatorname{IndCoh}(\widehat{\mathcal L}(U/G))^{t\mathrm{Tate}}
\end{CD} IndCoh ( L u ( X / G ) ) t Tate ( j u )! ↓ ⏐ IndCoh ( L u ( U / G ) ) t Tate ι X ! ι U ! IndCoh ( L ( X / G ) ) t Tate ↓ ⏐ j ! IndCoh ( L ( U / G ) ) t Tate
Thus, we see that if F ∈ IndCoh L u ( Z / G ) ( L u ( X / G ) ) t T a t e \mathcal F \in\operatorname{IndCoh}_{\mathcal L^u(Z/G)}(\mathcal L^u(X/G))^{t\mathrm{Tate}} F ∈ IndCoh L u ( Z / G ) ( L u ( X / G ) ) t Tate , then ι X ! F ∈ IndCoh L ^ ( Z / G ) ( L ^ ( X / G ) ) t T a t e \iota_X^!\mathcal F \in\operatorname{IndCoh}_{\widehat{\mathcal L}(Z/G)}(\widehat{\mathcal L}(X/G))^{t\mathrm{Tate}} ι X ! F ∈ IndCoh L ( Z / G ) ( L ( X / G ) ) t Tate . This implies that the equivalence ι X ! \iota_X^! ι X ! of Theorem 5.18 restricts to a fully faithful functor
IndCoh L u ( Z / G ) ( L u ( X / G ) ) t T a t e ↪ IndCoh L ^ ( Z / G ) ( L ^ ( X / G ) ) t T a t e . \operatorname{IndCoh}_{\mathcal L^u(Z/G)}(\mathcal L^u(X/G))^{t\mathrm{Tate}} \hookrightarrow\operatorname{IndCoh}_{\widehat{\mathcal L}(Z/G)}(\widehat{\mathcal L}(X/G))^{t\mathrm{Tate}}. IndCoh L u ( Z / G ) ( L u ( X / G ) ) t Tate ↪ IndCoh L ( Z / G ) ( L ( X / G ) ) t Tate .
We claim this functor is an equivalence. To see this, let G ∈ IndCoh L ^ ( Z / G ) ( L ^ ( X / G ) ) t T a t e \mathcal G \in\operatorname{IndCoh}_{\widehat{\mathcal L}(Z/G)}(\widehat{\mathcal L}(X/G))^{t\mathrm{Tate}} G ∈ IndCoh L ( Z / G ) ( L ( X / G ) ) t Tate . By Theorem 5.18, there is an F ∈ IndCoh ( L u ( X / G ) ) t T a t e \mathcal F \in\operatorname{IndCoh}(\mathcal L^u(X/G))^{t\mathrm{Tate}} F ∈ IndCoh ( L u ( X / G ) ) t Tate such that ι X ! F = G \iota_X^!\mathcal F = \mathcal G ι X ! F = G . By definition of the category with support, j ^ ! G ≃ 0 \widehat{j}^{!}\mathcal G \simeq0 j ! G ≃ 0 . Thus,
ι U ! ( j u ) ! F ≃ j ^ ! ι X ! F \iota_U^!(j^u)!\mathcal F \simeq\widehat{j}^{!}\iota_X^!\mathcal F ι U ! ( j u )! F ≃ j ! ι X ! F
Since ι U ! \iota_U^! ι U ! is also an equivalence by Theorem 5.18 applied to U / G U/G U / G , we find that ( j u ) ! F ≃ 0 (j^u)!\mathcal F \simeq0 ( j u )! F ≃ 0 , so that F ∈ IndCoh L u ( Z / G ) ( L u ( X / G ) ) t T a t e \mathcal F \in\operatorname{IndCoh}_{\mathcal L^u(Z/G)}(\mathcal L^u(X/G))^{t\mathrm{Tate}} F ∈ IndCoh L u ( Z / G ) ( L u ( X / G ) ) t Tate . Thus, we have shown that ι X ! \iota_X^! ι X ! restricts to an equivalence
IndCoh L u ( Z / G ) ( L u ( X / G ) ) t T a t e → ∼ IndCoh L ^ ( Z / G ) ( L ^ ( X / G ) ) t T a t e . (27) \operatorname{IndCoh}_{\mathcal L^u(Z/G)}(\mathcal L^u(X/G))^{t\mathrm{Tate}} \xrightarrow{\sim} \operatorname{IndCoh}_{\widehat{\mathcal L}(Z/G)}(\widehat{\mathcal L}(X/G))^{t\mathrm{Tate}}. \tag*{(27)} IndCoh L u ( Z / G ) ( L u ( X / G ) ) t Tate ∼ IndCoh L ( Z / G ) ( L ( X / G ) ) t Tate . (27)
We next claim that its left adjoint ι X ∗ \iota_{X*} ι X ∗ restricts to the inverse of this equivalence. For this, it is sufficient to show that ι X ∗ \iota_{X*} ι X ∗ restricts to a functor between the categories with support. Thus, let G ∈ IndCoh L ( Z / G ) ^ ( L ( X / G ) ^ ) t T a t e G \in\operatorname{IndCoh}_{\widehat{\mathcal{L}(Z/G)}}(\widehat{\mathcal{L}(X/G)})^{t\mathrm{Tate}} G ∈ IndCoh L ( Z / G ) ( L ( X / G ) ) t Tate as before. A similar argument to that in the proof of Lemma 3.59 shows that the diagram (26) is Cartesian. Thus, base change along this diagram gives that
( ι u ) ′ i X ∗ G ≃ i U ∗ ′ G ≃ 0 , (\iota^{u})' i_{X*}\mathcal{G} \simeq i_{U*}'\mathcal{G} \simeq0, ( ι u ) ′ i X ∗ G ≃ i U ∗ ′ G ≃ 0 ,
so that i X ∗ i_{X*} i X ∗ restricts as required.
We will now show that the functors i ^ ∗ \widehat{i}_* i ∗ and i ^ ! \widehat{i}^{!} i ! are adjoint equivalences. To see this, we apply Corollary 4.13 and Corollary 5.5 to each of the respective inclusions ι u : ( L u ( X / G ) ) L u ( Z / G ) → L u ( X / G ) \iota^{u} : (\mathcal{L}^{u}(X/G))_{\mathcal{L}^{u}(Z/G)} \to\mathcal{L}^{u}(X/G) ι u : ( L u ( X / G ) ) L u ( Z / G ) → L u ( X / G ) and ι ^ : ( L ( X / G ) ^ ) L ( Z / G ) ^ → L ( X / G ) ^ \iota^{\widehat{\ }} : (\widehat{\mathcal{L}(X/G)})_{\widehat{\mathcal{L}(Z/G)}} \to\widehat{\mathcal{L}(X/G)} ι : ( L ( X / G ) ) L ( Z / G ) → L ( X / G ) . We thus obtain adjoint equivalences
IndCoh L u ( Z / G ) ( L u ( X / G ) ) t T a t e ⇄ ( ι u ) ′ ! i ∗ u IndCoh ( L u ( X ^ Z / G ) ) t T a t e \operatorname{IndCoh}_{\mathcal{L}^{u}(Z/G)}(\mathcal{L}^{u}(X/G))^{t\mathrm{Tate}}
\underset{i_{*}^{u}}{\overset{(\iota^{u})'!}{\rightleftarrows}}
\operatorname{IndCoh}(\mathcal{L}^{u}(\widehat{X}_{Z}/G))^{t\mathrm{Tate}} IndCoh L u ( Z / G ) ( L u ( X / G ) ) t Tate i ∗ u ⇄ ( ι u ) ′ ! IndCoh ( L u ( X Z / G ) ) t Tate
and
IndCoh L ( Z / G ) ^ ( L ( X / G ) ^ ) t T a t e ⇄ ( ι ^ ) ′ ! i ^ ∗ IndCoh ( L ( X ^ Z / G ) ^ ) t T a t e . \operatorname{IndCoh}_{\widehat{\mathcal{L}(Z/G)}}(\widehat{\mathcal{L}(X/G)})^{t\mathrm{Tate}}
\underset{\widehat{i}^{*}}{\overset{(\iota^{\widehat{\ }})'!}{\rightleftarrows}}
\operatorname{IndCoh}(\widehat{\mathcal{L}(\widehat{X}_{Z}/G)})^{t\mathrm{Tate}}. IndCoh L ( Z / G ) ( L ( X / G ) ) t Tate i ∗ ⇄ ( ι ) ′ ! IndCoh ( L ( X Z / G ) ) t Tate .
Combining these with the natural adjoint equivalences of (27) , we obtain a commutative diagram
IndCoh L u ( Z / G ) ( L u ( X / G ) ) t T a t e → i ^ ! IndCoh L ( Z / G ) ^ ( L ( X / G ) ^ ) t T a t e i ∗ u ↓ ↓ i ^ ∗ IndCoh ( L u ( X ^ Z / G ) ) t T a t e → i ^ ! IndCoh ( L ( X ^ Z / G ) ^ ) t T a t e \begin{CD}
\operatorname{IndCoh}_{\mathcal{L}^{u}(Z/G)}(\mathcal{L}^{u}(X/G))^{t\mathrm{Tate}}
@>{\widehat{i}^{!}}>>
\operatorname{IndCoh}_{\widehat{\mathcal{L}(Z/G)}}(\widehat{\mathcal{L}(X/G)})^{t\mathrm{Tate}} \\
@V{i_{*}^{u}}VV @VV{\widehat{i}_{*}}V \\
\operatorname{IndCoh}(\mathcal{L}^{u}(\widehat{X}_{Z}/G))^{t\mathrm{Tate}}
@>{\widehat{i}^{!}}>>
\operatorname{IndCoh}(\widehat{\mathcal{L}(\widehat{X}_{Z}/G)})^{t\mathrm{Tate}}
\end{CD} IndCoh L u ( Z / G ) ( L u ( X / G ) ) t Tate i ∗ u ↓ ⏐ IndCoh ( L u ( X Z / G ) ) t Tate i ! i ! IndCoh L ( Z / G ) ( L ( X / G ) ) t Tate ↓ ⏐ i ∗ IndCoh ( L ( X Z / G ) ) t Tate
Since all functors in the diagram except i ^ ! \widehat{i}^{!} i ! and i ^ ∗ \widehat{i}_{*} i ∗ are equivalences, we deduce that those functors are equivalences, too. □
We now give the proof of nil-invariance of Coh ( − ) t S 1 \operatorname{Coh}(-)^{tS^{1}} Coh ( − ) t S 1 on the α \alpha α -potent loop spaces.
Proposition 5.22. Let U U U be a finite type derived scheme with affine diagonal and equipped with an action of a reductive group G G G , and let Z Z Z be a G G G -invariant closed subscheme of U U U such that the inclusion Z → U Z \to U Z → U is almost finitely presented. Let L α ( i ) : L α ( Z / G ) → L α ( U ^ Z / G ) \mathcal{L}^{\alpha}(i) : \mathcal{L}^{\alpha}(Z/G) \to\mathcal{L}^{\alpha}(\widehat{U}_{Z}/G) L α ( i ) : L α ( Z / G ) → L α ( U Z / G ) denote the standard inclusion. Then ( L α ( i ) ∗ ) t S 1 : Coh ( L α ( Z / G ) ) t S 1 → Coh ( L α ( U ^ Z / G ) ) t S 1 (\mathcal{L}^{\alpha}(i)_*)^{tS^{1}} : \operatorname{Coh}(\mathcal{L}^{\alpha}(Z/G))^{tS^{1}} \to\operatorname{Coh}(\mathcal{L}^{\alpha}(\widehat{U}_{Z}/G))^{tS^{1}} ( L α ( i ) ∗ ) t S 1 : Coh ( L α ( Z / G ) ) t S 1 → Coh ( L α ( U Z / G ) ) t S 1 is an equivalence.
Proof. We begin by showing that the related functor,
( L α ( i ) ∗ ) t T a t e : IndCoh ( L α ( Z / G ) ) t T a t e → IndCoh ( L α ( U ^ Z / G ) ) t T a t e , (\mathcal{L}^{\alpha}(i)_*)^{t\mathrm{Tate}} : \operatorname{IndCoh}(\mathcal{L}^{\alpha}(Z/G))^{t\mathrm{Tate}} \to\operatorname{IndCoh}(\mathcal{L}^{\alpha}(\widehat{U}_{Z}/G))^{t\mathrm{Tate}}, ( L α ( i ) ∗ ) t Tate : IndCoh ( L α ( Z / G ) ) t Tate → IndCoh ( L α ( U Z / G ) ) t Tate ,
is an equivalence. Afterwards, we will show that L α ( i ) ∗ \mathcal{L}^{\alpha}(i)_* L α ( i ) ∗ preserves coherent objects. Since L α ( i ) \mathcal{L}^{\alpha}(i) L α ( i ) is S 1 S^{1} S 1 -equivariant, we can restrict this equivalence to compact objects to conclude, by using [CW23, Proposition 5.30] and Corollary 3.36.
As noted above, by [BZCHN23, Corollary A.18] and Lemma 5.13 , respectively, the shifted localization maps s ℓ Z : L u ( Z α / G ( α ) ) → L α ( Z / G ) s\ell_Z : \mathcal{L}^{u}(Z^{\alpha}/G(\alpha)) \to\mathcal{L}^{\alpha}(Z/G) s ℓ Z : L u ( Z α / G ( α )) → L α ( Z / G ) and s ℓ U ^ Z : L u ( ( U ^ Z ) α / G ( α ) ) ≃ L u ( U ^ Z α α / G ( α ) ) → L α ( U ^ Z / G ) s\ell_{\widehat{U}_Z} : \mathcal{L}^{u}((\widehat{U}_Z)^{\alpha}/G(\alpha)) \simeq\mathcal{L}^{u}(\widehat{U}^{\alpha}_{Z^{\alpha}}/G(\alpha)) \to\mathcal{L}^{\alpha}(\widehat{U}_Z/G) s ℓ U Z : L u (( U Z ) α / G ( α )) ≃ L u ( U Z α α / G ( α )) → L α ( U Z / G ) are S 1 S^{1} S 1 -equivariant equivalences for the α \alpha α -twisted loop rotation S 1 S^{1} S 1 -action ρ ( α ) \rho(\alpha) ρ ( α ) on the source and the ordinary loop rotation S 1 S^{1} S 1 -action ρ \rho ρ on target. It is straightforward to see that these equivalences fit into a commutative diagram of S 1 S^{1} S 1 -equivariant maps,
L u ( Z α / G ( α ) ) → L u ( i α ) L u ( U ^ Z α α / G ( α ) ) s ℓ Z ↓ ↓ s ℓ U ^ Z L α ( Z / G ) → L α ( i ) L α ( U ^ Z / G ) , \begin{CD}
\mathcal{L}^{u}(Z^{\alpha}/G(\alpha))
@>{\mathcal{L}^{u}(i^{\alpha})}>>
\mathcal{L}^{u}(\widehat{U}^{\alpha}_{Z^{\alpha}}/G(\alpha)) \\
@V{s\ell_Z}VV @VV{s\ell_{\widehat{U}_Z}}V \\
\mathcal{L}^{\alpha}(Z/G)
@>{\mathcal{L}^{\alpha}(i)}>>
\mathcal{L}^{\alpha}(\widehat{U}_Z/G),
\end{CD} L u ( Z α / G ( α )) s ℓ Z ↓ ⏐ L α ( Z / G ) L u ( i α ) L α ( i ) L u ( U Z α α / G ( α )) ↓ ⏐ s ℓ U Z L α ( U Z / G ) ,
where i α : Z α / G ( α ) → U ^ Z α α i^\alpha: Z^\alpha/G(\alpha) \to\widehat{U}^\alpha_{Z_\alpha} i α : Z α / G ( α ) → U Z α α is map on fixed point loci induced by the S 1 S^1 S 1 -equivariant map i i i .
After applying IndCoh ( − ) \operatorname{IndCoh}(-) IndCoh ( − ) and the ( − ) t Tate (-)^{t\operatorname{Tate}} ( − ) t Tate construction (which respects composition of functors by [3 Lemma 4.6.1] ), this commutative diagram yields the following commutative diagram of functors,
IndCoh ( L u ( Z α / G ( α ) ) ) t Tate ( α ) → ( L u ( i α ) ∗ ) t Tate ( α ) IndCoh ( L u ( U ^ Z α α / G ( α ) ) ) t Tate ( α ) ≃ ↓ ↓ ≃ IndCoh ( L α ( Z / G ) ) t Tate → ( L α ( i ) ∗ ) t Tate IndCoh ( L α ( U Z / G ) ) t Tate , \begin{CD}
\operatorname{IndCoh}(\mathcal{L}^{u}(Z^\alpha/G(\alpha)))^{t\operatorname{Tate}(\alpha)} @>{(\mathcal{L}^{u}(i^\alpha)_*)^{t\operatorname{Tate}(\alpha)}}>> \operatorname{IndCoh}(\mathcal{L}^{u}(\widehat{U}^\alpha_{Z_\alpha}/G(\alpha)))^{t\operatorname{Tate}(\alpha)} \\
@V{\simeq}VV @VV{\simeq}V \\
\operatorname{IndCoh}(\mathcal{L}^{\alpha}(Z/G))^{t\operatorname{Tate}} @>{(\mathcal{L}^{\alpha}(i)_*)^{t\operatorname{Tate}}}>> \operatorname{IndCoh}(\mathcal{L}^{\alpha}(U_Z/G))^{t\operatorname{Tate}},
\end{CD} IndCoh ( L u ( Z α / G ( α )) ) t Tate ( α ) ≃ ↓ ⏐ IndCoh ( L α ( Z / G ) ) t Tate ( L u ( i α ) ∗ ) t Tate ( α ) ( L α ( i ) ∗ ) t Tate IndCoh ( L u ( U Z α α / G ( α )) ) t Tate ( α ) ↓ ⏐ ≃ IndCoh ( L α ( U Z / G ) ) t Tate ,
where t Tate ( α ) t\operatorname{Tate}(\alpha) t Tate ( α ) and t Tate t\operatorname{Tate} t Tate denote the renormalized Tate invariants taken with respect to the ρ ( α ) \rho(\alpha) ρ ( α ) and ρ \rho ρ S 1 S^1 S 1 -actions, respectively. Indeed, [CW23, Proposition 5.30] applies, as all stacks in questions are formal completions of coherent geometric stacks by Corollary 3.36, hence are coherent ind-geometric stacks by Lemma 3.35. Thus, it suffices to show that ( L u ( i α ) ∗ ) t Tate ( α ) (\mathcal{L}^{u}(i^\alpha)_*)^{t\operatorname{Tate}(\alpha)} ( L u ( i α ) ∗ ) t Tate ( α ) is an equivalence.
By Proposition 5.17 ,39 it in fact suffices to show that the induced functor on α \alpha α -trivial blocks,
( L u ( i α ) ∗ ) α t Tate ( α ) : IndCoh ( L u ( Z α / G ( α ) ) ) α t Tate ( α ) → IndCoh ( L u ( U ^ Z α α / G ( α ) ) ) α t Tate ( α ) (\mathcal{L}^{u}(i^\alpha)_*)^{t\operatorname{Tate}(\alpha)}_\alpha:\operatorname{IndCoh}(\mathcal{L}^{u}(Z^\alpha/G(\alpha)))^{t\operatorname{Tate}(\alpha)}_\alpha\to\operatorname{IndCoh}(\mathcal{L}^{u}(\widehat{U}^\alpha_{Z_\alpha}/G(\alpha)))^{t\operatorname{Tate}(\alpha)}_\alpha ( L u ( i α ) ∗ ) α t Tate ( α ) : IndCoh ( L u ( Z α / G ( α )) ) α t Tate ( α ) → IndCoh ( L u ( U Z α α / G ( α )) ) α t Tate ( α )
is an equivalence. In order to do so, we note that on the α \alpha α -trivial blocks, IndCoh ( L u ( Z α / G ( α ) ) ) α \operatorname{IndCoh}(\mathcal{L}^{u}(Z^\alpha/G(\alpha)))_\alpha IndCoh ( L u ( Z α / G ( α )) ) α and IndCoh ( L u ( U ^ Z α α / G ( α ) ) ) α \operatorname{IndCoh}(\mathcal{L}^{u}(\widehat{U}^\alpha_{Z_\alpha}/G(\alpha)))_\alpha IndCoh ( L u ( U Z α α / G ( α )) ) α , the α \alpha α -twisted and ordinary S 1 S^1 S 1 -actions coincide. Thus, there are equivalences that fit into a commutative diagram as follows:
\begin{CD}
\operatorname{IndCoh}(\mathcal{L}^{u}(Z^\alpha/G(\alpha)))^{t\operatorname{Tate}}_\alpha@>{(\mathcal{L}^{u}(i^\alpha)_*)^{t\operatorname{Tate}}}>> \operatorname{IndCoh}(\mathcal{L}^{u}(\widehat{U}^\alpha_{Z_\alpha}/G(\alpha)))^{t\operatorname{Tate}}_\alpha\\
@V{\simeq}VV @VV{\simeq}V \\
\operatorname{IndCoh}(\mathcal{L}^{u}(Z^\alpha/G(\alpha)))^{t\operatorname{Tate}(\alpha)}_\alpha@>{(\mathcal{L}^{u}(i^\alpha)_*)^{t\operatorname{Tate}(\alpha)}}>> \operatorname{IndCoh}(\mathcal{L}^{u}(\widehat{U}^\alpha_{Z_\alpha}/G(\alpha)))^{t\operatorname{Tate}(\alpha)}_\alpha.
\end{CD}
Because L u ( i α ) ∗ \mathcal{L}^{u}(i^\alpha)_* L u ( i α ) ∗ respects the decomposition of its source and target given by the α \alpha α -trivialization,40 and the S 1 S^1 S 1 -action preserves full subcategories, it therefore suffices to show the stronger statement that
( L u ( i α ) ∗ ) t Tate : IndCoh ( L u ( Z α / G ( α ) ) ) t Tate → IndCoh ( L u ( U ^ Z α α / G ( α ) ) ) t Tate (\mathcal{L}^{u}(i^\alpha)_*)^{t\operatorname{Tate}}:\operatorname{IndCoh}(\mathcal{L}^{u}(Z^\alpha/G(\alpha)))^{t\operatorname{Tate}} \to\operatorname{IndCoh}(\mathcal{L}^{u}(\widehat{U}^\alpha_{Z_\alpha}/G(\alpha)))^{t\operatorname{Tate}} ( L u ( i α ) ∗ ) t Tate : IndCoh ( L u ( Z α / G ( α )) ) t Tate → IndCoh ( L u ( U Z α α / G ( α )) ) t Tate
is an equivalence. By Theorem 5.18 and Corollary 5.19, we see that it suffices to show that
( L ^ ( i α ) ∗ ) t Tate : IndCoh ( L ^ ( Z α / G ( α ) ) ) t Tate → IndCoh ( L ^ ( U ^ Z α α / G ( α ) ) ) t Tate (\widehat{\mathcal{L}}(i^\alpha)_*)^{t\operatorname{Tate}}:\operatorname{IndCoh}(\widehat{\mathcal{L}}(Z^\alpha/G(\alpha)))^{t\operatorname{Tate}} \to\operatorname{IndCoh}(\widehat{\mathcal{L}}(\widehat{U}^\alpha_{Z_\alpha}/G(\alpha)))^{t\operatorname{Tate}} ( L ( i α ) ∗ ) t Tate : IndCoh ( L ( Z α / G ( α )) ) t Tate → IndCoh ( L ( U Z α α / G ( α )) ) t Tate
is an equivalence. To this end, let p ^ : L ^ ( Z α / G ( α ) ) → ( Z α / G ( α ) ) d R \widehat{p}:\widehat{\mathcal{L}}(Z^\alpha/G(\alpha)) \to(Z^\alpha/G(\alpha))_{\mathrm{dR}} p : L ( Z α / G ( α )) → ( Z α / G ( α ) ) dR and p ^ ′ : L ^ ( U ^ Z α α / G ( α ) ) → ( U ^ Z α α / G ( α ) ) d R \widehat{p}':\widehat{\mathcal{L}}(\widehat{U}^\alpha_{Z_\alpha}/G(\alpha)) \to(\widehat{U}^\alpha_{Z_\alpha}/G(\alpha))_{\mathrm{dR}} p ′ : L ( U Z α α / G ( α )) → ( U Z α α / G ( α ) ) dR denote the canonical maps from the formal loop space to the de Rham stack, where our notation p ^ \widehat{p} p and p ^ ′ \widehat{p}' p ′ is chosen to agree with that in [3 ] . We note that p ^ \widehat{p} p and p ^ ′ \widehat{p}' p ′ are canonically
L u ( U ^ Z α α / G ( α ) ) ≃ ( L u ( U α / G ( α ) ) ) L u ( Z α / G ( α ) ) ≃ L ( U α / G ( α ) ) × ( L ( U α / G ( α ) ) ) d R O 1 ( U α / G ( α ) ) d R × O 1 ( U α / G ( α ) ) d R O 1 ( Z α / G ( α ) ) d R ≃ L ( U α / G ( α ) ) × ( L ( U α / G ( α ) ) ) d R O 1 ( Z α / G ( α ) ) d R \begin{aligned}
\mathcal{L}^{u}(\widehat{U}^\alpha_{Z_\alpha}/G(\alpha))\simeq(\mathcal{L}^{u}(U^\alpha/G(\alpha)))_{\mathcal{L}^{u}(Z^\alpha/G(\alpha))} \\
&\simeq\mathcal{L}(U^\alpha/G(\alpha))\mathop{\times}_{(\mathcal{L}(U^\alpha/G(\alpha)))_{\mathrm{dR}}}\mathcal{O}_1(U^\alpha/G(\alpha))_{\mathrm{dR}}\mathop{\times}_{\mathcal{O}_1(U^\alpha/G(\alpha))_{\mathrm{dR}}}\mathcal{O}_1(Z^\alpha/G(\alpha))_{\mathrm{dR}} \\
&\simeq\mathcal{L}(U^\alpha/G(\alpha))\mathop{\times}_{(\mathcal{L}(U^\alpha/G(\alpha)))_{\mathrm{dR}}}\mathcal{O}_1(Z^\alpha/G(\alpha))_{\mathrm{dR}}
\end{aligned} L u ( U Z α α / G ( α )) ≃ ( L u ( U α / G ( α )) ) L u ( Z α / G ( α )) ≃ L ( U α / G ( α )) × ( L ( U α / G ( α )) ) dR O 1 ( U α / G ( α ) ) dR × O 1 ( U α / G ( α ) ) dR O 1 ( Z α / G ( α ) ) dR ≃ L ( U α / G ( α )) × ( L ( U α / G ( α )) ) dR O 1 ( Z α / G ( α ) ) dR
: = L ( U α / G ( α ) ) O 1 ( Z α / G ( α ) ) . :=\mathcal{L}(U^\alpha/G(\alpha))_{\mathcal{O}_1(Z^\alpha/G(\alpha))}. := L ( U α / G ( α ) ) O 1 ( Z α / G ( α )) .
where in the second line we use nil-invariance of ( − ) d R (-)_{\mathrm{dR}} ( − ) dR . We also note that the inclusion O 1 ( Z α / G ( α ) ) ↪ L ( U α / G ( α ) ) \mathcal{O}_1(Z^\alpha/G(\alpha))\hookrightarrow\mathcal{L}(U^\alpha/G(\alpha)) O 1 ( Z α / G ( α )) ↪ L ( U α / G ( α )) is an almost finitely presented closed immersion, because it is the composition of a base change of the inclusion of the point α \alpha α into G / / G G//G G // G with the almost finitely presented closed immersion L Z → L X \mathcal{L}Z\to\mathcal{L}X L Z → L X . S 1 S^1 S 1 -equivariant for the loop rotation action on the source and the trivial S 1 S^1 S 1 -action on the target. Hence, we have a commutative diagram,
IndCoh ( L ( Z α / G ( α ) ) ^ ) t T a t e → ( L ( i α ) ∗ ^ ) t T a t e IndCoh ( L ( U ~ Z α / G ( α ) ) ^ ) t T a t e ↓ ( p ^ ∗ ) t T a t e ↓ ( p ^ ∗ ′ ) t T a t e IndCoh ( ( Z α / G ( α ) ) d R ) t T a t e → ( i d R ∗ α ) t T a t e IndCoh ( ( U ~ Z α / G ( α ) ) d R ) t T a t e , (28) \begin{aligned}
\operatorname{IndCoh}\left(\widehat{\mathcal{L}(Z^\alpha/G(\alpha))}\right)^{t\mathrm{Tate}} &\xrightarrow{\left(\widehat{\mathcal{L}(i^\alpha)_*}\right)^{t\mathrm{Tate}}} \operatorname{IndCoh}\left(\widehat{\mathcal{L}(\widetilde{U}_{Z^\alpha}/G(\alpha))}\right)^{t\mathrm{Tate}} \\
\downarrow{(\widehat{p}_*)}^{t\mathrm{Tate}} & \qquad\qquad\qquad\qquad\qquad\qquad\downarrow{(\widehat{p}'_*)}^{t\mathrm{Tate}} \\
\operatorname{IndCoh}\left((Z^\alpha/G(\alpha))_{\mathrm{dR}}\right)^{t\mathrm{Tate}} &\xrightarrow{\left(i^\alpha_{\mathrm{dR}*}\right)^{t\mathrm{Tate}}} \operatorname{IndCoh}\left(\left(\widetilde{U}_{Z^\alpha}/G(\alpha)\right)_{\mathrm{dR}}\right)^{t\mathrm{Tate}}, \tag*{(28)}
\end{aligned} IndCoh ( L ( Z α / G ( α )) ) t Tate ↓ ( p ∗ ) t Tate IndCoh ( ( Z α / G ( α ) ) dR ) t Tate ( L ( i α ) ∗ ) t Tate IndCoh ( L ( U Z α / G ( α )) ) t Tate ↓ ( p ∗ ′ ) t Tate ( i dR ∗ α ) t Tate IndCoh ( ( U Z α / G ( α ) ) dR ) t Tate , (28)
coming from the identification p ^ ′ ∘ L ( i α ) ^ ≃ i d R α ∘ p ^ \widehat{p}' \circ\widehat{\mathcal{L}(i^\alpha)} \simeq i^\alpha_{\mathrm{dR}} \circ\widehat{p} p ′ ∘ L ( i α ) ≃ i dR α ∘ p and the fact that t T a t e t\mathrm{Tate} t Tate is a functor on Fun ( B S 1 , Coh t ) \operatorname{Fun}(BS^1,\operatorname{Coh}_{t}) Fun ( B S 1 , Coh t ) . The vertical arrows in (28) are equivalences by [3 Theorem 5.1.3] .41 Meanwhile, i d R α i^\alpha_{\mathrm{dR}} i dR α is an isomorphism because the map i α : Z α / G ( α ) ↪ U ~ Z α / G ( α ) i^\alpha: Z^\alpha/G(\alpha) \hookrightarrow\widetilde{U}_{Z^\alpha}/G(\alpha) i α : Z α / G ( α ) ↪ U Z α / G ( α ) is a formal completion by Lemma 3.48, hence induces an isomorphism on underlying reduced stacks. Thus, ( i d R ∗ α ) t T a t e (i^\alpha_{\mathrm{dR}*})^{t\mathrm{Tate}} ( i dR ∗ α ) t Tate is an equivalence, so ( L ( i α ) ∗ ^ ) t T a t e (\widehat{\mathcal{L}(i^\alpha)_*})^{t\mathrm{Tate}} ( L ( i α ) ∗ ) t Tate , hence ( L α ( i ) ∗ ) t T a t e (\mathcal{L}^\alpha(i)_*)^{t\mathrm{Tate}} ( L α ( i ) ∗ ) t Tate , is an equivalence as desired.
It remains to show then that ( L α i ) ∗ (\mathcal{L}^\alpha i)_* ( L α i ) ∗ preserves Coh ( − ) \operatorname{Coh}(-) Coh ( − ) . We begin by observing that L α i \mathcal{L}^\alpha i L α i is ind-inf-schematic and ind-proper. Indeed, the map i : Z / G → U ~ Z / G i: Z/G \to\widetilde{U}_Z/G i : Z / G → U Z / G is clearly an ind-schematic and ind-proper map of stacks.42 By the stability of ind-schematic and ind-proper morphisms under base change, we deduce that L ( Z / G ) → L ( U ~ Z / G ) \mathcal{L}(Z/G) \to\mathcal{L}(\widetilde{U}_Z/G) L ( Z / G ) → L ( U Z / G ) is an ind-schematic map of stacks, which implies, again by stability under base change, that the induced map of formal completions L α i : L α ( Z / G ) → L α ( U ~ Z / G ) \mathcal{L}^\alpha i: \mathcal{L}^\alpha(Z/G) \to\mathcal{L}^\alpha(\widetilde{U}_Z/G) L α i : L α ( Z / G ) → L α ( U Z / G ) is ind-schematic and ind-proper. In particular, it is ind-inf-schematic and ind-proper. Thus, by [GR17b, Chapter 3 Proposition 3.2.4], ( L α i ) ∗ (\mathcal{L}^\alpha i)_* ( L α i ) ∗ admits a continuous right adjoint, which implies that ( L α i ) ∗ (\mathcal{L}^\alpha i)_* ( L α i ) ∗ preserves compact objects, which for the stacks involved means Coh ( − ) \operatorname{Coh}(-) Coh ( − ) . □
The global quotient case
Lemma 6.1. Let X → S X \to S X → S be a reasonable ind-geometric stack over an affine scheme S = Spec A S = \operatorname{Spec} A S = Spec A . Let S ′ = Spec B → S S' = \operatorname{Spec} B \to S S ′ = Spec B → S be a flat map. Then for F , G ∈ Coh ( X ) \mathcal{F}, \mathcal{G} \in\operatorname{Coh}(X) F , G ∈ Coh ( X ) , the canonical map
Hom Coh ( X ) ( F , G ) ⊗ A B → Hom Coh ( X × S S ′ ) ( p ∗ F , p ∗ G ) \operatorname{Hom}_{\operatorname{Coh}(X)}(\mathcal{F},\mathcal{G}) \otimes_A B \to\operatorname{Hom}_{\operatorname{Coh}(X \times_S S')} (p^*\mathcal{F},p^*\mathcal{G}) Hom Coh ( X ) ( F , G ) ⊗ A B → Hom Coh ( X × S S ′ ) ( p ∗ F , p ∗ G ) is an isomorphism.
Proof. Consider the Cartesian square
X n × S S ′ → p n X n p ^ n ↓ ↓ f n S ′ → g S \begin{CD}
X_n \times_S S' @>{p_n}>> X_n \\
@V{\widehat{p}_n}VV @VV{f_n}V \\
S' @>{g}>> S
\end{CD} X n × S S ′ p n ↓ ⏐ S ′ p n g X n ↓ ⏐ f n S All objects in the square are geometric stacks, and both horizontal maps are flat. First, suppose that F , G ∈ Coh ( X n ) \mathcal{F},\mathcal{G} \in\operatorname{Coh}(X_n) F , G ∈ Coh ( X n ) . By [CW23, Propositions 7.2 and 7.3], we have
p n ∗ Hom QCoh ( X n ) ( F , G ) ≃ Hom QCoh ( X n × S S ′ ) ( p n ∗ F , p n ∗ G ) , p_n^*\operatorname{Hom}_{\operatorname{QCoh}(X_n)}(\mathcal{F},\mathcal{G}) \simeq\operatorname{Hom}_{\operatorname{QCoh}(X_n \times_S S')} (p_n^*\mathcal{F},p_n^*\mathcal{G}), p n ∗ Hom QCoh ( X n ) ( F , G ) ≃ Hom QCoh ( X n × S S ′ ) ( p n ∗ F , p n ∗ G ) , since F \mathcal{F} F is almost perfect, G \mathcal{G} G is bounded below, and p n p_n p n is flat. Now, Proposition 3.8 of op. cit. gives
g ∗ f n ∗ Hom QCoh ( X n ) ( F , G ) ≃ p ^ n ∗ p n ∗ Hom QCoh ( X n ) ( F , G ) . g^*f_{n*}\operatorname{Hom}_{\operatorname{QCoh}(X_n)}(\mathcal{F},\mathcal{G}) \simeq\widehat{p}_{n*}p_n^*\operatorname{Hom}_{\operatorname{QCoh}(X_n)}(\mathcal{F},\mathcal{G}). g ∗ f n ∗ Hom QCoh ( X n ) ( F , G ) ≃ p n ∗ p n ∗ Hom QCoh ( X n ) ( F , G ) .
Thus, we obtain
Hom X n ( F , G ) ⊗ A B ≃ Hom X n × S S ′ ( p n ∗ F , p n ∗ G ) . \operatorname{Hom}_{X_n}(\mathcal{F},\mathcal{G}) \otimes_A B \simeq\operatorname{Hom}_{X_n \times_S S'}(p_n^*\mathcal{F},p_n^*\mathcal{G}). Hom X n ( F , G ) ⊗ A B ≃ Hom X n × S S ′ ( p n ∗ F , p n ∗ G ) .
By Proposition 5.5 of op. cit. ,
Coh ( X ) ≃ colim n Coh ( X n ) Coh ( X × S S ′ ) ≃ colim n Coh ( X n × S S ′ ) . \begin{aligned}
\operatorname{Coh}(X) &\simeq\underset{n}{\operatorname{colim}}\,\operatorname{Coh}(X_n)\\
\operatorname{Coh}(X \times_S S') &\simeq\underset{n}{\operatorname{colim}}\,\operatorname{Coh}(X_n \times_S S').
\end{aligned} Coh ( X ) Coh ( X × S S ′ ) ≃ n colim Coh ( X n ) ≃ n colim Coh ( X n × S S ′ ) .
Now, let F , G ∈ Coh ( X ) \mathcal{F},\mathcal{G} \in\operatorname{Coh}(X) F , G ∈ Coh ( X ) , and suppose they are represented by F n , G n ∈ Coh ( X n ) \mathcal{F}_n,\mathcal{G}_n \in\operatorname{Coh}(X_n) F n , G n ∈ Coh ( X n ) . For m ≥ n m \ge n m ≥ n , write F m , G m \mathcal{F}_m,\mathcal{G}_m F m , G m for their pushforwards to X m X_m X m . We compute
Hom Coh ( X ) ( F , G ) ⊗ A B ≃ colim m ≥ n Hom Coh ( X m ) ( F m , G m ) ⊗ A B ≃ colim m ≥ n Hom Coh ( X m × S S ′ ) ( p m ∗ F m , p m ∗ G m ) ≃ Hom Coh ( X × S S ′ ) ( p ∗ F , p ∗ G ) . \begin{aligned}
\operatorname{Hom}_{\operatorname{Coh}(X)}(\mathcal{F},\mathcal{G}) \otimes_A B
&\simeq\underset{m \ge n}{\operatorname{colim}}\,\operatorname{Hom}_{\operatorname{Coh}(X_m)}(\mathcal{F}_m,\mathcal{G}_m) \otimes_A B\\
&\simeq\underset{m \ge n}{\operatorname{colim}}\,\operatorname{Hom}_{\operatorname{Coh}(X_m \times_S S')} (p_m^*\mathcal{F}_m,p_m^*\mathcal{G}_m)\\
&\simeq\operatorname{Hom}_{\operatorname{Coh}(X \times_S S')}(p^*\mathcal{F},p^*\mathcal{G}).
\end{aligned} Hom Coh ( X ) ( F , G ) ⊗ A B ≃ m ≥ n colim Hom Coh ( X m ) ( F m , G m ) ⊗ A B ≃ m ≥ n colim Hom Coh ( X m × S S ′ ) ( p m ∗ F m , p m ∗ G m ) ≃ Hom Coh ( X × S S ′ ) ( p ∗ F , p ∗ G ) .
Lemma 6.2. Let M ∈ A -mod M \in A\text{-mod} M ∈ A -mod be equipped with an S 1 S^1 S 1 -action, and suppose M M M is cohomologically bounded below. If A → B A \to B A → B is flat, then the natural map
M t S 1 ⊗ A [ u ± 1 ] B [ u ± 1 ] ⟶ ( M ⊗ A B ) t S 1 M^{tS^1} \otimes_{A[u^{\pm1}]} B[u^{\pm1}] \longrightarrow(M \otimes_A B)^{tS^1} M t S 1 ⊗ A [ u ± 1 ] B [ u ± 1 ] ⟶ ( M ⊗ A B ) t S 1 is an isomorphism. Here the S 1 S^1 S 1 -action on M ⊗ A B M \otimes_A B M ⊗ A B is the one induced by the trivial S 1 S^1 S 1 -actions on A A A and B B B .
Proof. We begin by proving that
M S 1 ⊗ A [ u ] B [ u ] → ≃ ( M ⊗ A B ) S 1 . M^{S^1} \otimes_{A[u]} B[u] \xrightarrow{\simeq} (M \otimes_A B)^{S^1}. M S 1 ⊗ A [ u ] B [ u ] ≃ ( M ⊗ A B ) S 1 .
By definition, M S 1 ≃ Hom A [ ε ] ( A , M ) M^{S^1} \simeq\operatorname{Hom}_{A[\varepsilon]}(A,M) M S 1 ≃ Hom A [ ε ] ( A , M ) , where deg ε = − 1 \deg\varepsilon=-1 deg ε = − 1 . We now use the standard resolution of A A A as an A [ ε ] A[\varepsilon] A [ ε ] -module,
P : = Tot ( ⋯ ⟶ A [ ε ] [ 2 ] ⟶ A [ ε ] [ 1 ] ⟶ A [ ε ] ) , P := \operatorname{Tot}(\cdots\longrightarrow A[\varepsilon][2] \longrightarrow A[\varepsilon][1] \longrightarrow A[\varepsilon]), P := Tot ( ⋯ ⟶ A [ ε ] [ 2 ] ⟶ A [ ε ] [ 1 ] ⟶ A [ ε ]) ,
where the map A [ ε ] [ j ] → A [ ε ] [ j − 1 ] A[\varepsilon][j] \to A[\varepsilon][j-1] A [ ε ] [ j ] → A [ ε ] [ j − 1 ] is multiplication by ε \varepsilon ε . Applying Hom A [ ε ] ( − , M ) \operatorname{Hom}_{A[\varepsilon]}(-,M) Hom A [ ε ] ( − , M ) to this resolution we find
K ( M ) : = Hom A [ ε ] ( A , M ) ≃ M [ [ u ] ] , K(M) := \operatorname{Hom}_{A[\varepsilon]}(A,M) \simeq M[[u]], K ( M ) := Hom A [ ε ] ( A , M ) ≃ M [[ u ]] ,
where the differential is d = d M + u ε d=d_M+u\varepsilon d = d M + u ε , deg u = 2 \deg u=2 deg u = 2 , and d M d_M d M is the differential of M M M . This is the usual Koszul duality for mixed complexes. Note that P ≃ A [ ε , v ] P \simeq A[\varepsilon,v] P ≃ A [ ε , v ] with deg v = − 2 \deg v=-2 deg v = − 2 and differential given by d ( a v j ) = ε a v j − 1 d(av^j)=\varepsilon av^{j-1} d ( a v j ) = ε a v j − 1 . Consider the fiber sequence
M [ [ u ] ] [ − 2 n − 2 ] ⟶ M [ [ u ] ] ⟶ M [ u ] / u n + 1 M [ u ] . M[[u]][-2n-2] \longrightarrow M[[u]] \longrightarrow M[u]/u^{n+1}M[u]. M [[ u ]] [ − 2 n − 2 ] ⟶ M [[ u ]] ⟶ M [ u ] / u n + 1 M [ u ] .
Suppose the lower bound on M M M is b b b . Then the fiber
M [ [ u ] ] [ − 2 n − 2 ] M[[u]][-2n-2] M [[ u ]] [ − 2 n − 2 ]
is bounded below by b + 2 n + 2 b+2n+2 b + 2 n + 2 . There is a commutative square
K ( M ) ⊗ A B → K ( M ⊗ A B ) ↓ ↓ K n ( M ) ⊗ A B → ≃ K n ( M ⊗ A B ) , \begin{CD}
K(M)\otimes_A B @>>> K(M\otimes_A B)\\
@VVV @VVV\\
K_n(M)\otimes_A B @>\simeq>> K_n(M\otimes_A B),
\end{CD} K ( M ) ⊗ A B ↓ ⏐ K n ( M ) ⊗ A B ≃ K ( M ⊗ A B ) ↓ ⏐ K n ( M ⊗ A B ) ,
where K n ( M ) = ( M [ u ] / u n − 1 , d ) K_n(M)=(M[u]/u^{n-1},d) K n ( M ) = ( M [ u ] / u n − 1 , d ) . The map
K n ( M ) ⊗ A B ⟶ K n ( M ⊗ A B ) K_n(M)\otimes_A B \longrightarrow K_n(M\otimes_A B) K n ( M ) ⊗ A B ⟶ K n ( M ⊗ A B )
is an isomorphism because it involves only finitely many summands, and tensoring preserves direct sums. Since B B B is flat over A A A , the bound b + 2 n + 2 b+2n+2 b + 2 n + 2 is preserved. Thus, both vertical maps in the square have fibers bounded below by b + 2 n + 2 b+2n+2 b + 2 n + 2 . Now, fixing j j j , choose n n n such that b + 2 n + 2 > j + 1 b+2n+2>j+1 b + 2 n + 2 > j + 1 , in which case both vertical maps give isomorphisms on H j H^j H j . Since we can do this for arbitrary j j j , we find
K ( M ) ⊗ A B ≃ K ( M ⊗ A B ) . K(M) \otimes_A B \simeq K(M \otimes_A B). K ( M ) ⊗ A B ≃ K ( M ⊗ A B ) .
Now observe that K ( M ) ⊗ A B ≃ K ( M ) ⊗ A [ u ] B [ u ] K(M) \otimes_A B \simeq K(M) \otimes_{A[u]} B[u] K ( M ) ⊗ A B ≃ K ( M ) ⊗ A [ u ] B [ u ] , and we conclude that M S 1 ⊗ A [ u ] B [ u ] ≃ ( M ⊗ A B ) S 1 M^{S^1} \otimes_{A[u]} B[u] \simeq(M \otimes_A B)^{S^1} M S 1 ⊗ A [ u ] B [ u ] ≃ ( M ⊗ A B ) S 1 , as desired. To conclude, we apply base change − ⊗ k [ u ] k [ u ± 1 ] -\otimes_{k[u]} k[u^{\pm1}] − ⊗ k [ u ] k [ u ± 1 ] to each side of the equivalence to obtain,
M t S 1 ⊗ A [ u ± 1 ] B [ u ± 1 ] ≃ ( M S 1 ⊗ A [ u ] B [ u ] ) ⊗ k [ u ] k [ u ± 1 ] ≃ ( M ⊗ A B ) S 1 ⊗ k [ u ] k [ u ± 1 ] , M^{tS^1} \otimes_{A[u^{\pm1}]} B[u^{\pm1}] \simeq(M^{S^1} \otimes_{A[u]} B[u]) \otimes_{k[u]} k[u^{\pm1}] \simeq(M \otimes_A B)^{S^1} \otimes_{k[u]} k[u^{\pm1}], M t S 1 ⊗ A [ u ± 1 ] B [ u ± 1 ] ≃ ( M S 1 ⊗ A [ u ] B [ u ]) ⊗ k [ u ] k [ u ± 1 ] ≃ ( M ⊗ A B ) S 1 ⊗ k [ u ] k [ u ± 1 ] ,
as desired. □
Proposition 6.3. Let X → S X \to S X → S be a stack over S = Spec A S = \operatorname{Spec} A S = Spec A such that X X X is reasonable ind-geometric, and let S ′ = Spec B → S S' = \operatorname{Spec} B \to S S ′ = Spec B → S be a flat map. Suppose that X X X has an S 1 S^1 S 1 -action; S S S and S ′ S' S ′ are equipped with the trivial S 1 S^1 S 1 -action; and the maps X → S X \to S X → S and S → S ′ S \to S' S → S ′ are S 1 S^1 S 1 -equivariant. The projection p : X × S S ′ → X p : X \times_S S' \to X p : X × S S ′ → X is then canonically S 1 S^1 S 1 -equivariant, so p ∗ p^* p ∗ is S 1 S^1 S 1 -equivariant and its target is B B B -linear. Let θ \theta θ denote the natural Perf B [ u ± 1 ] \operatorname{Perf} B[u^{\pm1}] Perf B [ u ± 1 ] -linear functor
θ : Coh ( X ) t S 1 ⊗ A B → Coh ( X × S S ′ ) t S 1 \theta: \operatorname{Coh}(X)^{tS^1} \otimes_A B \to\operatorname{Coh}(X \times_S S')^{tS^1} θ : Coh ( X ) t S 1 ⊗ A B → Coh ( X × S S ′ ) t S 1 obtained by extension of scalars. Then θ \theta θ is fully faithful.
Proof. Let F ′ \mathcal{F}' F ′ , G ′ ∈ Coh ( X ) S 1 \mathcal{G}' \in\operatorname{Coh}(X)^{S^1} G ′ ∈ Coh ( X ) S 1 whose underlying object in Coh ( X ) \operatorname{Coh}(X) Coh ( X ) are denoted by F \mathcal{F} F and G \mathcal{G} G , respectively. Set M : = Hom Coh ( X ) A ( F , G ) M := \operatorname{Hom}_{\operatorname{Coh}(X)}^A(\mathcal{F}, \mathcal{G}) M := Hom Coh ( X ) A ( F , G ) , considered with its induced S 1 S^1 S 1 -action. Let F B \mathcal{F}_B F B and G B \mathcal{G}_B G B denote the image of F ′ \mathcal{F}' F ′ , G ′ \mathcal{G}' G ′ , respectively, in Coh ( X ) S 1 ⊗ A B \operatorname{Coh}(X)^{S^1} \otimes_A B Coh ( X ) S 1 ⊗ A B under extension of scalars. Then
Hom Coh ( X ) S 1 ⊗ A B B [ u ± 1 ] ( F B , G B ) ≃ ( M S 1 ⊗ A [ u ] A [ u ± 1 ] ) ⊗ A [ u ± 1 ] B [ u ± 1 ] (29) \operatorname{Hom}_{\operatorname{Coh}(X)^{S^1} \otimes_A B}^{B[u^{\pm1}]}(\mathcal{F}_B,\mathcal{G}_B) \simeq(M^{S^1} \otimes_{A[u]} A[u^{\pm1}]) \otimes_{A[u^{\pm1}]} B[u^{\pm1}] \tag*{(29)} Hom Coh ( X ) S 1 ⊗ A B B [ u ± 1 ] ( F B , G B ) ≃ ( M S 1 ⊗ A [ u ] A [ u ± 1 ]) ⊗ A [ u ± 1 ] B [ u ± 1 ] (29)
≃ ( M ⊗ A B ) S 1 (30) \simeq(M \otimes_A B)^{S^1} \tag*{(30)} ≃ ( M ⊗ A B ) S 1 (30)
≃ ( Hom Coh ( X × S S ′ ) B ( p ∗ F , p ∗ G ) ) S 1 ⊗ B [ u ] B [ u ± 1 ] ≃ Hom Coh ( X × S S ′ ) t S 1 B [ u ± 1 ] ( θ ( F B ) , θ ( G B ) ) , (31) \begin{aligned}
\simeq\left(\operatorname{Hom}_{\operatorname{Coh}(X \times_S S')}^B(p^*\mathcal{F},p^*\mathcal{G})\right)^{S^1} \otimes_{B[u]} B[u^{\pm1}] \\
&\simeq\operatorname{Hom}_{\operatorname{Coh}(X \times_S S')^{tS^1}}^{B[u^{\pm1}]}(\theta(\mathcal{F}_B),\theta(\mathcal{G}_B)), \tag*{(31)}
\end{aligned} ≃ ( Hom Coh ( X × S S ′ ) B ( p ∗ F , p ∗ G ) ) S 1 ⊗ B [ u ] B [ u ± 1 ] ≃ Hom Coh ( X × S S ′ ) t S 1 B [ u ± 1 ] ( θ ( F B ) , θ ( G B )) , (31)
where equivalence (29) follows from Lemma 3.7 ; (30) follows from Lemma 6.2 since objects in Coh ( X ) \operatorname{Coh}(X) Coh ( X ) are bounded; and (31) follows from Lemma 6.1 . □
We first prove Kashiwara’s lemma for D p o t ( − ) \mathcal{D}^{\mathrm{pot}}(-) D pot ( − ) in the case of a global quotient of a derived scheme. The proof in this case is broken up into two propositions shown below, proving fully faithfulness of i ∗ i_* i ∗ and essential surjectivity, respectively.
Theorem 6.7. Let U U U be a Noetherian scheme almost of finite type and with affine diagonal equipped with an action by a reductive group G G G . Let Z ↪ U Z \hookrightarrow U Z ↪ U be an almost finitely presented inclusion of a G G G -invariant closed derived subscheme of U U U . Then the pushforward along i : Z / G → U / G i : Z/G \to U/G i : Z / G → U / G induces an equivalence,
i ∗ : D p o t ( Z / G ) → D Z / G p o t ( U / G ) . i_* : \mathcal{D}^{\mathrm{pot}}(Z/G) \to\mathcal{D}^{\mathrm{pot}}_{Z/G}(U/G). i ∗ : D pot ( Z / G ) → D Z / G pot ( U / G ) .
Proof. Proposition 6.11 and Proposition 6.22 together imply that ( L i ∗ ) t S 1 : Coh ( L ( Z / G ) ) t S 1 ≃ Coh Z / G ( L ( U / G ) ) t S 1 (Li_*)^{tS^1} : \operatorname{Coh}(\mathcal{L}(Z/G))^{tS^1} \simeq\operatorname{Coh}_{Z/G}(\mathcal{L}(U/G))^{tS^1} ( L i ∗ ) t S 1 : Coh ( L ( Z / G ) ) t S 1 ≃ Coh Z / G ( L ( U / G ) ) t S 1 is an equivalence. Taking the ind-extension of this equivalence gives the desired result. □
Fully faithfulness
We first need two technical lemmas, to prepare for the proof of fully faithfulness below.
Lemma 6.8. Let A ∈ CAlg k A \in\operatorname{CAlg}_{k} A ∈ CAlg k , and let C ∈ ( dgCat A ) B S 1 \mathcal{C} \in(\operatorname{dgCat}_{A})^{BS^{1}} C ∈ ( dgCat A ) B S 1 . Let a ′ , b ′ ∈ C S 1 a',b' \in\mathcal{C}^{S^{1}} a ′ , b ′ ∈ C S 1 , and let a , b ∈ C a,b \in\mathcal{C} a , b ∈ C denote their underlying objects. Denote by η ∗ : C S 1 → C t S 1 \eta^{*}:\mathcal{C}^{S^{1}}\to\mathcal{C}^{tS^{1}} η ∗ : C S 1 → C t S 1 the canonical functor given by extension of scalars. Let P ∈ Perf ( A ) P \in\operatorname{Perf}(A) P ∈ Perf ( A ) , and let P ∨ = Hom A ( P , A ) ∈ Perf ( A ) P^{\vee}=\operatorname{Hom}_{A}(P,A)\in\operatorname{Perf}(A) P ∨ = Hom A ( P , A ) ∈ Perf ( A ) denote the dual of P P P . Then there is a natural isomorphism
Hom C t S 1 A ( η ∗ ( a ′ ⊗ A [ u ] P [ u ] ) , η ∗ ( b ′ ⊗ A [ u ] P [ u ] ) ) ≃ Hom C t S 1 A ( a ′ , b ′ ) ⊗ A P ∨ ⊗ A P . \operatorname{Hom}_{\mathcal{C}^{tS^{1}}}^{A}\left(\eta^{*}(a'\otimes_{A[u]}P[u]),\eta^{*}(b'\otimes_{A[u]}P[u])\right)\simeq\operatorname{Hom}_{\mathcal{C}^{tS^{1}}}^{A}(a',b')\otimes_{A}P^{\vee}\otimes_{A}P. Hom C t S 1 A ( η ∗ ( a ′ ⊗ A [ u ] P [ u ]) , η ∗ ( b ′ ⊗ A [ u ] P [ u ]) ) ≃ Hom C t S 1 A ( a ′ , b ′ ) ⊗ A P ∨ ⊗ A P . where for any x , y ∈ C t S 1 x,y\in\mathcal{C}^{tS^{1}} x , y ∈ C t S 1 , Hom C t S 1 A ( x , y ) \operatorname{Hom}_{\mathcal{C}^{tS^{1}}}^{A}(x,y) Hom C t S 1 A ( x , y ) denotes the restriction of the A [ u ± 1 ] A[u^{\pm1}] A [ u ± 1 ] -module Hom C t S 1 A [ u ± 1 ] ( x , y ) \operatorname{Hom}_{\mathcal{C}^{tS^{1}}}^{A[u^{\pm1}]}(x,y) Hom C t S 1 A [ u ± 1 ] ( x , y ) along A → A [ u ± 1 ] A\to A[u^{\pm1}] A → A [ u ± 1 ] .
Proof. We first show that there is a natural isomorphism of A [ u ] A[u] A [ u ] -modules
Hom C A ( a , b ) S 1 ⊗ A P ⊗ A P ∨ ≃ Hom C A ( a ⊗ A P , b ⊗ A P ) S 1 . \operatorname{Hom}_{\mathcal{C}}^{A}(a,b)^{S^{1}}\otimes_{A}P\otimes_{A}P^{\vee}\simeq\operatorname{Hom}_{\mathcal{C}}^{A}(a\otimes_{A}P,b\otimes_{A}P)^{S^{1}}. Hom C A ( a , b ) S 1 ⊗ A P ⊗ A P ∨ ≃ Hom C A ( a ⊗ A P , b ⊗ A P ) S 1 . Indeed,
Hom C A ( x , y ) S 1 : = lim B S 1 Hom C A ( x , y ) \operatorname{Hom}_{\mathcal{C}}^{A}(x,y)^{S^{1}}:=\lim_{BS^{1}}\operatorname{Hom}_{\mathcal{C}}^{A}(x,y) Hom C A ( x , y ) S 1 := B S 1 lim Hom C A ( x , y ) for any x , y ∈ C x,y\in\mathcal{C} x , y ∈ C . Since P , P ∨ P,P^{\vee} P , P ∨ are dualizable A A A -modules, the functors − ⊗ A P -\otimes_{A}P − ⊗ A P and − ⊗ A P ∨ -\otimes_{A}P^{\vee} − ⊗ A P ∨ are both left and right adjoints; in particular, they preserve limits, which gives the desired isomorphism. We next claim that the underlying object under the forgetful functor C S 1 → C \mathcal{C}^{S^{1}}\to\mathcal{C} C S 1 → C of a ′ ⊗ A [ u ] P [ u ] a'\otimes_{A[u]}P[u] a ′ ⊗ A [ u ] P [ u ] is precisely a ⊗ A P a\otimes_{A}P a ⊗ A P , and similarly for b b b . Indeed, the forgetful functor is A A A -linear, and − ⊗ A [ u ] P [ u ] -\otimes_{A[u]}P[u] − ⊗ A [ u ] P [ u ] is canonically isomorphic to − ⊗ A P -\otimes_{A}P − ⊗ A P . Thus, by Lemma 3.7 , we obtain a natural, A [ u ] A[u] A [ u ] -linear isomorphism
Hom C A ( a , b ) S 1 ⊗ A P ⊗ A P ∨ ≃ Hom C S 1 A [ u ] ( a ′ ⊗ A [ u ] P [ u ] , b ′ ⊗ A [ u ] P [ u ] ) . \operatorname{Hom}_{\mathcal{C}}^{A}(a,b)^{S^{1}}\otimes_{A}P\otimes_{A}P^{\vee}\simeq\operatorname{Hom}_{\mathcal{C}^{S^{1}}}^{A[u]}(a'\otimes_{A[u]}P[u],b'\otimes_{A[u]}P[u]). Hom C A ( a , b ) S 1 ⊗ A P ⊗ A P ∨ ≃ Hom C S 1 A [ u ] ( a ′ ⊗ A [ u ] P [ u ] , b ′ ⊗ A [ u ] P [ u ]) . Given x ′ , y ′ ∈ C S 1 x',y'\in\mathcal{C}^{S^{1}} x ′ , y ′ ∈ C S 1 , by [Per18], Lemma 2.7, the functor η ∗ \eta^{*} η ∗ induces a natural isomorphism of A [ u ± 1 ] A[u^{\pm1}] A [ u ± 1 ] -modules
Hom C S 1 A [ u ] ( x ′ , y ′ ) ⊗ A [ u ] A [ u ± 1 ] → ∼ Hom C t S 1 A [ u ± 1 ] ( η ∗ x ′ , η ∗ y ′ ) . \operatorname{Hom}_{\mathcal{C}^{S^{1}}}^{A[u]}(x',y')\otimes_{A[u]}A[u^{\pm1}]\xrightarrow{\sim}\operatorname{Hom}_{\mathcal{C}^{tS^{1}}}^{A[u^{\pm1}]}(\eta^{*}x',\eta^{*}y'). Hom C S 1 A [ u ] ( x ′ , y ′ ) ⊗ A [ u ] A [ u ± 1 ] ∼ Hom C t S 1 A [ u ± 1 ] ( η ∗ x ′ , η ∗ y ′ ) . We thus have the following chain of natural isomorphisms, where res \operatorname{res} res is the restriction functor along A → A [ u ± 1 ] A\to A[u^{\pm1}] A → A [ u ± 1 ] .
Hom C t S 1 A ( η ∗ a ′ , η ∗ b ′ ) ⊗ A P ⊗ A P ∨ ≃ res ( Hom C S 1 A [ u ] ( a ′ , b ′ ) ⊗ A P ⊗ A P ∨ ⊗ A [ u ] A [ u ± 1 ] ) ≃ res ( Hom C A ( a , b ) S 1 ⊗ A P ⊗ A P ∨ ⊗ A [ u ] A [ u ± 1 ] ) ≃ res ( Hom C A ( a ⊗ A P , b ⊗ A P ) S 1 ⊗ A [ u ] A [ u ± 1 ] ) ≃ res ( Hom C S 1 A [ u ] ( a ′ ⊗ A [ u ] P [ u ] , b ′ ⊗ A [ u ] P [ u ] ) ⊗ A [ u ] A [ u ± 1 ] ) ≃ Hom C t S 1 A ( η ∗ ( a ′ ⊗ A [ u ] P [ u ] ) , η ∗ ( b ′ ⊗ A [ u ] P [ u ] ) ) . \begin{aligned}
\operatorname{Hom}_{\mathcal{C}^{tS^{1}}}^{A}(\eta^{*}a',\eta^{*}b')\otimes_{A}P\otimes_{A}P^{\vee}
&\simeq\operatorname{res}\left(\operatorname{Hom}_{\mathcal{C}^{S^{1}}}^{A[u]}(a',b')\otimes_{A}P\otimes_{A}P^{\vee}\otimes_{A[u]}A[u^{\pm1}]\right)\\
&\simeq\operatorname{res}\left(\operatorname{Hom}_{\mathcal{C}}^{A}(a,b)^{S^{1}}\otimes_{A}P\otimes_{A}P^{\vee}\otimes_{A[u]}A[u^{\pm1}]\right)\\
&\simeq\operatorname{res}\left(\operatorname{Hom}_{\mathcal{C}}^{A}(a\otimes_{A}P,b\otimes_{A}P)^{S^{1}}\otimes_{A[u]}A[u^{\pm1}]\right)\\
&\simeq\operatorname{res}\left(\operatorname{Hom}_{\mathcal{C}^{S^{1}}}^{A[u]}(a'\otimes_{A[u]}P[u],b'\otimes_{A[u]}P[u])\otimes_{A[u]}A[u^{\pm1}]\right)\\
&\simeq\operatorname{Hom}_{\mathcal{C}^{tS^{1}}}^{A}\left(\eta^{*}(a'\otimes_{A[u]}P[u]),\eta^{*}(b'\otimes_{A[u]}P[u])\right).
\end{aligned} Hom C t S 1 A ( η ∗ a ′ , η ∗ b ′ ) ⊗ A P ⊗ A P ∨ ≃ res ( Hom C S 1 A [ u ] ( a ′ , b ′ ) ⊗ A P ⊗ A P ∨ ⊗ A [ u ] A [ u ± 1 ] ) ≃ res ( Hom C A ( a , b ) S 1 ⊗ A P ⊗ A P ∨ ⊗ A [ u ] A [ u ± 1 ] ) ≃ res ( Hom C A ( a ⊗ A P , b ⊗ A P ) S 1 ⊗ A [ u ] A [ u ± 1 ] ) ≃ res ( Hom C S 1 A [ u ] ( a ′ ⊗ A [ u ] P [ u ] , b ′ ⊗ A [ u ] P [ u ]) ⊗ A [ u ] A [ u ± 1 ] ) ≃ Hom C t S 1 A ( η ∗ ( a ′ ⊗ A [ u ] P [ u ]) , η ∗ ( b ′ ⊗ A [ u ] P [ u ]) ) . □
Lemma 6.9. Let A ∈ CAlg k A\in\operatorname{CAlg}_{k} A ∈ CAlg k , and let C , D ∈ ( dgCat A ) B S 1 \mathcal{C},\mathcal{D}\in(\operatorname{dgCat}_{A})^{BS^{1}} C , D ∈ ( dgCat A ) B S 1 . Suppose that F : C → D F:\mathcal{C}\to\mathcal{D} F : C → D is an S 1 S^{1} S 1 -equivariant, A A A -linear functor, and let a ′ , b ′ ∈ C S 1 a',b'\in\mathcal{C}^{S^{1}} a ′ , b ′ ∈ C S 1 whose underlying objects of C \mathcal{C} C are denoted by a , b ∈ C a,b\in\mathcal{C} a , b ∈ C , respectively. Let η C ∗ \eta_{\mathcal{C}}^{*} η C ∗ (resp. η D ∗ \eta_{\mathcal{D}}^{*} η D ∗ ) denote the canonical extension of scalars functor, C S 1 → C S 1 ⊗ A [ u ] A [ u ± 1 ] \mathcal{C}^{S^{1}}\to\mathcal{C}^{S^{1}}\otimes_{A[u]}A[u^{\pm1}] C S 1 → C S 1 ⊗ A [ u ] A [ u ± 1 ] (resp. D S 1 → D S 1 ⊗ A [ u ] A [ u ± 1 ] \mathcal{D}^{S^{1}}\to\mathcal{D}^{S^{1}}\otimes_{A[u]}A[u^{\pm1}] D S 1 → D S 1 ⊗ A [ u ] A [ u ± 1 ] ). Then there is a commutative diagram of
A [ u ± 1 ] A[u^{\pm1}] A [ u ± 1 ] -modules,
Hom C t S 1 A [ u ± 1 ] ( η C ∗ a ′ , η C ∗ b ′ ) → F η C ∗ a ′ , η C ∗ b ′ t S 1 Hom D t S 1 A [ u ± 1 ] ( F t S 1 η C ∗ a ′ , F t S 1 η C ∗ b ′ ) ≅ ↓ ↓ ≅ ( Hom C A ( a , b ) ) t S 1 → F a , b ( Hom D A ( F a , F b ) ) t S 1 , \begin{aligned}
\operatorname{Hom}_{\mathcal{C}^{tS^1}}^{A[u^{\pm1}]}(\eta_{\mathcal{C}}^*a',\eta_{\mathcal{C}}^*b') &\xrightarrow{F^{tS^1}_{\eta_{\mathcal{C}}^*a',\eta_{\mathcal{C}}^*b'}} \operatorname{Hom}_{\mathcal{D}^{tS^1}}^{A[u^{\pm1}]}(F^{tS^1}\eta_{\mathcal{C}}^*a',F^{tS^1}\eta_{\mathcal{C}}^*b') \\
\mathrel{\cong}\downarrow\quad&\quad\downarrow\mathrel{\cong} \\
(\operatorname{Hom}_{\mathcal{C}}^A(a,b))^{tS^1} &\xrightarrow{F_{a,b}} (\operatorname{Hom}_{\mathcal{D}}^A(Fa,Fb))^{tS^1},
\end{aligned} Hom C t S 1 A [ u ± 1 ] ( η C ∗ a ′ , η C ∗ b ′ ) ≅ ↓ ( Hom C A ( a , b ) ) t S 1 F η C ∗ a ′ , η C ∗ b ′ t S 1 Hom D t S 1 A [ u ± 1 ] ( F t S 1 η C ∗ a ′ , F t S 1 η C ∗ b ′ ) ↓ ≅ F a , b ( Hom D A ( F a , F b ) ) t S 1 ,
in which the vertical arrows are isomorphisms, as indicated.
Proof. First, observe that there is a commutative square
C S 1 → F S 1 D S 1 η C ∗ ↓ ↓ η D ∗ C t S 1 → F t S 1 D t S 1 . \begin{CD}
\mathcal{C}^{S^1} @>{F^{S^1}}>> \mathcal{D}^{S^1} \\
@V{\eta_{\mathcal{C}}^*}VV @VV{\eta_{\mathcal{D}}^*}V \\
\mathcal{C}^{tS^1} @>{F^{tS^1}}>> \mathcal{D}^{tS^1}.
\end{CD} C S 1 η C ∗ ↓ ⏐ C t S 1 F S 1 F t S 1 D S 1 ↓ ⏐ η D ∗ D t S 1 .
Indeed, functoriality of the tensor product gives that
η C ∗ = id C ⊗ η , η D ∗ = id D ⊗ η , F S 1 ⊗ k [ u ] k [ u ± 1 ] = F S 1 ⊗ id Perf ( k [ u ± 1 ] ) , \eta_{\mathcal{C}}^*=\operatorname{id}_{\mathcal{C}}\otimes\eta,\qquad
\eta_{\mathcal{D}}^*=\operatorname{id}_{\mathcal{D}}\otimes\eta,\qquad
F^{S^1}\otimes_{k[u]}k[u^{\pm1}]=F^{S^1}\otimes\operatorname{id}_{\operatorname{Perf}(k[u^{\pm1}])}, η C ∗ = id C ⊗ η , η D ∗ = id D ⊗ η , F S 1 ⊗ k [ u ] k [ u ± 1 ] = F S 1 ⊗ id Perf ( k [ u ± 1 ]) ,
where η : Perf ( k [ u ] ) → Perf ( k [ u ± 1 ] ) \eta:\operatorname{Perf}(k[u])\to\operatorname{Perf}(k[u^{\pm1}]) η : Perf ( k [ u ]) → Perf ( k [ u ± 1 ]) is the usual extension of scalars. Thus, we have the equivalence
η D ∗ ∘ F S 1 = ( id D ⊗ η ) ∘ F S 1 \eta_{\mathcal{D}}^*\circ F^{S^1}=(\operatorname{id}_{\mathcal{D}}\otimes\eta)\circ F^{S^1} η D ∗ ∘ F S 1 = ( id D ⊗ η ) ∘ F S 1
≅ F S 1 ⊗ η \cong F^{S^1}\otimes\eta ≅ F S 1 ⊗ η
≅ F S 1 ∘ ( id C ⊗ η ) , \cong F^{S^1}\circ(\operatorname{id}_{\mathcal{C}}\otimes\eta), ≅ F S 1 ∘ ( id C ⊗ η ) ,
as desired. Thus, we obtain the following commutative diagram
Hom C S 1 A [ u ] ( a ′ , b ′ ) → F S 1 Hom D S 1 A [ u ] ( F S 1 a ′ , F S 1 b ′ ) η C ∗ ↓ ↓ η D ∗ Hom C t S 1 A [ u ± 1 ] ( η C ∗ a ′ , η C ∗ b ′ ) → F t S 1 Hom D t S 1 A [ u ± 1 ] ( η D ∗ F S 1 a ′ , η D ∗ F S 1 b ′ ) . \begin{CD}
\operatorname{Hom}_{\mathcal{C}^{S^1}}^{A[u]}(a',b') @>{F^{S^1}}>> \operatorname{Hom}_{\mathcal{D}^{S^1}}^{A[u]}(F^{S^1}a',F^{S^1}b') \\
@V{\eta_{\mathcal{C}}^*}VV @VV{\eta_{\mathcal{D}}^*}V \\
\operatorname{Hom}_{\mathcal{C}^{tS^1}}^{A[u^{\pm1}]}(\eta_{\mathcal{C}}^*a',\eta_{\mathcal{C}}^*b') @>{F^{tS^1}}>> \operatorname{Hom}_{\mathcal{D}^{tS^1}}^{A[u^{\pm1}]}(\eta_{\mathcal{D}}^*F^{S^1}a',\eta_{\mathcal{D}}^*F^{S^1}b').
\end{CD} Hom C S 1 A [ u ] ( a ′ , b ′ ) η C ∗ ↓ ⏐ Hom C t S 1 A [ u ± 1 ] ( η C ∗ a ′ , η C ∗ b ′ ) F S 1 F t S 1 Hom D S 1 A [ u ] ( F S 1 a ′ , F S 1 b ′ ) ↓ ⏐ η D ∗ Hom D t S 1 A [ u ± 1 ] ( η D ∗ F S 1 a ′ , η D ∗ F S 1 b ′ ) .
By [Per18, Lemma 2.10], this diagram induces a commutative diagram
Hom C S 1 A [ u ] ( a ′ , b ′ ) ⊗ A [ u ] A [ u ± 1 ] → F S 1 Hom D S 1 A [ u ] ( F S 1 a ′ , F S 1 b ′ ) ⊗ A [ u ] A [ u ± 1 ] ≅ ↓ ↓ ≅ Hom C t S 1 A [ u ± 1 ] ( η C ∗ a ′ , η C ∗ b ′ ) → F t S 1 Hom D t S 1 A [ u ± 1 ] ( η D ∗ F S 1 a ′ , η D ∗ F S 1 b ′ ) . (32) \begin{CD}
\operatorname{Hom}_{\mathcal{C}^{S^1}}^{A[u]}(a',b')\otimes_{A[u]}A[u^{\pm1}] @>{F^{S^1}}>> \operatorname{Hom}_{\mathcal{D}^{S^1}}^{A[u]}(F^{S^1}a',F^{S^1}b')\otimes_{A[u]}A[u^{\pm1}] \\
@V{\cong}VV @VV{\cong}V \\
\operatorname{Hom}_{\mathcal{C}^{tS^1}}^{A[u^{\pm1}]}(\eta_{\mathcal{C}}^*a',\eta_{\mathcal{C}}^*b') @>{F^{tS^1}}>> \operatorname{Hom}_{\mathcal{D}^{tS^1}}^{A[u^{\pm1}]}(\eta_{\mathcal{D}}^*F^{S^1}a',\eta_{\mathcal{D}}^*F^{S^1}b'). \tag*{(32)}
\end{CD} Hom C S 1 A [ u ] ( a ′ , b ′ ) ⊗ A [ u ] A [ u ± 1 ] ≅ ↓ ⏐ Hom C t S 1 A [ u ± 1 ] ( η C ∗ a ′ , η C ∗ b ′ ) F S 1 F t S 1 Hom D S 1 A [ u ] ( F S 1 a ′ , F S 1 b ′ ) ⊗ A [ u ] A [ u ± 1 ] ↓ ⏐ ≅ Hom D t S 1 A [ u ± 1 ] ( η D ∗ F S 1 a ′ , η D ∗ F S 1 b ′ ) . (32)
Now, by Lemma 3.7 , we also have a commutative diagram
Hom C A ( a , b ) S 1 → F Hom D A ( F a , F b ) S 1 ≅ ↓ ↓ ≅ Hom C S 1 A [ u ] ( a ′ , b ′ ) → F S 1 Hom D S 1 A [ u ] ( F S 1 a ′ , F S 1 b ′ ) . \begin{CD}
\operatorname{Hom}_{\mathcal{C}}^A(a,b)^{S^1} @>{F}>> \operatorname{Hom}_{\mathcal{D}}^A(Fa,Fb)^{S^1} \\
@V{\cong}VV @VV{\cong}V \\
\operatorname{Hom}_{\mathcal{C}^{S^1}}^{A[u]}(a',b') @>{F^{S^1}}>> \operatorname{Hom}_{\mathcal{D}^{S^1}}^{A[u]}(F^{S^1}a',F^{S^1}b').
\end{CD} Hom C A ( a , b ) S 1 ≅ ↓ ⏐ Hom C S 1 A [ u ] ( a ′ , b ′ ) F F S 1 Hom D A ( F a , F b ) S 1 ↓ ⏐ ≅ Hom D S 1 A [ u ] ( F S 1 a ′ , F S 1 b ′ ) .
Applying the functor − ⊗ A [ u ] A [ u ± 1 ] -\otimes_{A[u]}A[u^{\pm1}] − ⊗ A [ u ] A [ u ± 1 ] to this diagram and combining the resulting diagram with (32) , we obtain the commutative diagram below, in which the desired diagram is the outer square.
Hom C A ( a , b ) t S 1 → F Hom D A ( F a , F b ) t S 1 ∥ ∥ Hom C A ( a , b ) S 1 ⊗ A [ u ] A [ u ± 1 ] → F Hom D A ( F a , F b ) S 1 ⊗ A [ u ] A [ u ± 1 ] ≃ ↓ ≃ ↓ Hom C S 1 A [ u ] ( a ′ , b ′ ) ⊗ A [ u ] A [ u ± 1 ] → F S 1 Hom D S 1 A [ u ] ( F S 1 a ′ , F S 1 b ′ ) ⊗ A [ u ] A [ u ± 1 ] ≃ ↓ ≃ ↓ Hom C t S 1 A [ u ± 1 ] ( η C ∗ a ′ , η C ∗ b ′ ) → F t S 1 Hom D t S 1 A [ u ± 1 ] ( η D ∗ F S 1 a ′ , η D ∗ F S 1 b ′ ) ∥ ≃ ↓ Hom C t S 1 A [ u ± 1 ] ( η C ∗ a ′ , η C ∗ b ′ ) → F t S 1 Hom D t S 1 A [ u ± 1 ] ( F t S 1 η C ∗ a ′ , F t S 1 η C ∗ b ′ ) \begin{aligned}
&\operatorname{Hom}_{\mathcal C}^{A}(a,b)^{tS^1}
&&\xrightarrow{\ F\ }
&&\operatorname{Hom}_{\mathcal D}^{A}(Fa,Fb)^{tS^1}\\
&\big\Vert&&&&\big\Vert\\
&\operatorname{Hom}_{\mathcal C}^{A}(a,b)^{S^1}\otimes_{A[u]}A[u^{\pm1}]
&&\xrightarrow{\ F\ }
&&\operatorname{Hom}_{\mathcal D}^{A}(Fa,Fb)^{S^1}\otimes_{A[u]}A[u^{\pm1}]\\
&\mathrel{\simeq}\downarrow&&&&\mathrel{\simeq}\downarrow\\
&\operatorname{Hom}_{\mathcal C_{S^1}}^{A[u]}(a',b')\otimes_{A[u]}A[u^{\pm1}]
&&\xrightarrow{\ F^{S^1}\ }
&&\operatorname{Hom}_{\mathcal D_{S^1}}^{A[u]}(F^{S^1}a',F^{S^1}b')\otimes_{A[u]}A[u^{\pm1}]\\
&\mathrel{\simeq}\downarrow&&&&\mathrel{\simeq}\downarrow\\
&\operatorname{Hom}_{\mathcal C_{tS^1}}^{A[u^{\pm1}]}(\eta_{\mathcal C}^{*}a',\eta_{\mathcal C}^{*}b')
&&\xrightarrow{\ F^{tS^1}\ }
&&\operatorname{Hom}_{\mathcal D_{tS^1}}^{A[u^{\pm1}]}(\eta_{\mathcal D}^{*}F^{S^1}a',\eta_{\mathcal D}^{*}F^{S^1}b')\\
&\big\Vert&&&&\mathrel{\simeq}\downarrow\\
&\operatorname{Hom}_{\mathcal C_{tS^1}}^{A[u^{\pm1}]}(\eta_{\mathcal C}^{*}a',\eta_{\mathcal C}^{*}b')
&&\xrightarrow{\ F^{tS^1}\ }
&&\operatorname{Hom}_{\mathcal D_{tS^1}}^{A[u^{\pm1}]}(F^{tS^1}\eta_{\mathcal C}^{*}a',F^{tS^1}\eta_{\mathcal C}^{*}b')
\end{aligned} Hom C A ( a , b ) t S 1 Hom C A ( a , b ) S 1 ⊗ A [ u ] A [ u ± 1 ] ≃ ↓ Hom C S 1 A [ u ] ( a ′ , b ′ ) ⊗ A [ u ] A [ u ± 1 ] ≃ ↓ Hom C t S 1 A [ u ± 1 ] ( η C ∗ a ′ , η C ∗ b ′ ) Hom C t S 1 A [ u ± 1 ] ( η C ∗ a ′ , η C ∗ b ′ ) F F F S 1 F t S 1 F t S 1 Hom D A ( F a , F b ) t S 1 Hom D A ( F a , F b ) S 1 ⊗ A [ u ] A [ u ± 1 ] ≃ ↓ Hom D S 1 A [ u ] ( F S 1 a ′ , F S 1 b ′ ) ⊗ A [ u ] A [ u ± 1 ] ≃ ↓ Hom D t S 1 A [ u ± 1 ] ( η D ∗ F S 1 a ′ , η D ∗ F S 1 b ′ ) ≃ ↓ Hom D t S 1 A [ u ± 1 ] ( F t S 1 η C ∗ a ′ , F t S 1 η C ∗ b ′ )
We now prove fully faithfulness.
Proposition 6.11. Suppose that U U U is a derived scheme almost of finite type, with affine diagonal, and equipped with an action by a reductive group G G G . Let Z ↪ U Z \hookrightarrow U Z ↪ U be an almost finitely presented inclusion of a G G G -invariant closed derived subscheme of U U U . Let i : L ( Z / G ) → L ( U ‾ Z / G ) i:\mathcal L(Z/G)\to\mathcal L(\overline{U}_Z/G) i : L ( Z / G ) → L ( U Z / G ) denote the map induced by the inclusion. Then the functor ( i ∗ ) t S 1 : Coh ( L ( Z / G ) ) t S 1 → Coh ( L ( U ‾ Z / G ) ) t S 1 (i_*)^{tS^1}:\operatorname{Coh}(\mathcal L(Z/G))^{tS^1}\to\operatorname{Coh}(\mathcal L(\overline{U}_Z/G))^{tS^1} ( i ∗ ) t S 1 : Coh ( L ( Z / G ) ) t S 1 → Coh ( L ( U Z / G ) ) t S 1 is fully faithful.
Proof. Let η : k [ u ] → k [ u ± 1 ] \eta:k[u]\to k[u^{\pm1}] η : k [ u ] → k [ u ± 1 ] denote the canonical localization map, and let η ∗ : Coh ( L ( Z / G ) ) S 1 → Coh ( L ( Z / G ) ) t S 1 \eta^*:\operatorname{Coh}(\mathcal L(Z/G))^{S^1}\to\operatorname{Coh}(\mathcal L(Z/G))^{tS^1} η ∗ : Coh ( L ( Z / G ) ) S 1 → Coh ( L ( Z / G ) ) t S 1 denote the functor given by extension of scalars. Let F ′ , G ′ ∈ Coh ( L ( Z / G ) ) S 1 \mathcal F',\mathcal G'\in\operatorname{Coh}(\mathcal L(Z/G))^{S^1} F ′ , G ′ ∈ Coh ( L ( Z / G ) ) S 1 be arbitrary S 1 S^1 S 1 -equivariant objects whose images under η ∗ \eta^* η ∗ in Coh ( L ( Z / G ) ) t S 1 \operatorname{Coh}(\mathcal L(Z/G))^{tS^1} Coh ( L ( Z / G ) ) t S 1 are denoted by F \mathcal F F and G \mathcal G G , respectively. Since i i i commutes with the maps μ Z \mu_Z μ Z and μ U ‾ Z \mu_{\overline{U}_Z} μ U Z , the functor ( i ∗ ) t S 1 : Coh ( L ( Z / G ) ) t S 1 → Coh ( L ( U ‾ Z / G ) ) t S 1 (i_*)^{tS^1}:\operatorname{Coh}(\mathcal L(Z/G))^{tS^1}\to\operatorname{Coh}(\mathcal L(\overline{U}_Z/G))^{tS^1} ( i ∗ ) t S 1 : Coh ( L ( Z / G ) ) t S 1 → Coh ( L ( U Z / G ) ) t S 1 is canonically A A A -linear, where A : = O ( G / / G ) A:=\mathcal O(G/\!/G) A := O ( G / / G ) . With this in mind, we define the following A A A -module,
F : = ker ( Hom Coh ( L ( Z / G ) ) t S 1 A ( F , G ) ⟶ Hom Coh ( L ( U ‾ Z / G ) ) t S 1 A ( ( i ∗ ) t S 1 F , ( i ∗ ) t S 1 G ) ) . \mathcal F:=\ker\left(\operatorname{Hom}_{\operatorname{Coh}(\mathcal L(Z/G))^{tS^1}}^{A}(\mathcal F,\mathcal G)\longrightarrow\operatorname{Hom}_{\operatorname{Coh}(\mathcal L(\overline{U}_Z/G))^{tS^1}}^{A}((i_*)^{tS^1}\mathcal F,(i_*)^{tS^1}\mathcal G)\right). F := ker ( Hom Coh ( L ( Z / G ) ) t S 1 A ( F , G ) ⟶ Hom Coh ( L ( U Z / G ) ) t S 1 A (( i ∗ ) t S 1 F , ( i ∗ ) t S 1 G ) ) .
where, if C \mathcal C C is an A [ u ± 1 ] A[u^{\pm1}] A [ u ± 1 ] -linear category, Hom C A ( − , − ) \operatorname{Hom}_{\mathcal C}^{A}(-,-) Hom C A ( − , − ) denotes the A A A -module obtained by the restriction of scalars functor res A [ u ± 1 ] / A : A [ u ± 1 ] -mod → A -mod \operatorname{res}_{A[u^{\pm1}]/A}:A[u^{\pm1}]\text{-mod}\to A\text{-mod} res A [ u ± 1 ] / A : A [ u ± 1 ] -mod → A -mod given as the right adjoint to − ⊗ A A [ u ± 1 ] -\otimes_A A[u^{\pm1}] − ⊗ A A [ u ± 1 ] . Since res A [ u ± 1 ] / A \operatorname{res}_{A[u^{\pm1}]/A} res A [ u ± 1 ] / A is a right adjoint it preserves limits, meaning
F ≃ res A [ u ± 1 ] / A ( ker ( Hom Coh ( L ( Z / G ) ) t S 1 A [ u ± 1 ] ( F , G ) ⟶ Hom Coh ( L ( U ‾ Z / G ) ) t S 1 A [ u ± 1 ] ( ( i ∗ ) t S 1 F , ( i ∗ ) t S 1 G ) ) ) . \mathcal F\simeq\operatorname{res}_{A[u^{\pm1}]/A}\left(\ker\left(\operatorname{Hom}_{\operatorname{Coh}(\mathcal L(Z/G))^{tS^1}}^{A[u^{\pm1}]}(\mathcal F,\mathcal G)\longrightarrow\operatorname{Hom}_{\operatorname{Coh}(\mathcal L(\overline{U}_Z/G))^{tS^1}}^{A[u^{\pm1}]}((i_*)^{tS^1}\mathcal F,(i_*)^{tS^1}\mathcal G)\right)\right). F ≃ res A [ u ± 1 ] / A ( ker ( Hom Coh ( L ( Z / G ) ) t S 1 A [ u ± 1 ] ( F , G ) ⟶ Hom Coh ( L ( U Z / G ) ) t S 1 A [ u ± 1 ] (( i ∗ ) t S 1 F , ( i ∗ ) t S 1 G ) ) ) .
In order to show that ( i ∗ ) t S 1 (i_*)^{tS^1} ( i ∗ ) t S 1 is fully faithful, we begin by showing that F ≃ 0 \mathcal F\simeq0 F ≃ 0 . Since A A A is Noetherian, it suffices by [NB92, Lemma 2.12] to show that F ⊗ A k ( p ) ≃ 0 \mathcal F\otimes_A k(\mathfrak p)\simeq0 F ⊗ A k ( p ) ≃ 0 for every prime ideal p ⊂ A \mathfrak p\subset A p ⊂ A , where k ( p ) k(\mathfrak p) k ( p ) denotes the residue field of p \mathfrak p p . In fact, since the algebraic closure k ( p ) → k ( p ) ‾ k(\mathfrak p)\to\overline{k(\mathfrak p)} k ( p ) → k ( p ) is faithfully flat, it suffices to show that F ⊗ A k ( p ) ‾ ≃ 0 \mathcal F\otimes_A\overline{k(\mathfrak p)}\simeq0 F ⊗ A k ( p ) ≃ 0 for each prime p ⊂ A \mathfrak p\subset A p ⊂ A .
For any stack X / k X/k X / k , let X k ( p ) : = X × Spec k Spec k ( p ) X_{k(\mathfrak p)}:=X\times_{\operatorname{Spec}k}\operatorname{Spec}k(\mathfrak p) X k ( p ) := X × Spec k Spec k ( p ) denote the base change of X X X along Spec k ( p ) → Spec k \operatorname{Spec}k(\mathfrak p)\to\operatorname{Spec}k Spec k ( p ) → Spec k . Our key observation is that the point of ( G / / G ) k ( p ) (G/\!/G)_{k(\mathfrak p)} ( G / / G ) k ( p ) given by the canonical map α \alpha α : Spec k ( p ‾ ) → G / / G k ( p ) ‾ k(\overline{\mathfrak{p}}) \to G//G_{\overline{k(\mathfrak{p})}} k ( p ) → G // G k ( p ) is closed, even though the original point p ∈ G / / G \mathfrak{p} \in G//G p ∈ G // G might not be. Let m α ⊂ A ⊗ k k ( p ) ‾ = : A k ( p ) ‾ \mathfrak{m}_{\alpha} \subset A \otimes_k \overline{k(\mathfrak{p})} =: A_{\overline{k(\mathfrak{p})}} m α ⊂ A ⊗ k k ( p ) =: A k ( p ) denote the maximal ideal associated to the closed point α ∈ ( G / / G ) k ( p ) ‾ \alpha\in(G//G)_{\overline{k(\mathfrak{p})}} α ∈ ( G // G ) k ( p ) . Crucially, the statement of Proposition 5.22 is formulated over an arbitrary algebraically closed base field k k k of characteristic zero, which was fixed by convention in § 2 —in particular it applies to the case of the G k ( p ) ‾ G_{\overline{k(\mathfrak{p})}} G k ( p ) -equivariant closed immersion i k ( p ) ‾ : Z k ( p ) ‾ ↪ U k ( p ) ‾ i_{\overline{k(\mathfrak{p})}}: Z_{\overline{k(\mathfrak{p})}} \hookrightarrow U_{\overline{k(\mathfrak{p})}} i k ( p ) : Z k ( p ) ↪ U k ( p ) .
Since A k ( p ) ‾ A_{\overline{k(\mathfrak{p})}} A k ( p ) is Noetherian (as A A A is a finitely-generated k k k -algebra), m α \mathfrak{m}_{\alpha} m α is finitely generated. Choose a presentation m α = ( f 1 , … , f n ) \mathfrak{m}_{\alpha}=(f_1,\ldots,f_n) m α = ( f 1 , … , f n ) by a finite set of generators, and let P : = K A k ( p ) ‾ ( f ) P:=K_{A_{\overline{k(\mathfrak{p})}}}(f) P := K A k ( p ) ( f ) be the associated Koszul complex, which, as a finite sequence of free A k ( p ) ‾ A_{\overline{k(\mathfrak{p})}} A k ( p ) -modules, belongs to Perf ( A k ( p ) ‾ ) \operatorname{Perf}(A_{\overline{k(\mathfrak{p})}}) Perf ( A k ( p ) ) .
On the other hand, since H 0 ( K A k ( p ) ‾ ( f ) ) ≃ A k ( p ) ‾ / m α H^0(K_{A_{\overline{k(\mathfrak{p})}}}(f))\simeq A_{\overline{k(\mathfrak{p})}}/\mathfrak{m}_{\alpha} H 0 ( K A k ( p ) ( f )) ≃ A k ( p ) / m α , it is in fact supported on the closed point α \alpha α . In particular, this means that the A k ( p ) ‾ A_{\overline{k(\mathfrak{p})}} A k ( p ) -module,
( F k ( p ) ‾ ⊗ A k ( p ) ‾ P ∨ ) ⊗ A k ( p ) ‾ P \left(F_{\overline{k(\mathfrak{p})}}\otimes_{A_{\overline{k(\mathfrak{p})}}}P^\vee\right)\otimes_{A_{\overline{k(\mathfrak{p})}}}P ( F k ( p ) ⊗ A k ( p ) P ∨ ) ⊗ A k ( p ) P
is supported at α ∈ ( G / / G ) k ( p ) ‾ \alpha\in(G//G)_{\overline{k(\mathfrak{p})}} α ∈ ( G // G ) k ( p ) . Now we establish a sequence of equivalences that allows us to conclude that F k ( p ) ‾ ⊗ A k ( p ) ‾ P ⊗ A k ( p ) ‾ P ∨ ≃ 0 F_{\overline{k(\mathfrak{p})}}\otimes_{A_{\overline{k(\mathfrak{p})}}}P\otimes_{A_{\overline{k(\mathfrak{p})}}}P^\vee\simeq0 F k ( p ) ⊗ A k ( p ) P ⊗ A k ( p ) P ∨ ≃ 0 :
F k ( p ) ‾ = res A [ u ± 1 ] / A ( ker ( Hom Coh ( L ( Z / G ) ) t S 1 A [ u ± 1 ] ( F , G ) → Hom Coh ( L ( U Z / G ) ) t S 1 A [ u ± 1 ] ( ( i ∗ ) t S 1 F , ( i ∗ ) t S 1 G ) ) ) A k ( p ) ‾ ≃ res A k ( p ) ‾ [ u ± 1 ] / A k ( p ) ‾ ( ker ( Hom Coh ( L ( Z / G ) ) t S 1 A [ u ± 1 ] ( F , G ) → Hom Coh ( L ( U Z / G ) ) t S 1 A [ u ± 1 ] ( ( i ∗ ) t S 1 F , ( i ∗ ) t S 1 G ) ) ) A k ( p ) ‾ [ u ± 1 ] (33) \begin{aligned}
F_{\overline{k(\mathfrak{p})}}
&=\operatorname{res}_{A[u^{\pm1}]/A}\left(\ker\left(\operatorname{Hom}_{\operatorname{Coh}(\mathcal{L}(Z/G))^{tS^1}}^{A[u^{\pm1}]}(\mathcal{F},\mathcal{G})\to\operatorname{Hom}_{\operatorname{Coh}(\mathcal{L}(U_Z/G))^{tS^1}}^{A[u^{\pm1}]}((i_*)^{tS^1}\mathcal{F},(i_*)^{tS^1}\mathcal{G})\right)\right)_{A_{\overline{k(\mathfrak{p})}}}\\
&\simeq\operatorname{res}_{A_{\overline{k(\mathfrak{p})}}[u^{\pm1}]/A_{\overline{k(\mathfrak{p})}}}\left(\ker\left(\operatorname{Hom}_{\operatorname{Coh}(\mathcal{L}(Z/G))^{tS^1}}^{A[u^{\pm1}]}(\mathcal{F},\mathcal{G})\to\operatorname{Hom}_{\operatorname{Coh}(\mathcal{L}(U_Z/G))^{tS^1}}^{A[u^{\pm1}]}((i_*)^{tS^1}\mathcal{F},(i_*)^{tS^1}\mathcal{G})\right)\right)_{A_{\overline{k(\mathfrak{p})}}[u^{\pm1}]} \tag*{(33)}
\end{aligned} F k ( p ) = res A [ u ± 1 ] / A ( ker ( Hom Coh ( L ( Z / G ) ) t S 1 A [ u ± 1 ] ( F , G ) → Hom Coh ( L ( U Z / G ) ) t S 1 A [ u ± 1 ] (( i ∗ ) t S 1 F , ( i ∗ ) t S 1 G ) ) ) A k ( p ) ≃ res A k ( p ) [ u ± 1 ] / A k ( p ) ( ker ( Hom Coh ( L ( Z / G ) ) t S 1 A [ u ± 1 ] ( F , G ) → Hom Coh ( L ( U Z / G ) ) t S 1 A [ u ± 1 ] (( i ∗ ) t S 1 F , ( i ∗ ) t S 1 G ) ) ) A k ( p ) [ u ± 1 ] (33)
≃ res A k ( p ) ‾ [ u ± 1 ] / A k ( p ) ‾ ( ker ( Hom Coh ( L ( Z / G ) ) t S 1 A [ u ± 1 ] ( F , G ) ⊗ A [ u ± 1 ] A k ( p ) ‾ [ u ± 1 ] → Hom Coh ( L ( U Z / G ) ) t S 1 A [ u ± 1 ] ( ( i ∗ ) t S 1 F , ( i ∗ ) t S 1 G ) ⊗ A [ u ± 1 ] A k ( p ) ‾ [ u ± 1 ] ) ) (34) \begin{aligned}
&\simeq\operatorname{res}_{A_{\overline{k(\mathfrak{p})}}[u^{\pm1}]/A_{\overline{k(\mathfrak{p})}}}\left(\ker\left(\operatorname{Hom}_{\operatorname{Coh}(\mathcal{L}(Z/G))^{tS^1}}^{A[u^{\pm1}]}(\mathcal{F},\mathcal{G})\otimes_{A[u^{\pm1}]}A_{\overline{k(\mathfrak{p})}}[u^{\pm1}]\to\operatorname{Hom}_{\operatorname{Coh}(\mathcal{L}(U_Z/G))^{tS^1}}^{A[u^{\pm1}]}((i_*)^{tS^1}\mathcal{F},(i_*)^{tS^1}\mathcal{G})\otimes_{A[u^{\pm1}]}A_{\overline{k(\mathfrak{p})}}[u^{\pm1}]\right)\right) \tag*{(34)}
\end{aligned} ≃ res A k ( p ) [ u ± 1 ] / A k ( p ) ( ker ( Hom Coh ( L ( Z / G ) ) t S 1 A [ u ± 1 ] ( F , G ) ⊗ A [ u ± 1 ] A k ( p ) [ u ± 1 ] → Hom Coh ( L ( U Z / G ) ) t S 1 A [ u ± 1 ] (( i ∗ ) t S 1 F , ( i ∗ ) t S 1 G ) ⊗ A [ u ± 1 ] A k ( p ) [ u ± 1 ] ) ) (34)
\begin{aligned}
&\simeq\operatorname{res}_{A_{\overline{k(\mathfrak{p})}}[u^{\pm1}]/A_{\overline{k(\mathfrak{p})}}}\left(\ker\left(\operatorname{Hom}_{\operatorname{Coh}(\mathcal{L}(Z/G)_{A_{\overline{k(\mathfrak{p})}})^{tS^1}}^{A_{\overline{k(\mathfrak{p})}}[u^{\pm1}]}(\mathcal{F}_{A_{\overline{k(\mathfrak{p})}}},\mathcal{G}_{A_{\overline{k(\mathfrak{p})}}})\to\operatorname{Hom}_{\operatorname{Coh}(\mathcal{L}(U_Z/G)_{A_{\overline{k(\mathfrak{p})}})^{tS^1}}^{A_{\overline{k(\mathfrak{p})}}[u^{\pm1}]}((i_*)^{tS^1}\mathcal{F}_{A_{\overline{k(\mathfrak{p})}}},(i_*)^{tS^1}\mathcal{G}_{A_{\overline{k(\mathfrak{p})}}})\right)\right)
\end{aligned}}} \tag*{(35)}\end{aligned}
\begin{aligned}
&\simeq\operatorname{res}_{A_{\overline{k(\mathfrak{p})}}[u^{\pm1}]/A_{\overline{k(\mathfrak{p})}}}\left(\ker\left(\operatorname{Hom}_{\operatorname{Coh}(\mathcal{L}(Z/G)_{A_{\overline{k(\mathfrak{p})}})^{tS^1}}^{A_{\overline{k(\mathfrak{p})}}[u^{\pm1}]}(\mathcal{F}_{A_{\overline{k(\mathfrak{p})}}},\mathcal{G}_{A_{\overline{k(\mathfrak{p})}}})\to\operatorname{Hom}_{\operatorname{Coh}(\mathcal{L}(U_Z/G)_{A_{\overline{k(\mathfrak{p})}})^{tS^1}}^{A_{\overline{k(\mathfrak{p})}}[u^{\pm1}]}((i_{\mathfrak{p}*})^{tS^1}(\mathcal{F}_{A_{\overline{k(\mathfrak{p})}}}),(i_{\mathfrak{p}*})^{tS^1}(\mathcal{G}_{A_{\overline{k(\mathfrak{p})}}}))\right)\right).
\end{aligned}}} \tag*{(36)}\end{aligned}
where
the objects ( L ( Z / G ) ) A k ( p ) ‾ (\mathcal{L}(Z/G))_{A_{\overline{k(\mathfrak{p})}}} ( L ( Z / G ) ) A k ( p ) and ( L ( U ^ Z / G ) ) A k ( p ) ‾ (\mathcal{L}(\widehat{U}_Z/G))_{A_{\overline{k(\mathfrak{p})}}} ( L ( U Z / G ) ) A k ( p ) appearing in (35) denote the base changes of L ( Z / G ) , L ( U ^ Z / G ) ∈ Stk / A \mathcal{L}(Z/G),\mathcal{L}(\widehat{U}_Z/G)\in\operatorname{Stk}/A L ( Z / G ) , L ( U Z / G ) ∈ Stk / A , respectively, along the map Spec A k ( p ) ‾ → Spec A \operatorname{Spec}A_{\overline{k(\mathfrak{p})}}\to\operatorname{Spec}A Spec A k ( p ) → Spec A ;
the map i p : L ( Z / G ) A k ( p ) ‾ → L ( U ^ Z / G ) A k ( p ) ‾ i_{\mathfrak{p}}:\mathcal{L}(Z/G)_{A_{\overline{k(\mathfrak{p})}}}\to\mathcal{L}(\widehat{U}_Z/G)_{A_{\overline{k(\mathfrak{p})}}} i p : L ( Z / G ) A k ( p ) → L ( U Z / G ) A k ( p ) appearing in (36) denotes the base change of i i i along Spec A k ( p ) ‾ → Spec A \operatorname{Spec}A_{\overline{k(\mathfrak{p})}}\to\operatorname{Spec}A Spec A k ( p ) → Spec A ;
and where
equivalence (33) holds because restriction of scalars commutes with base change;43
equivalence (34) holds because A A A –mod is a stable category (meaning kernels are also finite colimits), and − ⊗ A [ u ± 1 ] A k ( p ) ‾ [ u ± 1 ] -\otimes_{A[u^{\pm1}]}A_{\overline{k(\mathfrak{p})}}[u^{\pm1}] − ⊗ A [ u ± 1 ] A k ( p ) [ u ± 1 ] is a left adjoint;
equivalence (35) follows from Proposition 6.3 and the fact that restriction of scalars res A k ( p ) ‾ [ u ± 1 ] / A k ( p ) ‾ \operatorname{res}_{A_{\overline{k(\mathfrak{p})}}[u^{\pm1}]/A_{\overline{k(\mathfrak{p})}}} res A k ( p ) [ u ± 1 ] / A k ( p ) satisfies the projection formula
res A k ( p ) ‾ [ u ± 1 ] / A k ( p ) ‾ ( M ) ⊗ A N ≃ res A k ( p ) ‾ [ u ± 1 ] / A k ( p ) ‾ ( M ⊗ A [ u ± 1 ] A k ( p ) ‾ [ u ± 1 ] ) ; \operatorname{res}_{A_{\overline{k(\mathfrak{p})}}[u^{\pm1}]/A_{\overline{k(\mathfrak{p})}}}(M)\otimes_A N\simeq\operatorname{res}_{A_{\overline{k(\mathfrak{p})}}[u^{\pm1}]/A_{\overline{k(\mathfrak{p})}}}(M\otimes_{A[u^{\pm1}]}A_{\overline{k(\mathfrak{p})}}[u^{\pm1}]); res A k ( p ) [ u ± 1 ] / A k ( p ) ( M ) ⊗ A N ≃ res A k ( p ) [ u ± 1 ] / A k ( p ) ( M ⊗ A [ u ± 1 ] A k ( p ) [ u ± 1 ]) ;
For any stack X ∈ Stk / A X \in\operatorname{Stk}/A X ∈ Stk / A , every square in the commutative diagram
X A k ( p ) → Spec A k ( p ) → Spec k ( p ) ↓ ↓ ↓ X → Spec A → Spec k \begin{CD}
X_{A_{k(\mathfrak p)}} @>>> \operatorname{Spec} A_{k(\mathfrak p)} @>>> \operatorname{Spec} k(\mathfrak p) \\
@VVV @VVV @VVV \\
X @>>> \operatorname{Spec} A @>>> \operatorname{Spec} k
\end{CD} X A k ( p ) ↓ ⏐ X Spec A k ( p ) ↓ ⏐ Spec A Spec k ( p ) ↓ ⏐ Spec k
is Cartesian. Thus, we have a canonical isomorphism
X A k ( p ) ≃ X ‾ k ( p ) . X_{A_{k(\mathfrak p)}} \simeq\overline{X}_{k(\mathfrak p)}. X A k ( p ) ≃ X k ( p ) .
Now, for any X ∈ Stk k X \in\operatorname{Stk}_k X ∈ Stk k , we have ( X k ( p ) ) d R ≃ ( X d R ) k ( p ) (X_{k(\mathfrak p)})_{\mathrm{dR}} \simeq(X_{\mathrm{dR}})_{k(\mathfrak p)} ( X k ( p ) ) dR ≃ ( X dR ) k ( p ) , from which we deduce
( U Z ‾ ) k ( p ) ≃ ( U k ( p ) ‾ ) Z k ( p ) . (\overline{U_Z})_{k(\mathfrak p)} \simeq\left(\overline{U_{k(\mathfrak p)}}\right)_{Z_{k(\mathfrak p)}}. ( U Z ) k ( p ) ≃ ( U k ( p ) ) Z k ( p ) .
To compute the loop spaces of these stacks, we need to distinguish between the loop space functor L \mathcal{L} L with respect to the base Spec k \operatorname{Spec} k Spec k and that with respect to the base Spec k ( p ) \operatorname{Spec} k(\mathfrak p) Spec k ( p ) . Denote the former by L : = L k \mathcal{L} := \mathcal{L}_k L := L k and the latter by L k ( p ) \mathcal{L}_{k(\mathfrak p)} L k ( p ) . Using that L k \mathcal{L}_k L k and L k ( p ) \mathcal{L}_{k(\mathfrak p)} L k ( p ) preserve products, we therefore obtain equivalences
L k ( Z / G ) A k ( p ) ≃ L k ( Z / G ) k ( p ) ≃ L k ( p ) ( ( Z / G ) k ( p ) ) ≃ L k ( p ) ( Z k ( p ) / G k ( p ) ) , \mathcal{L}_k(Z/G)_{A_{k(\mathfrak p)}} \simeq\mathcal{L}_k(Z/G)_{k(\mathfrak p)} \simeq\mathcal{L}_{k(\mathfrak p)}\left((Z/G)_{k(\mathfrak p)}\right) \simeq\mathcal{L}_{k(\mathfrak p)}(Z_{k(\mathfrak p)}/G_{k(\mathfrak p)}), L k ( Z / G ) A k ( p ) ≃ L k ( Z / G ) k ( p ) ≃ L k ( p ) ( ( Z / G ) k ( p ) ) ≃ L k ( p ) ( Z k ( p ) / G k ( p ) ) ,
and
L k ( U Z ‾ / G ) A k ( p ) ≃ L k ( U Z ‾ / G ) k ( p ) ≃ L k ( p ) ( ( U Z ‾ / G ) k ( p ) ) ≃ L k ( p ) ( U k ( p ) ‾ Z k ( p ) / G k ( p ) ) . \mathcal{L}_k(\overline{U_Z}/G)_{A_{k(\mathfrak p)}} \simeq\mathcal{L}_k(\overline{U_Z}/G)_{k(\mathfrak p)} \simeq\mathcal{L}_{k(\mathfrak p)}\left((\overline{U_Z}/G)_{k(\mathfrak p)}\right) \simeq\mathcal{L}_{k(\mathfrak p)}\left(\overline{U_{k(\mathfrak p)}}_{Z_{k(\mathfrak p)}}/G_{k(\mathfrak p)}\right). L k ( U Z / G ) A k ( p ) ≃ L k ( U Z / G ) k ( p ) ≃ L k ( p ) ( ( U Z / G ) k ( p ) ) ≃ L k ( p ) ( U k ( p ) Z k ( p ) / G k ( p ) ) .
With these equivalences in hand, we continue by computing the expression (35) ⊗ A k ( p ) P ⊗ A k ( p ) P ∨ \otimes_{A_{k(\mathfrak p)}}P\otimes_{A_{k(\mathfrak p)}}P^\vee ⊗ A k ( p ) P ⊗ A k ( p ) P ∨ .
In what follows,
C : = Coh ( L k ( p ) ( Z k ( p ) ‾ / G k ( p ) ) ) , D : = Coh ( L k ( p ) ( U k ( p ) ‾ Z k ( p ) / G k ( p ) ) ) . ≃ res A k ( p ) [ u ± 1 ] / A k ( p ) ( ker ( Hom C A k ( p ) [ u ± 1 ] t S 1 ( F k ( p ) , G k ( p ) ) → Hom D A k ( p ) [ u ± 1 ] t S 1 ( ( i p ∗ ) t S 1 F k ( p ) , ( i p ∗ ) t S 1 G k ( p ) ) ) ) ⊗ A k ( p ) P ⊗ A k ( p ) P ∨ (37) \begin{aligned}
\mathcal{C} := \operatorname{Coh}\left(\mathcal{L}_{k(\mathfrak p)}\left(\overline{Z_{k(\mathfrak p)}}/G_{k(\mathfrak p)}\right)\right), \qquad\mathcal{D} := \operatorname{Coh}\left(\mathcal{L}_{k(\mathfrak p)}\left(\overline{U_{k(\mathfrak p)}}_{Z_{k(\mathfrak p)}}/G_{k(\mathfrak p)}\right)\right). \\
&\simeq\operatorname{res}_{A_{k(\mathfrak p)}[u^{\pm1}]/A_{k(\mathfrak p)}}\left(\ker\left(\operatorname{Hom}_{\mathcal{C}^{tS^1}_{A_{k(\mathfrak p)}[u^{\pm1}]}}\left(\mathcal{F}_{k(\mathfrak p)},\mathcal{G}_{k(\mathfrak p)}\right)\to\operatorname{Hom}_{\mathcal{D}^{tS^1}_{A_{k(\mathfrak p)}[u^{\pm1}]}}\left((i_{p*})^{tS^1}\mathcal{F}_{k(\mathfrak p)},(i_{p*})^{tS^1}\mathcal{G}_{k(\mathfrak p)}\right)\right)\right)\otimes_{A_{k(\mathfrak p)}}P\otimes_{A_{k(\mathfrak p)}}P^\vee \tag*{(37)}
\end{aligned} C := Coh ( L k ( p ) ( Z k ( p ) / G k ( p ) ) ) , D := Coh ( L k ( p ) ( U k ( p ) Z k ( p ) / G k ( p ) ) ) . ≃ res A k ( p ) [ u ± 1 ] / A k ( p ) ( ker ( Hom C A k ( p ) [ u ± 1 ] t S 1 ( F k ( p ) , G k ( p ) ) → Hom D A k ( p ) [ u ± 1 ] t S 1 ( ( i p ∗ ) t S 1 F k ( p ) , ( i p ∗ ) t S 1 G k ( p ) ) ) ) ⊗ A k ( p ) P ⊗ A k ( p ) P ∨ (37)
≃ ker ( res A k ( p ) [ u ± 1 ] / A k ( p ) ( Hom C A k ( p ) [ u ± 1 ] t S 1 ( F k ( p ) , G k ( p ) ) → Hom D A k ( p ) [ u ± 1 ] t S 1 ( ( i p ∗ ) t S 1 F k ( p ) , ( i p ∗ ) t S 1 G k ( p ) ) ) ) ⊗ A k ( p ) P ⊗ A k ( p ) P ∨ ≃ ker ( res A k ( p ) [ u ± 1 ] / A k ( p ) ( ( Hom C A k ( p ) [ u ± 1 ] t S 1 ( η ∗ ( F k ( p ) ) , η ∗ ( G k ( p ) ) ) ) ⊗ A k ( p ) [ u ± 1 ] η ∗ P [ u ] ⊗ A k ( p ) [ u ± 1 ] η ∗ P ∨ [ u ] (38) \begin{aligned}
\simeq\ker\left(\operatorname{res}_{A_{k(\mathfrak p)}[u^{\pm1}]/A_{k(\mathfrak p)}}\left(\operatorname{Hom}_{\mathcal{C}^{tS^1}_{A_{k(\mathfrak p)}[u^{\pm1}]}}\left(\mathcal{F}_{k(\mathfrak p)},\mathcal{G}_{k(\mathfrak p)}\right)\to\operatorname{Hom}_{\mathcal{D}^{tS^1}_{A_{k(\mathfrak p)}[u^{\pm1}]}}\left((i_{p*})^{tS^1}\mathcal{F}_{k(\mathfrak p)},(i_{p*})^{tS^1}\mathcal{G}_{k(\mathfrak p)}\right)\right)\right)\otimes_{A_{k(\mathfrak p)}}P\otimes_{A_{k(\mathfrak p)}}P^\vee\\
&\simeq\ker(\operatorname{res}_{A_{k(\mathfrak p)}[u^{\pm1}]/A_{k(\mathfrak p)}}((\operatorname{Hom}_{\mathcal{C}^{tS^1}_{A_{k(\mathfrak p)}[u^{\pm1}]}}(\eta^*(\mathcal{F}_{k(\mathfrak p)}),\eta^*(\mathcal{G}_{k(\mathfrak p)})))\otimes_{A_{k(\mathfrak p)}[u^{\pm1}]}\eta^*P[u]\otimes_{A_{k(\mathfrak p)}[u^{\pm1}]}\eta^*P^\vee[u] \tag*{(38)}
\end{aligned} ≃ ker ( res A k ( p ) [ u ± 1 ] / A k ( p ) ( Hom C A k ( p ) [ u ± 1 ] t S 1 ( F k ( p ) , G k ( p ) ) → Hom D A k ( p ) [ u ± 1 ] t S 1 ( ( i p ∗ ) t S 1 F k ( p ) , ( i p ∗ ) t S 1 G k ( p ) ) ) ) ⊗ A k ( p ) P ⊗ A k ( p ) P ∨ ≃ ker ( res A k ( p ) [ u ± 1 ] / A k ( p ) (( Hom C A k ( p ) [ u ± 1 ] t S 1 ( η ∗ ( F k ( p ) ) , η ∗ ( G k ( p ) ))) ⊗ A k ( p ) [ u ± 1 ] η ∗ P [ u ] ⊗ A k ( p ) [ u ± 1 ] η ∗ P ∨ [ u ] (38)
⟶ ( Hom D A k ( p ) [ u ± 1 ] t S 1 ( ( i p ∗ ) t S 1 η ∗ ( F k ( p ) ) , ( i p ∗ ) t S 1 η ∗ ( G k ( p ) ) ) ) ⊗ A k ( p ) [ u ± 1 ] η ∗ P [ u ] ⊗ A k ( p ) [ u ± 1 ] η ∗ P ∨ [ u ] ) ) (39) \longrightarrow(\operatorname{Hom}_{\mathcal{D}^{tS^1}_{A_{k(\mathfrak p)}[u^{\pm1}]}}((i_{p*})^{tS^1}\eta^*(\mathcal{F}_{k(\mathfrak p)}),(i_{p*})^{tS^1}\eta^*(\mathcal{G}_{k(\mathfrak p)})))\otimes_{A_{k(\mathfrak p)}[u^{\pm1}]}\eta^*P[u]\otimes_{A_{k(\mathfrak p)}[u^{\pm1}]}\eta^*P^\vee[u])) \tag*{(39)} ⟶ ( Hom D A k ( p ) [ u ± 1 ] t S 1 (( i p ∗ ) t S 1 η ∗ ( F k ( p ) ) , ( i p ∗ ) t S 1 η ∗ ( G k ( p ) ))) ⊗ A k ( p ) [ u ± 1 ] η ∗ P [ u ] ⊗ A k ( p ) [ u ± 1 ] η ∗ P ∨ [ u ])) (39)
≃ ker ( Hom C A k ( p ) t S 1 ( η ∗ F A k ( p ) , η ∗ G A k ( p ) ) ⊗ A k ( p ) P ⊗ A k ( p ) P ∨ → Hom D A k ( p ) t S 1 ( η ∗ ( i p ∗ ) A k ( p ) F A k ( p ) , η ∗ ( i p ∗ ) A k ( p ) G A k ( p ) ) ⊗ A k ( p ) P ⊗ A k ( p ) P ∨ ) ≃ ker ( Hom C A k ( p ) t S 1 ( η ∗ ( F k ( p ) ) ⊗ A k ( p ) [ u ± 1 ] η ∗ P [ u ] , η ∗ ( G k ( p ) ) ⊗ A k ( p ) [ u ± 1 ] η ∗ P [ u ] ) (40) \begin{aligned}
\simeq\ker\left(\operatorname{Hom}_{\mathcal{C}^{tS^1}_{A_{k(\mathfrak p)}}}\left(\eta^*\mathcal{F}_{A_{k(\mathfrak p)}},\eta^*\mathcal{G}_{A_{k(\mathfrak p)}}\right)\otimes_{A_{k(\mathfrak p)}}P\otimes_{A_{k(\mathfrak p)}}P^\vee\to\operatorname{Hom}_{\mathcal{D}^{tS^1}_{A_{k(\mathfrak p)}}}\left(\eta^*(i_{p*})_{A_{k(\mathfrak p)}}\mathcal{F}_{A_{k(\mathfrak p)}},\eta^*(i_{p*})_{A_{k(\mathfrak p)}}\mathcal{G}_{A_{k(\mathfrak p)}}\right)\otimes_{A_{k(\mathfrak p)}}P\otimes_{A_{k(\mathfrak p)}}P^\vee\right) \\
&\simeq\ker(\operatorname{Hom}_{\mathcal{C}^{tS^1}_{A_{k(\mathfrak p)}}}(\eta^*(\mathcal{F}_{k(\mathfrak p)})\otimes_{A_{k(\mathfrak p)}[u^{\pm1}]}\eta^*P[u],\eta^*(\mathcal{G}_{k(\mathfrak p)})\otimes_{A_{k(\mathfrak p)}[u^{\pm1}]}\eta^*P[u]) \tag*{(40)}
\end{aligned} ≃ ker ( Hom C A k ( p ) t S 1 ( η ∗ F A k ( p ) , η ∗ G A k ( p ) ) ⊗ A k ( p ) P ⊗ A k ( p ) P ∨ → Hom D A k ( p ) t S 1 ( η ∗ ( i p ∗ ) A k ( p ) F A k ( p ) , η ∗ ( i p ∗ ) A k ( p ) G A k ( p ) ) ⊗ A k ( p ) P ⊗ A k ( p ) P ∨ ) ≃ ker ( Hom C A k ( p ) t S 1 ( η ∗ ( F k ( p ) ) ⊗ A k ( p ) [ u ± 1 ] η ∗ P [ u ] , η ∗ ( G k ( p ) ) ⊗ A k ( p ) [ u ± 1 ] η ∗ P [ u ]) (40)
⟶ Hom D A k ( p ) t S 1 ( ( i p ∗ ) t S 1 ( η ∗ ( F k ( p ) ) ⊗ A k ( p ) [ u ± 1 ] η ∗ P [ u ] ) , ( i p ∗ ) t S 1 ( η ∗ ( G k ( p ) ) ⊗ A k ( p ) [ u ± 1 ] η ∗ P [ u ] ) ) ) (41) \longrightarrow\operatorname{Hom}_{\mathcal{D}^{tS^1}_{A_{k(\mathfrak p)}}}((i_{p*})^{tS^1}(\eta^*(\mathcal{F}_{k(\mathfrak p)})\otimes_{A_{k(\mathfrak p)}[u^{\pm1}]}\eta^*P[u]),(i_{p*})^{tS^1}(\eta^*(\mathcal{G}_{k(\mathfrak p)})\otimes_{A_{k(\mathfrak p)}[u^{\pm1}]}\eta^*P[u]))) \tag*{(41)} ⟶ Hom D A k ( p ) t S 1 (( i p ∗ ) t S 1 ( η ∗ ( F k ( p ) ) ⊗ A k ( p ) [ u ± 1 ] η ∗ P [ u ]) , ( i p ∗ ) t S 1 ( η ∗ ( G k ( p ) ) ⊗ A k ( p ) [ u ± 1 ] η ∗ P [ u ]))) (41)
where
equivalence (37) follows from the preceding discussion;
equivalence (38) holds because, as a right adjoint, res A [ u ± 1 ] / A \operatorname{res}_{A[u^{\pm1}]/A} res A [ u ± 1 ] / A commutes with limits (e.g. kernels)
(39) holds because restriction of scalars along B → C B\to C B → C is tensoring with the ( B , C ) (B,C) ( B , C ) -bimodule C C C , hence commutes with tensor products over the base B B B ;
equivalence (40) follows from the fact that the A k ( p ) ‾ A_{\overline{k(p)}} A k ( p ) -linear Hom objects are equivalent to image of the A k ( p ) ‾ [ u ± 1 ] A_{\overline{k(p)}}[u^{\pm1}] A k ( p ) [ u ± 1 ] -linear Hom objects under restriction of scalars;
and equivalence (41) follows from Lemma 6.8 .
Because P , P ∨ P,P^{\vee} P , P ∨ are bounded, the sheaves F k ( p ) ‾ ⊗ A k ( p ) ‾ P , G k ( p ) ‾ ⊗ A k ( p ) ‾ P \mathcal{F}_{\overline{k(p)}}\otimes_{A_{\overline{k(p)}}}P,\mathcal{G}_{\overline{k(p)}}\otimes_{A_{\overline{k(p)}}}P F k ( p ) ⊗ A k ( p ) P , G k ( p ) ⊗ A k ( p ) P are coherent, as well as supported on the fiber over α \alpha α by the discussion above, so they belong to Coh ( ( L k ( p ) ‾ u ( Z k ( p ) ‾ / G k ( p ) ‾ ) ) t S 1 \operatorname{Coh}((\mathcal{L}_{\overline{k(p)}}^{u}(Z_{\overline{k(p)}}/G_{\overline{k(p)}}))^{tS^{1}} Coh (( L k ( p ) u ( Z k ( p ) / G k ( p ) ) ) t S 1 . 44 Thus, we can apply Proposition 5.22 to obtain that
F k ( p ) ‾ ⊗ A k ( p ) ‾ P ⊗ A k ( p ) ‾ P ∨ ≃ 0. (42) \mathcal{F}_{\overline{k(p)}}\otimes_{A_{\overline{k(p)}}}P\otimes_{A_{\overline{k(p)}}}P^{\vee}\simeq0. \tag*{(42)} F k ( p ) ⊗ A k ( p ) P ⊗ A k ( p ) P ∨ ≃ 0. (42)
Base changing this A k ( p ) ‾ A_{\overline{k(p)}} A k ( p ) -module along A k ( p ) ‾ → k ( p ) ‾ A_{\overline{k(p)}}\to\overline{k(p)} A k ( p ) → k ( p ) , we obtain an object that belongs to the thick subcategory generated by F k ( p ) ‾ \mathcal{F}_{\overline{k(p)}} F k ( p ) , again by perfectness of P P P over A k ( p ) ‾ A_{\overline{k(p)}} A k ( p ) . By definition of the Koszul complex P P P , base changing it along A k ( p ) ‾ → k ( p ) ‾ A_{\overline{k(p)}}\to\overline{k(p)} A k ( p ) → k ( p ) sends the Koszul differential to zero. Thus, we see that the complex (42) is a direct sum of shifts of F ⊗ A k ( p ) ‾ F\otimes_A\overline{k(p)} F ⊗ A k ( p ) . We thus conclude that
F ⊗ A k ( p ) ‾ = F k ( p ) ‾ ≃ 0 , F\otimes_A\overline{k(p)}=F_{\overline{k(p)}}\simeq0, F ⊗ A k ( p ) = F k ( p ) ≃ 0 ,
as desired.
Finally, in order to conclude, we apply Lemma 3.10 . □
Essential surjectivity
Proposition 6.22. Let i : Z ↪ X i:Z\hookrightarrow X i : Z ↪ X be a closed substack of a stack X X X such that X X X is an algebraic stack almost of finite type with affine diagonal and i i i is almost of finite presentation. Denote by i ^ : Z → X ^ Z \widehat{i}:Z\to\widehat{X}_Z i : Z → X Z the inclusion of Z Z Z into the formal completion, and by L i ^ : L Z → L X ^ Z \mathcal{L}\widehat{i}:\mathcal{L}Z\to\mathcal{L}\widehat{X}_Z L i : L Z → L X Z the corresponding map on loop spaces. Then the image of the functor ( L i ^ ∗ ) t S 1 : Coh ( L Z ) t S 1 → Coh ( L X ^ Z ) t S 1 (\mathcal{L}\widehat{i}_*)^{tS^1}:\operatorname{Coh}(\mathcal{L}Z)^{tS^1}\to\operatorname{Coh}(\mathcal{L}\widehat{X}_Z)^{tS^1} ( L i ∗ ) t S 1 : Coh ( L Z ) t S 1 → Coh ( L X Z ) t S 1 thickly generates the target.
Proof. We begin by showing that Coh ( L X ^ Z ) S 1 \operatorname{Coh}(\mathcal{L}\widehat{X}_Z)^{S^1} Coh ( L X Z ) S 1 is generated as a thick category by the image of the restriction of ( L i ^ ∗ ) t T a t e (\mathcal{L}\widehat{i}_*)^{tTate} ( L i ∗ ) tT a t e to Coh ( L Z ) S 1 \operatorname{Coh}(\mathcal{L}Z)^{S^1} Coh ( L Z ) S 1 . For this, it suffices by Lemma 6.23 to show that Coh ( L X ^ Z ) S 1 \operatorname{Coh}(\mathcal{L}\widehat{X}_Z)^{S^1} Coh ( L X Z ) S 1 is generated as a thick category by the image of the functor Coh ( L Z ) S 1 → Coh ( L X ^ Z ) S 1 \operatorname{Coh}(\mathcal{L}Z)^{S^1}\to\operatorname{Coh}(\mathcal{L}\widehat{X}_Z)^{S^1} Coh ( L Z ) S 1 → Coh ( L X Z ) S 1 .
From the natural isomorphism L X ^ Z → L X ^ Z \mathcal{L}\widehat{X}_Z\to\mathcal{L}\widehat{X}_Z L X Z → L X Z (Lemma 3.57), we have the equivalence Coh ( L X ^ Z ) ≃ Coh ( L X ^ Z ) \operatorname{Coh}(\mathcal{L}\widehat{X}_Z)\simeq\operatorname{Coh}(\widehat{\mathcal{L}X}_Z) Coh ( L X Z ) ≃ Coh ( L X Z ) . We recall that pushforward along the map L X ^ Z → L X \widehat{\mathcal{L}X}_Z\to\mathcal{L}X L X Z → L X induces an equivalence Coh ( L X ^ Z ) → ∼ Coh Z ( L X ) ⊂ Coh ( L X ) \operatorname{Coh}(\widehat{\mathcal{L}X}_Z)\xrightarrow{\sim}\operatorname{Coh}_Z(\mathcal{L}X)\subset\operatorname{Coh}(\mathcal{L}X) Coh ( L X Z ) ∼ Coh Z ( L X ) ⊂ Coh ( L X ) , so it suffices to show that the target of Coh ( L Z ) S 1 → Coh Z ( L X ) S 1 \operatorname{Coh}(\mathcal{L}Z)^{S^1}\to\operatorname{Coh}_Z(\mathcal{L}X)^{S^1} Coh ( L Z ) S 1 → Coh Z ( L X ) S 1 induced by L Z → L X \mathcal{L}Z\to\mathcal{L}X L Z → L X is generated as a thick category by the image.
The t t t -structures on both Coh ( L Z ) S 1 \operatorname{Coh}(\mathcal{L}Z)^{S^1} Coh ( L Z ) S 1 and Coh Z ( L X ) S 1 ≃ Coh ( L X ^ Z ) t S 1 \operatorname{Coh}_Z(\mathcal{L}X)^{S^1}\simeq\operatorname{Coh}(\widehat{\mathcal{L}X}_Z)^{tS^1} Coh Z ( L X ) S 1 ≃ Coh ( L X Z ) t S 1 are bounded since they’re induced by those on Coh ( L Z ) \operatorname{Coh}(\mathcal{L}Z) Coh ( L Z ) and Coh Z ( L X ) \operatorname{Coh}_Z(\mathcal{L}X) Coh Z ( L X ) , respectively, so it follows from Lemma 6.27 that it suffices just to show that the heart ( Coh Z ( L X ) S 1 ) ♡ (\operatorname{Coh}_Z(\mathcal{L}X)^{S^1})^{\heartsuit} ( Coh Z ( L X ) S 1 ) ♡ is generated by the image of Coh ( L Z ) S 1 \operatorname{Coh}(\mathcal{L}Z)^{S^1} Coh ( L Z ) S 1 .
Note that Lemma 3.15 and Lemma 6.24 together imply that
( Coh Z ( L X ) S 1 ) ♡ ≃ ( Coh ( L Z ) c l ( ( L X ) c l ) ♡ ) S 1 . (\operatorname{Coh}_Z(\mathcal{L}X)^{S^1})^{\heartsuit}\simeq(\operatorname{Coh}_{(\mathcal{L}Z)^{\mathrm{cl}}}((\mathcal{L}X)^{\mathrm{cl}})^{\heartsuit})^{S^1}. ( Coh Z ( L X ) S 1 ) ♡ ≃ ( Coh ( L Z ) cl (( L X ) cl ) ♡ ) S 1 .
There is, of course, a functor oblv : ( Coh ( L Z ) c l ( ( L X ) c l ) ♡ ) S 1 → Coh ( L Z ) c l ( ( L X ) c l ) ♡ \operatorname{oblv}:(\operatorname{Coh}_{(\mathcal{L}Z)^{\mathrm{cl}}}((\mathcal{L}X)^{\mathrm{cl}})^{\heartsuit})^{S^1}\to\operatorname{Coh}_{(\mathcal{L}Z)^{\mathrm{cl}}}((\mathcal{L}X)^{\mathrm{cl}})^{\heartsuit} oblv : ( Coh ( L Z ) cl (( L X ) cl ) ♡ ) S 1 → Coh ( L Z ) cl (( L X ) cl ) ♡ which forgets the S 1 S^1 S 1 -equivariant structure of an object on the left. Since X X X is a 1 1 1 -stack, L X \mathcal{L}X L X is also a 1 1 1 -stack, meaning that the restriction of L X : CAlg k c n → Spc \mathcal{L}X:\operatorname{CAlg}^{\mathrm{cn}}_k\to\operatorname{Spc} L X : CAlg k cn → Spc to classical rings CAlg k c l → Spc \operatorname{CAlg}^{\mathrm{cl}}_k\to\operatorname{Spc} CAlg k cl → Spc is an ordinary stack in the sense of e.g. [Ols16]. In this case, ( L X ) c l (\mathcal{L}X)^{\mathrm{cl}} ( L X ) cl possesses a structure sheaf of commutative rings, O ( L X ) c l \mathcal{O}_{(\mathcal{L}X)^{\mathrm{cl}}} O ( L X ) cl . The t t t -structure on Coh ( ( L X ) c l ) \operatorname{Coh}((\mathcal{L}X)^{\mathrm{cl}}) Coh (( L X ) cl ) , hence on Coh ( L Z ) c l ( ( L X ) c l ) \operatorname{Coh}_{(\mathcal{L}Z)^{\mathrm{cl}}}((\mathcal{L}X)^{\mathrm{cl}}) Coh ( L Z ) cl (( L X ) cl ) , is compatible with the fully faithful embedding Coh ( ( L X ) c l ) ↪ O ( L X ) c l -mod \operatorname{Coh}((\mathcal{L}X)^{\mathrm{cl}})\hookrightarrow\mathcal{O}_{(\mathcal{L}X)^{\mathrm{cl}}}\text{-mod} Coh (( L X ) cl ) ↪ O ( L X ) cl -mod . The t t t -structure on the latter is the canonical one on any category of modules over a sheaf of commutative rings, so an object of F ∈ Coh ( L Z ) c l ( ( L X ) c l ) ♡ \mathcal{F}\in\operatorname{Coh}_{(\mathcal{L}Z)^{\mathrm{cl}}}((\mathcal{L}X)^{\mathrm{cl}})^{\heartsuit} F ∈ Coh ( L Z ) cl (( L X ) cl ) ♡ is a discrete module over O ( L X ) c l \mathcal{O}_{(\mathcal{L}X)^{\mathrm{cl}}} O ( L X ) cl . Moreover, because F \mathcal{F} F is supported on ( L Z ) c l (\mathcal{L}Z)^{\mathrm{cl}} ( L Z ) cl , the defining ideal of ( L Z ) c l (\mathcal{L}Z)^{\mathrm{cl}} ( L Z ) cl ,
I : = ker ( O ( L X ) c l ⟶ ( L i ) ∗ O ( L Z ) c l ) , I := \ker\left(\mathcal{O}_{(\mathcal{L}X)^{\mathrm{cl}}} \longrightarrow(\mathcal{L}i)_*\mathcal{O}_{(\mathcal{L}Z)^{\mathrm{cl}}}\right), I := ker ( O ( L X ) cl ⟶ ( L i ) ∗ O ( L Z ) cl ) ,
is contained in Ann F \sqrt{\operatorname{Ann}\mathcal{F}} Ann F . By Lemma 6.28, the classical stack ( L X ) c l (\mathcal{L}X)^{\mathrm{cl}} ( L X ) cl is Noetherian, so I I I is locally finitely generated. For a local chart U a U_a U a , let I a = ( f a 1 , … , f a m ) I_a=(f_{a1},\ldots,f_{am}) I a = ( f a 1 , … , f am ) be a finite local presentation of I I I . Since each f a i f_{ai} f ai belongs to Ann F \sqrt{\operatorname{Ann}\mathcal{F}} Ann F , there exists in a neighborhood of each point p ∈ U a p\in U_a p ∈ U a an integer n a i ≥ 1 n_{ai}\geq1 n ai ≥ 1 such that f a i n a i F = 0 f_{ai}^{n_{ai}}\mathcal{F}=0 f ai n ai F = 0 . Set N a = ∑ i n a i N_a=\sum_i n_{ai} N a = ∑ i n ai , so that on the chart U a U_a U a , I N a F = 0 I^{N_a}\mathcal{F}=0 I N a F = 0 . Since ( L X ) c l (\mathcal{L}X)^{\mathrm{cl}} ( L X ) cl is quasi-compact, we may pick a maximum such integer N = max N a N=\max N_a N = max N a for a finite open cover of X X X , so that globally I N F = 0 I^N\mathcal{F}=0 I N F = 0 .
Thus, F \mathcal{F} F possesses a finite filtration F ⊃ I F ⊃ ⋯ ⊃ I N F \mathcal{F}\supset I\mathcal{F}\supset\cdots\supset I^N\mathcal{F} F ⊃ I F ⊃ ⋯ ⊃ I N F , each of whose associated graded pieces are annihilated by I I I . Thus, each associated graded piece I i F / I i + 1 F I^i\mathcal{F}/I^{i+1}\mathcal{F} I i F / I i + 1 F belongs to the image of i ∗ : Coh ( ( L Z ) c l ) → Coh ( ( L X ) c l ) i_*:\operatorname{Coh}((\mathcal{L}Z)^{\mathrm{cl}})\to\operatorname{Coh}((\mathcal{L}X)^{\mathrm{cl}}) i ∗ : Coh (( L Z ) cl ) → Coh (( L X ) cl ) . Since each term of the filtration on F \mathcal{F} F fits into an exact sequence,
0 ⟶ I i + 1 F ⟶ I i F ⟶ I i F / I i + 1 F ⟶ 0 , 0\longrightarrow I^{i+1}\mathcal{F}\longrightarrow I^i\mathcal{F}\longrightarrow I^i\mathcal{F}/I^{i+1}\mathcal{F}\longrightarrow0, 0 ⟶ I i + 1 F ⟶ I i F ⟶ I i F / I i + 1 F ⟶ 0 ,
we see by induction that F \mathcal{F} F itself belongs to the thick subcategory of Coh ( ( L X ) c l ) \operatorname{Coh}((\mathcal{L}X)^{\mathrm{cl}}) Coh (( L X ) cl ) generated by the image of i ∗ i_* i ∗ .
Now, I I I is an S 1 S^1 S 1 -equivariant subsheaf of O ( L X ) c l \mathcal{O}_{(\mathcal{L}X)^{\mathrm{cl}}} O ( L X ) cl . Thus, the argument above can be performed S 1 S^1 S 1 -equivariantly to show that every F ‾ ∈ ( Coh ( L Z ) c l ( ( L X ) c l ) ♡ ) S 1 \overline{\mathcal{F}}\in(\operatorname{Coh}_{(\mathcal{L}Z)^{\mathrm{cl}}}((\mathcal{L}X)^{\mathrm{cl}})^\heartsuit)^{S^1} F ∈ ( Coh ( L Z ) cl (( L X ) cl ) ♡ ) S 1 is in the thick subcategory generated by the image of ( L i ∗ ) S 1 : Coh ( ( L Z ) c l ) S 1 → Coh ( ( L X ) c l ) S 1 (\mathcal{L}i_*)^{S^1}:\operatorname{Coh}((\mathcal{L}Z)^{\mathrm{cl}})^{S^1}\to\operatorname{Coh}((\mathcal{L}X)^{\mathrm{cl}})^{S^1} ( L i ∗ ) S 1 : Coh (( L Z ) cl ) S 1 → Coh (( L X ) cl ) S 1 . This in turn finishes the argument that Coh L Z ( L X ) S 1 \operatorname{Coh}_{\mathcal{L}Z}(\mathcal{L}X)^{S^1} Coh L Z ( L X ) S 1 is thickly generated by the image of Coh ( L Z ) S 1 \operatorname{Coh}(\mathcal{L}Z)^{S^1} Coh ( L Z ) S 1 under ( L i ∗ ) S 1 (\mathcal{L}i_*)^{S^1} ( L i ∗ ) S 1 .
To conclude the proof, we apply Lemma 6.23 to base change along the map k [ u ] → k [ u ± 1 ] k[u]\to k[u^{\pm1}] k [ u ] → k [ u ± 1 ] and the functor ( L i ∗ ) S 1 : Coh ( ( L Z ) ) S 1 → Coh L Z ( ( L X ) ) S 1 (\mathcal{L}i_*)^{S^1}:\operatorname{Coh}((\mathcal{L}Z))^{S^1}\to\operatorname{Coh}_{\mathcal{L}Z}((\mathcal{L}X))^{S^1} ( L i ∗ ) S 1 : Coh (( L Z ) ) S 1 → Coh L Z (( L X ) ) S 1 .
Lemma 6.23. Let R R R be an E ∞ E_\infty E ∞ -algebra, and let R → R ′ R\to R' R → R ′ be an R R R -algebra. Suppose that F : C → D F:\mathcal{C}\to\mathcal{D} F : C → D is an R R R -linear functor such that D \mathcal{D} D is thickly generated by the image of F F F . Then D ⊗ R R ′ \mathcal{D}\otimes_R R' D ⊗ R R ′ is thickly generated by the image of F ⊗ R R ′ F\otimes_R R' F ⊗ R R ′ .
Proof. Indeed, since − ⊗ R R ′ -\otimes_R R' − ⊗ R R ′ commutes with colimits, we see that the thick subcategory generated by the image of F ⊗ R R ′ F\otimes_R R' F ⊗ R R ′ is precisely D ⊗ R R ′ \mathcal{D}\otimes_R R' D ⊗ R R ′ , since D \mathcal{D} D is the thick subcategory generated by the image of F F F .
Lemma 6.24. Let X X X be a coherent derived algebraic 1 1 1 -stack, and let Z ↪ X Z\hookrightarrow X Z ↪ X be a closed derived substack. Then, pushforward along the natural map j : ι ( X c l ) 45 → X j:\iota(X^{\mathrm{cl}})^{45}\to X j : ι ( X cl ) 45 → X induces an equivalence
Coh Z c l ( X c l ) ♡ → ∼ ( Coh Z ( X ) ) ♡ . (43) \operatorname{Coh}_{Z^{\mathrm{cl}}}(X^{\mathrm{cl}})^\heartsuit\xrightarrow{\sim} \left(\operatorname{Coh}_Z(X)\right)^\heartsuit. \tag*{(43)} Coh Z cl ( X cl ) ♡ ∼ ( Coh Z ( X ) ) ♡ . (43) Moreover, if X X X is equipped with an S 1 S^1 S 1 -action such that Z Z Z is an S 1 S^1 S 1 -invariant substack, then X c l X^{\mathrm{cl}} X cl possesses a natural induced S 1 S^1 S 1 -action from that on X X X ; the canonical t t t -structures on Coh ( X ) \operatorname{Coh}(X) Coh ( X ) and Coh ( X c l ) \operatorname{Coh}(X^{\mathrm{cl}}) Coh ( X cl ) are compatible with the induced S 1 S^1 S 1 -actions on Coh ( X ) \operatorname{Coh}(X) Coh ( X ) and Coh ( X c l ) \operatorname{Coh}(X^{\mathrm{cl}}) Coh ( X cl ) ; and the equivalence (43) is canonically S 1 S^1 S 1 -equivariant.
Proof. We first check that pushforward induces an equivalence j ∗ : Coh ( X c l ) ♡ → ∼ Coh ( X ) ♡ j_*:\operatorname{Coh}(X^{\mathrm{cl}})^\heartsuit\xrightarrow{\sim}\operatorname{Coh}(X)^\heartsuit j ∗ : Coh ( X cl ) ♡ ∼ Coh ( X ) ♡ . The proof of [4 Proposition 2.5.9.1] applies to the map j : X c l → X j:X^{\mathrm{cl}}\to X j : X cl → X locally in the smooth topology on the target. Indeed, if U = Spec A → X U=\operatorname{Spec}A\to X U = Spec A → X is a smooth affine chart, then its base change j A : Spec H 0 ( A ) → Spec A j_A:\operatorname{Spec}H^0(A)\to\operatorname{Spec}A j A : Spec H 0 ( A ) → Spec A has 1 1 1 -connective fiber. This implies that j A ∗ : QCoh ( Spec A ) ♡ → QCoh ( Spec H 0 ( A ) ) ♡ j_{A*}:\operatorname{QCoh}(\operatorname{Spec}A)^\heartsuit\to\operatorname{QCoh}(\operatorname{Spec}H^0(A))^\heartsuit j A ∗ : QCoh ( Spec A ) ♡ → QCoh ( Spec H 0 ( A ) ) ♡ is an equivalence. Moreover, it is apparent from the proof of [4 Proposition 2.5.9.1] that j A ∗ j_{A*} j A ∗ preserves the subcategory of finitely generated objects, so the equivalence
restricts to coherent objects. Since being coherent and being in the heart are local properties in the smooth topology, it follows that j ∗ j_* j ∗ is an equivalence, as desired.46
We now check that we obtain the induced equivalence (43) . We claim to have a Cartesian diagram
U c l → i U c l X c l j U ↓ ↓ j U → i U X (44) \begin{CD}
U^{\mathrm{cl}} @>{i_{U^{\mathrm{cl}}}}>> X^{\mathrm{cl}} \\
@V{j_U}VV @VV{j}V \\
U @>{i_U}>> X \tag*{(44)}
\end{CD} U cl j U ↓ ⏐ U i U cl i U X cl ↓ ⏐ j X (44)
Indeed, there is a natural map U c l → U × X X c l U^{\mathrm{cl}} \to U \times_X X^{\mathrm{cl}} U cl → U × X X cl . To see that it is an equivalence, it is sufficient to check the following. Let x : S → X x:S \to X x : S → X be an S S S -point of X X X . Then S × X Z = ∅ S \times_X Z = \varnothing S × X Z = ∅ if and only if S × X Z c l = ∅ S \times_X Z^{\mathrm{cl}} = \varnothing S × X Z cl = ∅ . One direction is clear, to see the other, it suffices to observe that a stack Y Y Y is empty if and only if its classical truncation is. Indeed, any T T T -point T → Y T \to Y T → Y restricts to a point T c l → Y c l T^{\mathrm{cl}} \to Y^{\mathrm{cl}} T cl → Y cl , and T c l T^{\mathrm{cl}} T cl is nonempty whenever T T T is nonempty. Thus,
U c l → ∼ U × X X c l U^{\mathrm{cl}} \xrightarrow{\sim} U \times_X X^{\mathrm{cl}} U cl ∼ U × X X cl
is an equivalence as desired. We can then use base change for the diagram (44) to see that if F ∈ Coh Z c l ( X c l ) ♡ \mathcal{F} \in\operatorname{Coh}_{Z^{\mathrm{cl}}}(X^{\mathrm{cl}})^\heartsuit F ∈ Coh Z cl ( X cl ) ♡ , then i U ! j ∗ F ≅ 0 i_U^!j_*\mathcal{F} \cong0 i U ! j ∗ F ≅ 0 . Thus, we see that j ∗ j_* j ∗ defines a functor
Coh Z c l ( X c l ) ♡ ⟶ Coh Z ( X ) ♡ . \operatorname{Coh}_{Z^{\mathrm{cl}}}(X^{\mathrm{cl}})^\heartsuit\longrightarrow\operatorname{Coh}_Z(X)^\heartsuit. Coh Z cl ( X cl ) ♡ ⟶ Coh Z ( X ) ♡ .
Clearly, the functor is fully faithful, because in the commutative diagram
\begin{CD}
\operatorname{Coh}_{Z^{\mathrm{cl}}}(X^{\mathrm{cl}})^\heartsuit@>>> \operatorname{Coh}_Z(X)^\heartsuit\\
@VVV @VVV \\
\operatorname{Coh}(X^{\mathrm{cl}})^\heartsuit@>{\simeq}>> \operatorname{Coh}(X)^\heartsuit
\end{CD}
the vertical arrows are fully faithful inclusions and the bottom horizontal arrow is an equivalence. To see it is essentially surjective, suppose F ∈ Coh Z ( X ) ♡ \mathcal{F} \in\operatorname{Coh}_Z(X)^\heartsuit F ∈ Coh Z ( X ) ♡ . Under the inclusion Coh Z ( X ) ♡ ↪ Coh ( X ) ♡ \operatorname{Coh}_Z(X)^\heartsuit\hookrightarrow\operatorname{Coh}(X)^\heartsuit Coh Z ( X ) ♡ ↪ Coh ( X ) ♡ and the equivalence
Coh ( X c l ) ♡ → ≃ Coh ( X ) ♡ , \operatorname{Coh}(X^{\mathrm{cl}})^\heartsuit\xrightarrow{\simeq} \operatorname{Coh}(X)^\heartsuit, Coh ( X cl ) ♡ ≃ Coh ( X ) ♡ ,
there is an F 0 ∈ Coh ( X c l ) ♡ \mathcal{F}_0 \in\operatorname{Coh}(X^{\mathrm{cl}})^\heartsuit F 0 ∈ Coh ( X cl ) ♡ such that j ∗ F 0 = F j_*\mathcal{F}_0 = \mathcal{F} j ∗ F 0 = F . It is sufficient to show that j U ! j ∗ F 0 ≅ 0 j_U^!j_*\mathcal{F}_0 \cong0 j U ! j ∗ F 0 ≅ 0 , which follows from base change along (44) .
If X X X is equipped with an S 1 S^1 S 1 -action such that Z Z Z is an S 1 S^1 S 1 -invariant substack, then there is an induced S 1 S^1 S 1 -action on X c l X^{\mathrm{cl}} X cl by pulling back the action map S 1 × X → X S^1 \times X \to X S 1 × X → X to X c l X^{\mathrm{cl}} X cl . Clearly, the map X c l → X X^{\mathrm{cl}} \to X X cl → X is S 1 S^1 S 1 -equivariant. The truncation functors on QCoh ( X ) \operatorname{QCoh}(X) QCoh ( X ) and QCoh ( X c l ) \operatorname{QCoh}(X^{\mathrm{cl}}) QCoh ( X cl ) are S 1 S^1 S 1 -equivariant by Lemma 3.14. Thus, the S 1 S^1 S 1 -equivariant functor j ∗ j_* j ∗ preserves the hearts as desired. □ \square □
Lemma 6.27. Suppose that C \mathcal{C} C and D \mathcal{D} D are stable ∞ \infty ∞ -categories such that D \mathcal{D} D is equipped with a bounded t t t -structure. Then D \mathcal{D} D is thickly generated by the image of a functor F : C → D F:\mathcal{C} \to\mathcal{D} F : C → D if and only if its heart D ♡ \mathcal{D}^\heartsuit D ♡ is.
Proof. One direction is clear, so suppose D ♡ \mathcal{D}^\heartsuit D ♡ is generated by im ( F ) \operatorname{im}(F) im ( F ) . Denote by D ′ \mathcal{D}' D ′ the category thickly generated by im ( F ) \operatorname{im}(F) im ( F ) . Let d ∈ D d \in\mathcal{D} d ∈ D . Since the t t t -structure on D \mathcal{D} D is bounded, there are integers ℓ ≤ k \ell\leq k ℓ ≤ k such that d ∈ D ≥ ℓ ∩ D ≤ k d \in\mathcal{D}^{\geq\ell} \cap\mathcal{D}^{\leq k} d ∈ D ≥ ℓ ∩ D ≤ k . We proceed by induction on n = k − ℓ n = k-\ell n = k − ℓ . If n = 0 n=0 n = 0 , then d ≃ H ℓ ( d ) [ − ℓ ] d \simeq H^\ell(d)[-\ell] d ≃ H ℓ ( d ) [ − ℓ ] . Since H ℓ ( d ) ∈ D ♡ ⊂ D ′ H^\ell(d) \in\mathcal{D}^\heartsuit\subset\mathcal{D}' H ℓ ( d ) ∈ D ♡ ⊂ D ′ and D ′ \mathcal{D}' D ′ is closed under shifts, we see that d ∈ D ′ d \in\mathcal{D}' d ∈ D ′ . For n > 0 n>0 n > 0 , use the triangle
τ ≤ k − 1 d ⟶ d ⟶ H k ( d ) [ − k ] . \tau^{\leq k-1}d \longrightarrow d \longrightarrow H^k(d)[-k]. τ ≤ k − 1 d ⟶ d ⟶ H k ( d ) [ − k ] .
The first term lies in D ′ \mathcal{D}' D ′ by induction. The last term also lies in D ′ \mathcal{D}' D ′ because H k ( d ) ∈ D ♡ H^k(d) \in\mathcal{D}^{\heartsuit} H k ( d ) ∈ D ♡ . Since D ′ \mathcal{D}' D ′ is closed under extensions, we conclude. □
Lemma 6.28. Suppose that X X X is a Noetherian derived algebraic 1 1 1 -stack. Then ( L X ) c l (\mathcal{L}X)^{\mathrm{cl}} ( L X ) cl is a Noetherian classical stack. In particular, ( L X ) c l (\mathcal{L}X)^{\mathrm{cl}} ( L X ) cl is quasi-compact.
Proof. Denote X c l = : X 0 X^{\mathrm{cl}} =: X_0 X cl =: X 0 . Classical truncation preserves limits, as it is a right adjoint. Hence,
( L X ) c l ≃ X 0 × X 0 × X 0 X 0 , (\mathcal{L}X)^{\mathrm{cl}} \simeq X_0 \mathop{\times}_{X_0 \times X_0} X_0, ( L X ) cl ≃ X 0 × X 0 × X 0 X 0 ,
and the canonical map
p : ( L X ) c l ⟶ X 0 p : (\mathcal{L}X)^{\mathrm{cl}} \longrightarrow X_0 p : ( L X ) cl ⟶ X 0
is a base change of the diagonal Δ 0 \Delta_0 Δ 0 of X 0 X_0 X 0 . Now, X 0 X_0 X 0 is a classical algebraic stack; hence, its diagonal is representable by algebraic spaces and is locally of finite type. Moreover, X 0 X_0 X 0 is Noetherian; hence, its diagonal is qcqs, since X 0 X_0 X 0 is in particular quasi-separated. This immediately gives that ( L X ) c l (\mathcal{L}X)^{\mathrm{cl}} ( L X ) cl is locally Noetherian. It is also qcqs, since this holds for p p p and X 0 X_0 X 0 . Thus, it is Noetherian as desired. □
The linearly scalloped case
Nisnevich descent
In this section, we prove that potent D \mathcal{D} D -modules on linearly scalloped stacks satisfy Kashiwara’s Lemma. Our proof relies heavily on the fact that D p o t ( − ) \mathcal{D}^{\mathrm{pot}}(-) D pot ( − ) satisfies Nisnevich descent, which we prove below after recalling the following technical lemma. Let Rep / X \operatorname{Rep}/X Rep / X denote the category of derived stacks Y Y Y with affine diagonal over X X X for which the structure map Y → X Y \to X Y → X is representable by algebraic spaces and of finite presentation.
Remark 7.1. The category Rep / X \operatorname{Rep}/X Rep / X we define here differs slightly from the one in [KR24], where they do not require objects to have affine diagonal.
Proposition 7.2. Let X X X be an algebraic stack almost of finite type and with affine diagonal47 Then the presheaf D p o t : Rep / X → DGCat k [ u ± 1 ] \mathcal{D}^{\mathrm{pot}} : \operatorname{Rep}/X \to\operatorname{DGCat}_{k[u^{\pm1}]} D pot : Rep / X → DGCat k [ u ± 1 ] satisfies descent with respect to the Nisnevich topology on Rep / X \operatorname{Rep}/X Rep / X .
Proof. It suffices to show that D p o t ( − ) \mathcal{D}^{\mathrm{pot}}(-) D pot ( − ) is Nisnevich excisive, i.e. takes Nisnevich squares48 to pullback diagrams, by [Kha19, Theorem 2.2.7].
Now, given a Nisnevich square
W → j ′ V p ′ ↓ ↓ p U → j S \begin{CD}
W @>{j'}>> V \\
@V{p'}VV @VV{p}V \\
U @>{j}>> S
\end{CD} W p ′ ↓ ⏐ U j ′ j V ↓ ⏐ p S
where we recall that j j j is an immersion and p p p is an étale map that restricts to an isomorphism along the complement to j j j . From this, we obtain another Nisnevich square
L W → L j ′ L V L p ′ ↓ ↓ L p L U → L j L S (45) \begin{CD}
\mathcal{L}W @>{\mathcal{L}j'}>> \mathcal{L}V \\
@V{\mathcal{L}p'}VV @VV{\mathcal{L}p}V \\
\mathcal{L}U @>{\mathcal{L}j}>> \mathcal{L}S \tag*{(45)}
\end{CD} L W L p ′ ↓ ⏐ L U L j ′ L j L V ↓ ⏐ L p L S (45)
since L \mathcal{L} L is a Nisnevich cosheaf by [BK]. We note that each morphism in (45) is canonically S 1 S^1 S 1 -equivariant. Since IndCoh satisfies fppf, hence Nisnevich, descent ([Gai13, Corollary 10.4.5]), we obtain an S 1 S^1 S 1 -equivariant equivalence,
IndCoh ( L S ) → ≃ IndCoh ( L V ) × IndCoh ( L W ) IndCoh ( L U ) . \operatorname{IndCoh}(\mathcal{L}S) \xrightarrow{\simeq} \operatorname{IndCoh}(\mathcal{L}V) \mathop{\times}_{\operatorname{IndCoh}(\mathcal{L}W)} \operatorname{IndCoh}(\mathcal{L}U). IndCoh ( L S ) ≃ IndCoh ( L V ) × IndCoh ( L W ) IndCoh ( L U ) .
We conclude by checking that the set-up above satisfies the conditions of Lemma 7.6 . By [CW23, Proposition 5.30], the t t t -structure on IndCoh ( S ) \operatorname{IndCoh}(\mathcal{S}) IndCoh ( S ) for any coherent ind-geometric stack S \mathcal{S} S is regular, meaning in particular that IndCoh ( S ) \operatorname{IndCoh}(\mathcal{S}) IndCoh ( S ) is compactly generated. We claim that the functors ( L j ′ ) ! : IndCoh ( V ) → IndCoh ( W ) (\mathcal{L}j')^!:\operatorname{IndCoh}(\mathcal{V})\to\operatorname{IndCoh}(\mathcal{W}) ( L j ′ ) ! : IndCoh ( V ) → IndCoh ( W ) and ( L p ′ ) ! : IndCoh ( U ) → IndCoh ( W ) (\mathcal{L}p')^!:\operatorname{IndCoh}(\mathcal{U})\to\operatorname{IndCoh}(\mathcal{W}) ( L p ′ ) ! : IndCoh ( U ) → IndCoh ( W ) are each t t t -exact and preserve compact objects. The morphism j ′ j' j ′ is an open immersion as the base change of an open immersion, so L j ′ \mathcal{L}j' L j ′ is an open immersion by Lemma 3.51. In particular, L j ′ \mathcal{L}j' L j ′ is smooth. On the other hand L p \mathcal{L}p L p is étale by virtue of (45) being a Nisnevich square, so L p ′ \mathcal{L}p' L p ′ is étale as the base change of an étale morphism. In particular, L p ′ \mathcal{L}p' L p ′ is smooth. Thus, by Corollary 3.42, ( L p ′ ) ! (\mathcal{L}p')^! ( L p ′ ) ! and ( L j ′ ) ! (\mathcal{L}j')^! ( L j ′ ) ! each preserve compact objects, and they are t t t -exact by Lemma 3.39. Finally, ( L j ′ ) ! (\mathcal{L}j')^! ( L j ′ ) ! has a fully faithful t t t -bounded right adjoint by Lemma 3.43 and Lemma 3.44 .
Lemma 7.4. Let A → F C ← G B \mathcal{A} \xrightarrow{F} \mathcal{C} \xleftarrow{G} \mathcal{B} A F C G B be a span in R e g t \mathbf{Reg}_t Reg t , such that G : B → C G:\mathcal{B}\to\mathcal{C} G : B → C has a fully faithful right adjoint in Pr L \operatorname{Pr}^{\mathrm{L}} Pr L , and such that F F F and G G G are t t t -exact and preserve compact objects. Assume that A × C B \mathcal{A}\times_{\mathcal{C}}\mathcal{B} A × C B is compactly generated. Then the lax product A × ⃗ C B \mathcal{A}\mathbin{\vec{\times}}_{\mathcal{C}}\mathcal{B} A × C B 49 and the product A × C B \mathcal{A}\times_{\mathcal{C}}\mathcal{B} A × C B have canonical regular t t t -structures.
Proof. We first prove the claim that the canonical t t t -structure on the lax product A × ⃗ C B \mathcal{A}\mathbin{\vec{\times}}_{\mathcal{C}}\mathcal{B} A × C B is coherent. We begin by considering the pointwise t t t -structure on Coh ( C ) [ 1 ] \operatorname{Coh}(\mathcal{C})^{[1]} Coh ( C ) [ 1 ] , where [ 1 ] [1] [ 1 ] is the standard 1 1 1 -simplex. By [HP23, Lemma 2.1.9(iv)], the products Coh ( A ) × Coh ( B ) \operatorname{Coh}(\mathcal{A})\times\operatorname{Coh}(\mathcal{B}) Coh ( A ) × Coh ( B ) and Coh ( C ) × Coh ( C ) \operatorname{Coh}(\mathcal{C})\times\operatorname{Coh}(\mathcal{C}) Coh ( C ) × Coh ( C ) have canonical t t t -structures such that the projection morphisms are t t t -exact. By the same, the lax product Coh ( A ) × ⃗ Coh ( C ) Coh ( B ) : = ( Coh ( A ) × Coh ( B ) ) × Coh ( C ) × Coh ( C ) Coh ( A ) × Coh ( B ) \operatorname{Coh}(\mathcal{A})\mathbin{\vec{\times}}_{\operatorname{Coh}(\mathcal{C})}\operatorname{Coh}(\mathcal{B}):=(\operatorname{Coh}(\mathcal{A})\times\operatorname{Coh}(\mathcal{B}))\times_{\operatorname{Coh}(\mathcal{C})\times\operatorname{Coh}(\mathcal{C})}\operatorname{Coh}(\mathcal{A})\times\operatorname{Coh}(\mathcal{B}) Coh ( A ) × Coh ( C ) Coh ( B ) := ( Coh ( A ) × Coh ( B )) × Coh ( C ) × Coh ( C ) Coh ( A ) × Coh ( B ) possesses a unique t t t -structure such that the projection functors to Coh ( C ) [ 1 ] \operatorname{Coh}(\mathcal{C})^{[1]} Coh ( C ) [ 1 ] , Coh ( A ) × Coh ( B ) \operatorname{Coh}(\mathcal{A})\times\operatorname{Coh}(\mathcal{B}) Coh ( A ) × Coh ( B ) , and Coh ( C ) × Coh ( C ) \operatorname{Coh}(\mathcal{C})\times\operatorname{Coh}(\mathcal{C}) Coh ( C ) × Coh ( C ) . By [AGH19, Proposition 2.13], this t t t -structure induces one on the ind-completion Ind ( Coh ( A ) × ⃗ Coh ( C ) Coh ( B ) ) \operatorname{Ind}(\operatorname{Coh}(\mathcal{A})\mathbin{\vec{\times}}_{\operatorname{Coh}(\mathcal{C})}\operatorname{Coh}(\mathcal{B})) Ind ( Coh ( A ) × Coh ( C ) Coh ( B )) , which is compatible with filtered colimits, and right complete if the t t t -structure on Coh ( A ) × ⃗ Coh ( C ) Coh ( B ) \operatorname{Coh}(\mathcal{A})\mathbin{\vec{\times}}_{\operatorname{Coh}(\mathcal{C})}\operatorname{Coh}(\mathcal{B}) Coh ( A ) × Coh ( C ) Coh ( B ) is bounded below.
Since A \mathcal{A} A , B \mathcal{B} B , and C \mathcal{C} C belong to C o h t \mathbf{Coh}_t Coh t , their coherent subcategories Coh ( A ) \operatorname{Coh}(\mathcal{A}) Coh ( A ) , Coh ( B ) \operatorname{Coh}(\mathcal{B}) Coh ( B ) , and Coh ( C ) \operatorname{Coh}(\mathcal{C}) Coh ( C ) have bounded t t t -structures. Hence, the categories Coh ( C ) [ 1 ] \operatorname{Coh}(\mathcal{C})^{[1]} Coh ( C ) [ 1 ] , Coh ( A ) × Coh ( B ) \operatorname{Coh}(\mathcal{A})\times\operatorname{Coh}(\mathcal{B}) Coh ( A ) × Coh ( B ) , and Coh ( C ) × Coh ( C ) \operatorname{Coh}(\mathcal{C})\times\operatorname{Coh}(\mathcal{C}) Coh ( C ) × Coh ( C ) inherit bounded t t t -structures. From this, it is clear that the t t t -structure on the fiber product Coh ( A ) × ⃗ Coh ( C ) Coh ( B ) \operatorname{Coh}(\mathcal{A})\mathbin{\vec{\times}}_{\operatorname{Coh}(\mathcal{C})}\operatorname{Coh}(\mathcal{B}) Coh ( A ) × Coh ( C ) Coh ( B ) is also bounded—in particular, it is bounded below.
In order to show that the t t t -structure on Ind ( Coh ( A ) × ⃗ Coh ( C ) Coh ( B ) ) \operatorname{Ind}(\operatorname{Coh}(\mathcal{A})\mathbin{\vec{\times}}_{\operatorname{Coh}(\mathcal{C})}\operatorname{Coh}(\mathcal{B})) Ind ( Coh ( A ) × Coh ( C ) Coh ( B )) is coherent, it therefore remains to show that its heart is a locally coherent abelian category, meaning that it is compactly generated (as an abelian category) and that the subcategory of compact objects in the heart is abelian. From the construction of the t t t -structure, we have
Ind ( Coh ( A ) × ⃗ Coh ( C ) Coh ( B ) ) ♡ ≃ Ind 1 ( ( Coh ( A ) × ⃗ Coh ( C ) Coh ( B ) ) ♡ ) (46) \operatorname{Ind}(\operatorname{Coh}(\mathcal{A})\mathbin{\vec{\times}}_{\operatorname{Coh}(\mathcal{C})}\operatorname{Coh}(\mathcal{B}))^\heartsuit\simeq\operatorname{Ind}_1\left((\operatorname{Coh}(\mathcal{A})\mathbin{\vec{\times}}_{\operatorname{Coh}(\mathcal{C})}\operatorname{Coh}(\mathcal{B}))^\heartsuit\right) \tag*{(46)} Ind ( Coh ( A ) × Coh ( C ) Coh ( B ) ) ♡ ≃ Ind 1 ( ( Coh ( A ) × Coh ( C ) Coh ( B ) ) ♡ ) (46)
where Ind 1 \operatorname{Ind}_1 Ind 1 denotes the 1 1 1 -categorical ind-construction. Because Coh ( A ) × ⃗ Coh ( C ) Coh ( B ) \operatorname{Coh}(\mathcal{A})\mathbin{\vec{\times}}_{\operatorname{Coh}(\mathcal{C})}\operatorname{Coh}(\mathcal{B}) Coh ( A ) × Coh ( C ) Coh ( B ) is a limit in Cat ∞ s m \operatorname{Cat}^{\mathrm{sm}}_\infty Cat ∞ sm , it is an ω \omega ω -small stable category. In particular, any of its subcategories is also essentially small. Thus, ( Coh ( A ) × ⃗ Coh ( C ) Coh ( B ) ) ♡ (\operatorname{Coh}(\mathcal{A})\mathbin{\vec{\times}}_{\operatorname{Coh}(\mathcal{C})}\operatorname{Coh}(\mathcal{B}))^\heartsuit ( Coh ( A ) × Coh ( C ) Coh ( B ) ) ♡ is a skeletally small abelian category. Thus, by [1 Theorem 4.1] , its Ind 1 \operatorname{Ind}_1 Ind 1 -category is a Grothendieck locally coherent abelian category whose subcategory of compact objects is precisely ( Coh ( A ) × ⃗ Coh ( C ) Coh ( B ) ) ♡ (\operatorname{Coh}(\mathcal{A})\mathbin{\vec{\times}}_{\operatorname{Coh}(\mathcal{C})}\operatorname{Coh}(\mathcal{B}))^\heartsuit ( Coh ( A ) × Coh ( C ) Coh ( B ) ) ♡ , which shows that the t t t -structure on Ind ( Coh ( A ) × ⃗ Coh ( C ) Coh ( B ) ) \operatorname{Ind}(\operatorname{Coh}(\mathcal{A})\mathbin{\vec{\times}}_{\operatorname{Coh}(\mathcal{C})}\operatorname{Coh}(\mathcal{B})) Ind ( Coh ( A ) × Coh ( C ) Coh ( B )) is coherent. To conclude that the t t t -structure on A × ⃗ C B \mathcal{A}\mathbin{\vec{\times}}_{\mathcal{C}}\mathcal{B} A × C B is coherent, we note that
Ind ( Coh ( A ) × ⃗ Coh ( C ) Coh ( B ) ) ≃ Ind ( Coh ( A ) ) × ⃗ Ind ( Coh ( C ) ) Ind ( Coh ( B ) ) ≃ A × ⃗ C B , \begin{aligned}
\operatorname{Ind}(\operatorname{Coh}(\mathcal{A})\mathbin{\vec{\times}}_{\operatorname{Coh}(\mathcal{C})}\operatorname{Coh}(\mathcal{B})) \simeq\operatorname{Ind}(\operatorname{Coh}(\mathcal{A}))\mathbin{\vec{\times}}_{\operatorname{Ind}(\operatorname{Coh}(\mathcal{C}))}\operatorname{Ind}(\operatorname{Coh}(\mathcal{B})) \\
&\simeq\mathcal{A}\mathbin{\vec{\times}}_{\mathcal{C}}\mathcal{B},
\end{aligned} Ind ( Coh ( A ) × Coh ( C ) Coh ( B )) ≃ Ind ( Coh ( A )) × Ind ( Coh ( C )) Ind ( Coh ( B )) ≃ A × C B ,
by [5 Proposition 13] . The exact same argument as above shows that the canonical t t t -structure on Ind ( Coh ( A ) × Coh ( C ) Coh ( B ) ) \operatorname{Ind}(\operatorname{Coh}(\mathcal{A})\times_{\operatorname{Coh}(\mathcal{C})}\operatorname{Coh}(\mathcal{B})) Ind ( Coh ( A ) × Coh ( C ) Coh ( B )) is coherent. Since the analogue of [5 Proposition 13] for products is not true in general, it remains to show in our case that Ind ( Coh ( A ) × Coh ( C ) Coh ( B ) ) ≃ A × C B \operatorname{Ind}(\operatorname{Coh}(\mathcal{A}) \times_{\operatorname{Coh}(\mathcal{C})} \operatorname{Coh}(\mathcal{B})) \simeq\mathcal{A} \times_{\mathcal{C}} \mathcal{B} Ind ( Coh ( A ) × Coh ( C ) Coh ( B )) ≃ A × C B in order to prove that the t t t -structure on the latter is coherent.
To begin, we note that, because G G G has a fully faithful adjoint, there exists an exact sequence \mathcal{A} \times_{\mathcal{C}} \mathcal{B} \xrightarrow{\iota} \mathcal{A} \mathbin{\times\limits^{\rightarrow}_{\mathcal{C}}} \mathcal{B} \xrightarrow{e} \mathcal{C} by [5 Corollary 12] . Since F F F and G G G preserve compact objects by assumption, the latter functor restricts to a functor (\mathcal{A} \mathbin{\times\limits^{\rightarrow}_{\mathcal{C}}} \mathcal{B})^\omega\to\mathcal{C}^\omega . At the same time, by [5 Proposition 13] , there is an equivalence (\mathcal{A} \mathbin{\times\limits^{\rightarrow}_{\mathcal{C}}} \mathcal{B})^\omega\simeq\mathcal{A}^\omega\mathbin{\times\limits^{\rightarrow}_{\mathcal{C}^\omega} \mathcal{B}^\omega} . Since the t t t -structures on A \mathcal{A} A , B \mathcal{B} B , and C \mathcal{C} C are regular, their subcategories of compact objects coincide with their subcategories of coherent objects, so we see that the functor e^\omega: \operatorname{Coh}(\mathcal{A}) \mathbin{\times\limits^{\rightarrow}_{\operatorname{Coh}(\mathcal{C})}} \operatorname{Coh}(\mathcal{B}) \to\operatorname{Coh}(\mathcal{C}) is the restriction of the functor e : \mathcal{A} \mathbin{\times\limits^{\rightarrow}_{\mathcal{C}}} \mathcal{B} \to\mathcal{C} to compact objects. Now, observe that the kernel of e ω e^\omega e ω is precisely the ordinary product, Coh ( A ) × Coh ( C ) Coh ( B ) \operatorname{Coh}(\mathcal{A}) \times_{\operatorname{Coh}(\mathcal{C})} \operatorname{Coh}(\mathcal{B}) Coh ( A ) × Coh ( C ) Coh ( B ) . We now claim that ker ( e ) ω ≃ ker ( e ω ) \ker(e)^\omega\simeq\ker(e^\omega) ker ( e ) ω ≃ ker ( e ω ) .
Indeed, since e e e is a localization, it has a right adjoint e R e^R e R . Since e e e preserves compact objects (because each of F F F and G G G does), e R e^R e R preserves filtered colimits. Since \iota: \mathcal{A} \times_{\mathcal{C}} \mathcal{B} \to\mathcal{A} \mathbin{\times\limits^{\rightarrow}_{\mathcal{C}}} \mathcal{B} is the kernel of the map e e e , its right adjoint ι R \iota^R ι R given object-wise by ι R ( X ) ≃ ker ( X → e R e ( X ) ) \iota^R(X) \simeq\ker(X \to e^R e(X)) ι R ( X ) ≃ ker ( X → e R e ( X )) . Since e R e^R e R preserves filtered colimits, and e e e preserves colimits as a left-adjoint functor, we find that, for X = colim n X n X = \operatorname{colim}_n X_n X = colim n X n ,
ι R ( X ) = ι R ( colim n X n ) ≃ ker ( colim n X n → e R e ( colim n X n ) ) ≃ ker ( colim n ( X n → e R e ( X n ) ) ) ≃ colim n ker ( X n → e R e ( X n ) ) , \begin{aligned}
\iota^R(X) &= \iota^R\left(\operatorname{colim}_n X_n\right) \simeq\ker\left(\operatorname{colim}_n X_n \to e^R e\left(\operatorname{colim}_n X_n\right)\right) \\
&\simeq\ker\left(\operatorname{colim}_n\left(X_n \to e^R e(X_n)\right)\right) \\
&\simeq\operatorname{colim}_n \ker\left(X_n \to e^R e(X_n)\right),
\end{aligned} ι R ( X ) = ι R ( colim n X n ) ≃ ker ( colim n X n → e R e ( colim n X n ) ) ≃ ker ( colim n ( X n → e R e ( X n ) ) ) ≃ colim n ker ( X n → e R e ( X n ) ) ,
where the last equivalence follows from the fact that in a stable category, kernels are both limits and colimits. Since ι R \iota^R ι R thus preserves filtered colimits, ι \iota ι preserves compact objects, meaning \ker(e)^\omega\simeq\ker(e) \cap(\mathcal{A} \mathbin{\times\limits^{\rightarrow}_{\mathcal{C}}} \mathcal{B})^\omega , while the latter object is precisely \ker(e|_{(\mathcal{A} \mathbin{\times\limits^{\rightarrow}_{\mathcal{C}}} \mathcal{B})^\omega}) = \ker(e^\omega) . Since ker ( e ) ≃ A × C B \ker(e) \simeq\mathcal{A} \times_{\mathcal{C}} \mathcal{B} ker ( e ) ≃ A × C B , is compactly generated by assumption, we finally obtain that ( A × C B ) ω ≃ Coh ( A ) × Coh ( C ) Coh ( B ) (\mathcal{A} \times_{\mathcal{C}} \mathcal{B})^\omega\simeq\operatorname{Coh}(\mathcal{A}) \times_{\operatorname{Coh}(\mathcal{C})} \operatorname{Coh}(\mathcal{B}) ( A × C B ) ω ≃ Coh ( A ) × Coh ( C ) Coh ( B ) , which implies that the equivalence Ind ( Coh ( A ) × Coh ( C ) Coh ( B ) ) ≃ A × C B \operatorname{Ind}(\operatorname{Coh}(\mathcal{A}) \times_{\operatorname{Coh}(\mathcal{C})} \operatorname{Coh}(\mathcal{B})) \simeq\mathcal{A} \times_{\mathcal{C}} \mathcal{B} Ind ( Coh ( A ) × Coh ( C ) Coh ( B )) ≃ A × C B as desired.
By [HA, Lemma 5.3.2.9], ind-completion furnishes an equivalence Ind ( − ) : Cat ∞ i d m → Pr ω L , ω \operatorname{Ind}(-) : \operatorname{Cat}^{\mathrm{idm}}_\infty\to\operatorname{Pr}^{\mathrm{L},\omega}_\omega Ind ( − ) : Cat ∞ idm → Pr ω L , ω , where Cat ∞ s m \operatorname{Cat}^{\mathrm{sm}}_\infty Cat ∞ sm denotes the category of ω \omega ω -small, idempotent-complete ∞ \infty ∞ -categories, and Pr ω L , ω \operatorname{Pr}^{\mathrm{L},\omega}_\omega Pr ω L , ω denotes the category of ω \omega ω -presentable, stable ∞ \infty ∞ -categories. In particular, this implies that the universal functor,
Ind ( Coh ( A ) × Coh ( C ) Coh ( B ) ) ≃ Ind ( Coh ( A ) ) × Ind ( Coh ( C ) ) Ind ( Coh ( B ) ) ≃ A × C B \begin{aligned}
\operatorname{Ind}(\operatorname{Coh}(\mathcal{A}) \times_{\operatorname{Coh}(\mathcal{C})} \operatorname{Coh}(\mathcal{B})) &\simeq\operatorname{Ind}(\operatorname{Coh}(\mathcal{A})) \times_{\operatorname{Ind}(\operatorname{Coh}(\mathcal{C}))} \operatorname{Ind}(\operatorname{Coh}(\mathcal{B})) \\
&\simeq\mathcal{A} \times_{\mathcal{C}} \mathcal{B}
\end{aligned} Ind ( Coh ( A ) × Coh ( C ) Coh ( B )) ≃ Ind ( Coh ( A )) × Ind ( Coh ( C )) Ind ( Coh ( B )) ≃ A × C B
is an equivalence.
In order to show that the t t t -structure on \mathcal{A} \mathbin{\times\limits^{\rightarrow}_{\mathcal{C}}} \mathcal{B} is regular, we need to show that
\operatorname{Coh}(\mathcal{A}) \mathbin{\times\limits^{\rightarrow}_{\operatorname{Coh}(\mathcal{C})}} \operatorname{Coh}(\mathcal{B}) \simeq\operatorname{Coh}\left(\mathcal{A} \mathbin{\times\limits^{\rightarrow}_{\mathcal{C}}} \mathcal{B}\right).
By [BNP17, Lemma 6.2.5], since the t t t -structure on \mathcal{A} \mathbin{\times\limits^{\rightarrow}_{\mathcal{C}}} \mathcal{B} is coherent, an object X \in\mathcal{A} \mathbin{\times\limits^{\rightarrow}_{\mathcal{C}}} \mathcal{B} belongs to \operatorname{Coh}(\mathcal{A} \mathbin{\times\limits^{\rightarrow}_{\mathcal{C}}} \mathcal{B}) if and only if H_nX \in\operatorname{Coh}(\mathcal{A} \mathbin{\times\limits^{\rightarrow}_{\mathcal{C}}} \mathcal{B}) for all n n n , and H n X = 0 H_nX = 0 H n X = 0 for all but finitely many n n n . By the same, we also know that \operatorname{Coh}(\mathcal{A} \mathbin{\times\limits^{\rightarrow}_{\mathcal{C}}} \mathcal{B}) consists precisely of the compact objects in (\mathcal{A} \mathbin{\times\limits^{\rightarrow}_{\mathcal{C}}} \mathcal{B})^\heartsuit . By the characterization (46) above, we have
\left((\mathcal{A} \mathbin{\times\limits^{\rightarrow}_{\mathcal{C}}} \mathcal{B})^\omega\right)^\heartsuit\simeq\left(\operatorname{Coh}(\mathcal{A}) \mathbin{\times\limits^{\rightarrow}_{\operatorname{Coh}(\mathcal{C})}} \operatorname{Coh}(\mathcal{B})\right)^\heartsuit.
That is, X \in\mathcal{A} \mathbin{\times\limits^{\rightarrow}_{\mathcal{C}}} \mathcal{B} belongs to \operatorname{Coh}(\mathcal{A} \mathbin{\times\limits^{\rightarrow}_{\mathcal{C}}} \mathcal{B}) if and only if
H_nX \in\left(\operatorname{Coh}(\mathcal{A}) \mathbin{\times\limits^{\rightarrow}_{\operatorname{Coh}(\mathcal{C})}} \operatorname{Coh}(\mathcal{B})\right)^\heartsuit , for all n n n ,
and H n X = 0 H_nX = 0 H n X = 0 for all but finitely many n n n .
Let X \in\operatorname{Coh}(\mathcal{A} \mathbin{\times\limits^{\rightarrow}_{\mathcal{C}}} \mathcal{B}) . Because \operatorname{Coh}(\mathcal{A} \mathbin{\times\limits^{\rightarrow}_{\mathcal{C}}} \mathcal{B}) has a bounded t t t -structure, condition (2) implies that X X X is equivalent to one of its truncations τ ≤ n X \tau_{\leq n}X τ ≤ n X , for some n n n . On the other hand, if m m m is the least integer such that H m X ≠ 0 H_mX \ne0 H m X = 0 , then H m ( X ) [ m ] ≃ τ ≤ m X H_m(X)[m] \simeq\tau_{\leq m}X H m ( X ) [ m ] ≃ τ ≤ m X . Since each truncation is obtained from the previous one by the fiber sequence,
H n + 1 ( X ) [ n + 1 ] ⟶ τ ≤ n + 1 X ⟶ τ ≤ n X , H_{n+1}(X)[n+1]\longrightarrow\tau_{\leq n+1}X\longrightarrow\tau_{\leq n}X, H n + 1 ( X ) [ n + 1 ] ⟶ τ ≤ n + 1 X ⟶ τ ≤ n X ,
we see that X X X is obtained by a series of extensions by shifts of its homology objects. Since H n X ∈ ( Coh ( A ) × → Coh ( C ) Coh ( B ) ) ♡ H_nX \in\left(\operatorname{Coh}(\mathcal{A}) \overrightarrow{\times}_{\operatorname{Coh}(\mathcal{C})} \operatorname{Coh}(\mathcal{B})\right)^\heartsuit H n X ∈ ( Coh ( A ) × Coh ( C ) Coh ( B ) ) ♡ , for all n n n by condition (1), and since Coh ( A ) × → Coh ( C ) Coh ( B ) \operatorname{Coh}(\mathcal{A}) \overrightarrow{\times}_{\operatorname{Coh}(\mathcal{C})} \operatorname{Coh}(\mathcal{B}) Coh ( A ) × Coh ( C ) Coh ( B ) is a stable subcategory, this implies that X ∈ Coh ( A × → C B ) X \in\operatorname{Coh}(\mathcal{A}\overrightarrow{\times}_{\mathcal{C}}\mathcal{B}) X ∈ Coh ( A × C B ) . On the other hand, it is obvious that every object of Coh ( A ) × → C B \operatorname{Coh}(\mathcal{A})\overrightarrow{\times}_{\mathcal{C}}\mathcal{B} Coh ( A ) × C B satisfies conditions (1) and (2). Since Coh ( A × → C B ) \operatorname{Coh}(\mathcal{A}\overrightarrow{\times}_{\mathcal{C}}\mathcal{B}) Coh ( A × C B ) and Coh ( A ) × → Coh ( C ) Coh ( B ) \operatorname{Coh}(\mathcal{A})\overrightarrow{\times}_{\operatorname{Coh}(\mathcal{C})}\operatorname{Coh}(\mathcal{B}) Coh ( A ) × Coh ( C ) Coh ( B ) are full subcategories of A × → C B \mathcal{A}\overrightarrow{\times}_{\mathcal{C}}\mathcal{B} A × C B containing the same objects, they are equivalent.
The argument that the t t t -structure on A × → C B \mathcal{A}\overrightarrow{\times}_{\mathcal{C}}\mathcal{B} A × C B is regular proceeds in exactly the same way. □
Lemma 7.6. Let A → F C ← G B \mathcal{A}\xrightarrow{F}\mathcal{C}\xleftarrow{G}\mathcal{B} A F C G B be a span in Reg t B S 1 \operatorname{Reg}_{t}^{BS^1} Reg t B S 1 , such that G : B → C G:\mathcal{B}\to\mathcal{C} G : B → C has a fully faithful, t t t -bounded right adjoint in Pr L \operatorname{Pr}^{L} Pr L , and F F F and G G G are t t t -exact and preserve compact objects. If A × C B \mathcal{A}\times_{\mathcal{C}}\mathcal{B} A × C B is compactly generated, then
A t T a t e × → C t T a t e B t T a t e ≃ ( A × C B ) t T a t e , \mathcal{A}^{t\mathrm{Tate}}\overrightarrow{\times}_{\mathcal{C}^{t\mathrm{Tate}}}\mathcal{B}^{t\mathrm{Tate}}\simeq(\mathcal{A}\times_{\mathcal{C}}\mathcal{B})^{t\mathrm{Tate}}, A t Tate × C t Tate B t Tate ≃ ( A × C B ) t Tate , where the renormalized Tate invariants on the right-hand side are taken with respect to the coherent t t t -structure defined in Lemma 7.4 , and all fiber products are taken in Pr L \operatorname{Pr}^{L} Pr L .
Proof. We first prove that
( A × → C B ) t T a t e ≃ A t T a t e × → C t T a t e B t T a t e , (\mathcal{A}\overrightarrow{\times}_{\mathcal{C}}\mathcal{B})^{t\mathrm{Tate}}\simeq\mathcal{A}^{t\mathrm{Tate}}\overrightarrow{\times}_{\mathcal{C}^{t\mathrm{Tate}}}\mathcal{B}^{t\mathrm{Tate}}, ( A × C B ) t Tate ≃ A t Tate × C t Tate B t Tate ,
where t T a t e t\mathrm{Tate} t Tate on the left-hand side is taken with respect to the regular t t t -structure defined in Lemma 7.4 . Since F F F and G G G preserve compact objects and A \mathcal{A} A and B \mathcal{B} B are compactly generated, [5 Proposition 13] gives an equivalence
Coh ( A × → C B ) ≃ ( A × → C B ) ω ≃ A ω × → C ω B ω ≃ Coh ( A ) × → Coh ( C ) Coh ( B ) . \operatorname{Coh}(\mathcal{A}\overrightarrow{\times}_{\mathcal{C}}\mathcal{B})\simeq(\mathcal{A}\overrightarrow{\times}_{\mathcal{C}}\mathcal{B})^\omega\simeq\mathcal{A}^\omega\overrightarrow{\times}_{\mathcal{C}^\omega}\mathcal{B}^\omega\simeq\operatorname{Coh}(\mathcal{A})\overrightarrow{\times}_{\operatorname{Coh}(\mathcal{C})}\operatorname{Coh}(\mathcal{B}). Coh ( A × C B ) ≃ ( A × C B ) ω ≃ A ω × C ω B ω ≃ Coh ( A ) × Coh ( C ) Coh ( B ) .
Since S 1 S^1 S 1 -invariants are defined by a limit operation, and limits commute with one another, we obtain an equivalence
Coh ( A × → C B ) S 1 ≃ Coh ( A ) S 1 × → Coh ( C ) S 1 Coh ( B ) S 1 . (47) \operatorname{Coh}(\mathcal{A}\overrightarrow{\times}_{\mathcal{C}}\mathcal{B})^{S^1}\simeq\operatorname{Coh}(\mathcal{A})^{S^1}\overrightarrow{\times}_{\operatorname{Coh}(\mathcal{C})^{S^1}}\operatorname{Coh}(\mathcal{B})^{S^1}. \tag*{(47)} Coh ( A × C B ) S 1 ≃ Coh ( A ) S 1 × Coh ( C ) S 1 Coh ( B ) S 1 . (47)
Now observe that the induced functors Ind ( F ∣ Coh ( A ) S 1 ) : Ind ( Coh ( A ) S 1 ) → Ind ( Coh ( C ) S 1 ) \operatorname{Ind}(F|_{\operatorname{Coh}(\mathcal{A})^{S^1}}):\operatorname{Ind}(\operatorname{Coh}(\mathcal{A})^{S^1})\to\operatorname{Ind}(\operatorname{Coh}(\mathcal{C})^{S^1}) Ind ( F ∣ Coh ( A ) S 1 ) : Ind ( Coh ( A ) S 1 ) → Ind ( Coh ( C ) S 1 ) and Ind ( G ∣ Coh ( B ) S 1 ) : Ind ( Coh ( B ) S 1 ) → Ind ( Coh ( C ) S 1 ) \operatorname{Ind}(G|_{\operatorname{Coh}(\mathcal{B})^{S^1}}):\operatorname{Ind}(\operatorname{Coh}(\mathcal{B})^{S^1})\to\operatorname{Ind}(\operatorname{Coh}(\mathcal{C})^{S^1}) Ind ( G ∣ Coh ( B ) S 1 ) : Ind ( Coh ( B ) S 1 ) → Ind ( Coh ( C ) S 1 ) obviously preserve compact objects. Thus, we have the sequence of equivalences,
Ind ( Coh ( A ) S 1 ) × → Ind ( Coh ( C ) S 1 ) Ind ( Coh ( B ) S 1 ) ≃ Ind ( Coh ( A ) S 1 × → Coh ( C ) S 1 Coh ( B ) S 1 ) (48) \operatorname{Ind}\left(\operatorname{Coh}(\mathcal{A})^{S^1}\right)\overrightarrow{\times}_{\operatorname{Ind}(\operatorname{Coh}(\mathcal{C})^{S^1})}\operatorname{Ind}\left(\operatorname{Coh}(\mathcal{B})^{S^1}\right)\simeq\operatorname{Ind}\left(\operatorname{Coh}(\mathcal{A})^{S^1}\overrightarrow{\times}_{\operatorname{Coh}(\mathcal{C})^{S^1}}\operatorname{Coh}(\mathcal{B})^{S^1}\right) \tag*{(48)} Ind ( Coh ( A ) S 1 ) × Ind ( Coh ( C ) S 1 ) Ind ( Coh ( B ) S 1 ) ≃ Ind ( Coh ( A ) S 1 × Coh ( C ) S 1 Coh ( B ) S 1 ) (48)
≃ Ind ( ( Coh ( A ) × → Coh ( C ) Coh ( B ) ) S 1 ) (49) \simeq\operatorname{Ind}\left(\left(\operatorname{Coh}(\mathcal{A})\overrightarrow{\times}_{\operatorname{Coh}(\mathcal{C})}\operatorname{Coh}(\mathcal{B})\right)^{S^1}\right) \tag*{(49)} ≃ Ind ( ( Coh ( A ) × Coh ( C ) Coh ( B ) ) S 1 ) (49)
≃ Ind ( Coh ( A × → C B ) S 1 ) , (50) \simeq\operatorname{Ind}\left(\operatorname{Coh}(\mathcal{A}\overrightarrow{\times}_{\mathcal{C}}\mathcal{B})^{S^1}\right), \tag*{(50)} ≃ Ind ( Coh ( A × C B ) S 1 ) , (50)
where (48) follows from a second application of [5 Proposition 13] ; (49) is the equivalence (47) established above; and (50) follows from the regularity of the t t t -structure on A × → C B \mathcal{A}\overrightarrow{\times}_{\mathcal{C}}\mathcal{B} A × C B . Since k [ u ± 1 ] -mod k[u^{\pm1}]\text{-mod} k [ u ± 1 ] -mod is dualizable as a k [ u ] k[u] k [ u ] -linear category, base change along k [ u ] -mod → k [ u ± 1 ] -mod k[u]\text{-mod}\to k[u^{\pm1}]\text{-mod} k [ u ] -mod → k [ u ± 1 ] -mod preserves limits. Using this, we deduce from the sequence of equivalences (48) –(50) above the equivalence,
( A × → C B ) t T a t e ≃ A t T a t e × → C t T a t e B t T a t e , (\mathcal{A}\overrightarrow{\times}_{\mathcal{C}}\mathcal{B})^{t\mathrm{Tate}}\simeq\mathcal{A}^{t\mathrm{Tate}}\overrightarrow{\times}_{\mathcal{C}^{t\mathrm{Tate}}}\mathcal{B}^{t\mathrm{Tate}}, ( A × C B ) t Tate ≃ A t Tate × C t Tate B t Tate ,
as desired.
Now, since G G G has a fully faithful right adjoint, denoted G R G^{R} G R , we apply [5 Corollary 12] to obtain a localization sequence of S 1 S^{1} S 1 -equivariant functors,
\mathcal{A}\times_{\mathcal{C}}\mathcal{B}\xrightarrow{i}\mathcal{A}\mathop{\times\limits^{\rightarrow}}_{\mathcal{C}}\mathcal{B}\xrightarrow{\pi}\mathcal{C}.
In order to apply Theorem 3.12 and obtain a localization after applying ( − ) t T a t e (-)^{\mathrm{tTate}} ( − ) tTate , we show that π R \pi^{R} π R and i R i^{R} i R are t t t -bounded. Using the decomposition π R = r ∘ G R \pi^{R}=r\circ G^{R} π R = r ∘ G R found therein,^51 we see that π R \pi^{R} π R is t t t -bounded if and only if G G G and G R G^{R} G R are (as it is clear from the definition that r r r is t t t -bounded if and only if G G G is). Since both G G G and G R G^{R} G R are t t t -bounded by assumption, it follows that π R \pi^{R} π R is t t t -bounded. On the other hand, the functor i i i is a t t t -exact inclusion, so its right adjoint i R i^{R} i R is left t t t -exact, hence left t t t -bounded. To obtain a right t t t -bound on i R i^{R} i R , consider the fiber sequence
i i R ⟶ id ⟶ π R π . ii^{R}\longrightarrow\operatorname{id}\longrightarrow\pi^{R}\pi. i i R ⟶ id ⟶ π R π .
Since π R \pi^{R} π R and π \pi π are t t t -bounded, we see that the fiber i i R ii^{R} i i R is t t t -bounded. Since, i i i furthermore is t t t -exact, we see that i R i^{R} i R is t t t -bounded. Thus, Theorem 3.12 applies, and we obtain a localization sequence
(\mathcal{A}\times_{\mathcal{C}}\mathcal{B})^{\mathrm{tTate}}\longrightarrow(\mathcal{A}\mathop{\times\limits^{\rightarrow}}_{\mathcal{C}}\mathcal{B})^{\mathrm{tTate}}\xrightarrow{\pi^{\mathrm{tTate}}}\mathcal{C}^{\mathrm{tTate}} \tag*{(51)}
Under the identification (\mathcal{A}\mathop{\times\limits^{\rightarrow}}_{\mathcal{C}}\mathcal{B})^{\mathrm{tTate}}\simeq\mathcal{A}^{\mathrm{tTate}}\mathop{\times\limits^{\rightarrow}}_{\mathcal{C}^{\mathrm{tTate}}}\mathcal{B}^{\mathrm{tTate}} , the right-hand functor π t T a t e \pi^{\mathrm{tTate}} π tTate corresponds to the canonical map
\pi:\mathcal{A}^{\mathrm{tTate}}\mathop{\times\limits^{\rightarrow}}_{\mathcal{C}^{\mathrm{tTate}}}\mathcal{B}^{\mathrm{tTate}}\longrightarrow\mathcal{C}^{\mathrm{tTate}},
from the lax product of the span A t T a t e ⟶ C t T a t e ⟵ B t T a t e \mathcal{A}^{\mathrm{tTate}}\longrightarrow\mathcal{C}^{\mathrm{tTate}}\longleftarrow\mathcal{B}^{\mathrm{tTate}} A tTate ⟶ C tTate ⟵ B tTate to the base C t T a t e \mathcal{C}^{\mathrm{tTate}} C tTate . The fiber of this map is precisely A t T a t e × C t T a t e B t T a t e \mathcal{A}^{\mathrm{tTate}}\times_{\mathcal{C}^{\mathrm{tTate}}}\mathcal{B}^{\mathrm{tTate}} A tTate × C tTate B tTate , so we conclude from the localization sequence (51) that A t T a t e × C t T a t e B t T a t e ≃ ( A × C B ) t T a t e \mathcal{A}^{\mathrm{tTate}}\times_{\mathcal{C}^{\mathrm{tTate}}}\mathcal{B}^{\mathrm{tTate}}\simeq(\mathcal{A}\times_{\mathcal{C}}\mathcal{B})^{\mathrm{tTate}} A tTate × C tTate B tTate ≃ ( A × C B ) tTate , as desired. □
Given a six-functor formalism, we may define the following presheaves of categories.
Definition 7.13. Let D \mathcal{D} D be a six-functor formalism, and let X X X be a stack. Let Z ↪ X Z\hookrightarrow X Z ↪ X be a closed immersion, and let U ↪ X U\hookrightarrow X U ↪ X its open complement. For each Y ∈ Rep / X Y\in\operatorname{Rep}/X Y ∈ Rep / X , the assignment of the category
D Z ( Y ) : = fib ( j Y ! : D ( Y ) ⟶ D ( U × X Y ) ) , \mathcal{D}_{Z}(Y):=\operatorname{fib}(j_{Y}^{!}:\mathcal{D}(Y)\longrightarrow\mathcal{D}(U\times_{X}Y)), D Z ( Y ) := fib ( j Y ! : D ( Y ) ⟶ D ( U × X Y )) , where j Y : U × X Y ↪ Y j_{Y}:U\times_{X}Y\hookrightarrow Y j Y : U × X Y ↪ Y denotes the canonical open immersion, defines a presheaf on Rep / X \operatorname{Rep}/X Rep / X .
Definition 7.15. Let D \mathcal{D} D be a six-functor formalism, and let X X X be a stack. For a fixed Z ↪ X Z\hookrightarrow X Z ↪ X , the assignment to each Y ∈ Rep / X Y\in\operatorname{Rep}/X Y ∈ Rep / X of the category D ( Z × X Y ) \mathcal{D}(Z\times_{X}Y) D ( Z × X Y ) defines a presheaf which we denote by D [ Z ] ( − ) : Rep / X ⟶ DGCat k \mathcal{D}[Z](-):\operatorname{Rep}/X\longrightarrow\operatorname{DGCat}_{k} D [ Z ] ( − ) : Rep / X ⟶ DGCat k .
For a given closed immersion Z ↪ X Z\hookrightarrow X Z ↪ X , there is an obvious natural transformation of presheaves, D [ Z ] ( − ) ⟶ D Z ( − ) \mathcal{D}[Z](-)\longrightarrow\mathcal{D}_{Z}(-) D [ Z ] ( − ) ⟶ D Z ( − ) , whose component at Y ∈ Rep / X Y\in\operatorname{Rep}/X Y ∈ Rep / X is given by pushforward along the inclusion Z × X Y ↪ Y Z\times_{X}Y\hookrightarrow Y Z × X Y ↪ Y .
Lemma 7.16. Let X X X be a stack with affine diagonal, and let D : Rep / X → DGCat R \mathcal{D} : \operatorname{Rep}/X \to\operatorname{DGCat}_{R} D : Rep / X → DGCat R be a six-functor formalism satisfying Nisnevich descent, where R R R is any E ∞ \mathbb{E}_{\infty} E ∞ -algebra over k k k . For any closed immersion Z ↪ X Z \hookrightarrow X Z ↪ X , the presheaf D Z \mathcal{D}_{Z} D Z satisfies Nisnevich descent.
Proof. As before, it suffices to show that D Z ( − ) \mathcal{D}_{Z}(-) D Z ( − ) satisfies Nisnevich excision. Let
W → V ↓ ↓ U → S \begin{CD}
W @>>> V \\
@VVV @VVV \\
U @>>> S
\end{CD} W ↓ ⏐ U V ↓ ⏐ S
be a Nisnevich square in Rep / X \operatorname{Rep}/X Rep / X . For Y ∈ Rep / X Y \in\operatorname{Rep}/X Y ∈ Rep / X , denote by Y c Y^{c} Y c the complement of the closed immersion Z × X Y ↪ Y Z \times_{X} Y \hookrightarrow Y Z × X Y ↪ Y . Observe that the square
W c → V c ↓ ↓ U c → S c \begin{CD}
W^{c} @>>> V^{c} \\
@VVV @VVV \\
U^{c} @>>> S^{c}
\end{CD} W c ↓ ⏐ U c V c ↓ ⏐ S c
is also Nisnevich. Indeed, it is Cartesian because we may write W c ≃ W × S U c W^{c} \simeq W \times_{S} U^{c} W c ≃ W × S U c . Moreover, the base changes V c → S c V^{c} \to S^{c} V c → S c and U c → S c U^{c} \to S^{c} U c → S c are respectively étale and open, as these properties are preserved by base change. Similarly, if V → S V \to S V → S was an isomorphism away from U U U , then its base change will also be an isomorphism away from the base change of U U U . Thus, by descent for D \mathcal{D} D , we have pullbacks
D ( S ) = D ( V ) × D ( W ) D ( U ) \mathcal{D}(S) = \mathcal{D}(V) \times_{\mathcal{D}(W)} \mathcal{D}(U) D ( S ) = D ( V ) × D ( W ) D ( U )
and
D ( S c ) = D ( V c ) × D ( W c ) D ( U c ) . \mathcal{D}(S^{c}) = \mathcal{D}(V^{c}) \times_{\mathcal{D}(W^{c})} \mathcal{D}(U^{c}). D ( S c ) = D ( V c ) × D ( W c ) D ( U c ) .
In order to identify
D Z ( S ) = D Z ( V ) × D Z ( W ) D Z ( U ) , \mathcal{D}_{Z}(S) = \mathcal{D}_{Z}(V) \times_{\mathcal{D}_{Z}(W)} \mathcal{D}_{Z}(U), D Z ( S ) = D Z ( V ) × D Z ( W ) D Z ( U ) ,
we need only check that the kernel of the map D ( S ) → D ( S c ) \mathcal{D}(S) \to\mathcal{D}(S^{c}) D ( S ) → D ( S c ) induced by j ! ∣ V × j ! ∣ U j^{!}|_{V} \times j^{!}|_{U} j ! ∣ V × j ! ∣ U has an analogous presentation as a fiber product. We now use the following general observation. Let Λ \Lambda Λ denote the category ∗ → ∗ ← ∗ * \to* \leftarrow* ∗ → ∗ ← ∗ . Then lim Λ \lim_{\Lambda} lim Λ sends diagrams of shape Λ \Lambda Λ to their pullbacks. Moreover, if we have a diagram A → C ← B A \to C \leftarrow B A → C ← B in Map ( Λ , Cat ∞ ) \operatorname{Map}(\Lambda, \operatorname{Cat}_{\infty}) Map ( Λ , Cat ∞ ) , where A = A 1 → A 3 ← A 2 A = A_{1} \to A_{3} \leftarrow A_{2} A = A 1 → A 3 ← A 2 and similarly for B , C B,C B , C , then we have an object of Map ( Λ , Map ( Λ , Cat ∞ ) ) ≃ Map ( Λ × Λ , Cat ∞ ) \operatorname{Map}(\Lambda,\operatorname{Map}(\Lambda,\operatorname{Cat}_{\infty})) \simeq\operatorname{Map}(\Lambda\times\Lambda,\operatorname{Cat}_{\infty}) Map ( Λ , Map ( Λ , Cat ∞ )) ≃ Map ( Λ × Λ , Cat ∞ ) . Writing this diagram out, we have
A 1 → B 1 ← C 1 ↓ ↓ ↓ A 3 → B 3 ← C 3 ↑ ↑ ↑ A 2 → B 2 ← C 2 \begin{CD}
A_{1} @>>> B_{1} @<<< C_{1} \\
@VVV @VVV @VVV \\
A_{3} @>>> B_{3} @<<< C_{3} \\
@AAA @AAA @AAA \\
A_{2} @>>> B_{2} @<<< C_{2}
\end{CD} A 1 ↓ ⏐ A 3 ⏐ ↑ A 2 B 1 ↓ ⏐ B 3 ⏐ ↑ B 2 C 1 ↓ ⏐ C 3 ⏐ ↑ C 2
Applying lim Λ × Λ \lim_{\Lambda\times\Lambda} lim Λ × Λ to this diagram and using that limits commute with limits, we find
lim Λ × Λ ( A → C ← B ) = ( lim Λ A ) × lim Λ C ( lim Λ B ) = ( A 1 × C 1 B 1 ) × A 3 × B 3 C 3 ( A 2 × C 2 B 2 ) . \lim_{\Lambda\times\Lambda}(A \to C \leftarrow B) = (\lim_{\Lambda} A) \times_{\lim_{\Lambda} C} (\lim_{\Lambda} B) = (A_{1} \times_{C_{1}} B_{1}) \times_{A_{3} \times_{B_{3}} C_{3}} (A_{2} \times_{C_{2}} B_{2}). Λ × Λ lim ( A → C ← B ) = ( Λ lim A ) × l i m Λ C ( Λ lim B ) = ( A 1 × C 1 B 1 ) × A 3 × B 3 C 3 ( A 2 × C 2 B 2 ) .
Now, let A A A be the span D ( V ) → D ( V c ) ← 0 \mathcal{D}(V) \to\mathcal{D}(V^{c}) \leftarrow0 D ( V ) → D ( V c ) ← 0 , C C C the span D ( W ) → D ( W c ) ← 0 \mathcal{D}(W) \to\mathcal{D}(W^{c}) \leftarrow0 D ( W ) → D ( W c ) ← 0 , and B B B the span D ( U ) → D ( U c ) ← 0 \mathcal{D}(U) \to\mathcal{D}(U^{c}) \leftarrow0 D ( U ) → D ( U c ) ← 0 . □
The main theorem
Theorem 7.17. Let X X X be a linearly scalloped stack almost of finite type and with affine diagonal, and let i : Z ↪ X i: Z \hookrightarrow X i : Z ↪ X be an almost finitely presented inclusion of a closed substack. Then pushforward along i i i induces an equivalence,
i ∗ : D p o t ( Z ) → ∼ D Z p o t ( X ) . i_*:\mathcal{D}^{\mathrm{pot}}(Z)\xrightarrow{\sim}\mathcal{D}^{\mathrm{pot}}_Z(X). i ∗ : D pot ( Z ) ∼ D Z pot ( X ) .
Proof. Note that X X X is in particular a quasi-compact algebraic stack with affine diagonal; hence, it is geometric. We may thus use the results of § 7.1 . We will proceed by induction on scallop diagrams. Recall that a linearly scalloped stack X X X admits a finite filtration by quasi-compact open substacks
∅ = X 0 ⊂ X 1 ⊂ ⋯ ⊂ X n = X \varnothing=X_0\subset X_1\subset\cdots\subset X_n=X ∅ = X 0 ⊂ X 1 ⊂ ⋯ ⊂ X n = X
along with Nisnevich squares
W i → V i ↓ u i ↓ X i − 1 → X i (52) \begin{CD}
W_i @>>> V_i\\
@VVV @V{u_i}VV\\
X_{i-1} @>>> X_i \tag*{(52)}
\end{CD} W i ↓ ⏐ X i − 1 V i u i ↓ ⏐ X i (52)
where the u i u_i u i are representable (by algebraic spaces) étale morphisms of finite presentation and the V i V_i V i are linearly quasi-fundamental, which means that each V i V_i V i admits a quasi-projective morphism to B G i BG_i B G i for a linearly reductive embeddable group scheme G i G_i G i .52
We recall the presheaf D p o t [ Z ] \mathcal{D}^{\mathrm{pot}}[Z] D pot [ Z ] defined in Definition 7.15 above, and the natural transformation D p o t [ Z ] ( − ) → D Z p o t ( − ) \mathcal{D}^{\mathrm{pot}}[Z](-)\to\mathcal{D}^{\mathrm{pot}}_Z(-) D pot [ Z ] ( − ) → D Z pot ( − ) , whose component at Y ∈ Rep / X Y\in\operatorname{Rep}/X Y ∈ Rep / X is given by the pushforward, ( i Z × X Y ) ∗ : D p o t ( Z × X Y ) → D Z × X Y p o t ( Y ) (i_{Z\times_XY})_*:\mathcal{D}^{\mathrm{pot}}(Z\times_XY)\to\mathcal{D}^{\mathrm{pot}}_{Z\times_XY}(Y) ( i Z × X Y ) ∗ : D pot ( Z × X Y ) → D Z × X Y pot ( Y ) . For a given Y ∈ Rep / X Y\in\operatorname{Rep}/X Y ∈ Rep / X , let i Y : Z × X Y → Y i_Y:Z\times_XY\to Y i Y : Z × X Y → Y denote the canonical closed immersion. Since D p o t [ Z ] ( − ) → D Z p o t ( − ) \mathcal{D}^{\mathrm{pot}}[Z](-)\to\mathcal{D}^{\mathrm{pot}}_Z(-) D pot [ Z ] ( − ) → D Z pot ( − ) is a map of presheaves, from the Nisnevich square (52) , we obtain the following commutative diagram,
D p o t ( Z × X X i ) → ( 1 ) D p o t ( Z × X V i ) × D p o t ( Z × X W i ) D p o t ( Z × X X i − 1 ) ( i X i ) ∗ ↓ ↓ ( i V i ) ∗ × ( i W i ) ∗ × ( i X i − 1 ) ∗ D Z p o t ( X i ) → ( 2 ) D Z p o t ( V i ) × D Z p o t ( W i ) D Z p o t ( X i − 1 ) \begin{CD}
\mathcal{D}^{\mathrm{pot}}(Z\times_X X_i) @>{(1)}>> \mathcal{D}^{\mathrm{pot}}(Z\times_X V_i)\times_{\mathcal{D}^{\mathrm{pot}}(Z\times_XW_i)}\mathcal{D}^{\mathrm{pot}}(Z\times_X X_{i-1})\\
@V{(i_{X_i})_*}VV @VV{(i_{V_i})_*\times(i_{W_i})_*\times(i_{X_{i-1}})_*}V\\
\mathcal{D}^{\mathrm{pot}}_{Z}(X_i) @>{(2)}>> \mathcal{D}^{\mathrm{pot}}_{Z}(V_i)\times_{\mathcal{D}^{\mathrm{pot}}_{Z}(W_i)}\mathcal{D}^{\mathrm{pot}}_{Z}(X_{i-1})
\end{CD} D pot ( Z × X X i ) ( i X i ) ∗ ↓ ⏐ D Z pot ( X i ) ( 1 ) ( 2 ) D pot ( Z × X V i ) × D pot ( Z × X W i ) D pot ( Z × X X i − 1 ) ↓ ⏐ ( i V i ) ∗ × ( i W i ) ∗ × ( i X i − 1 ) ∗ D Z pot ( V i ) × D Z pot ( W i ) D Z pot ( X i − 1 )
where the horizontal arrow (1) and (2) are the functors induced by the universal property of the fiber products on the right-hand side of the diagram. By Lemma 7.16 , the presheaf D Z p o t ( − ) \mathcal{D}^{\mathrm{pot}}_Z(-) D Z pot ( − ) satisfies Nisnevich excision, so arrow (2) is actually an equivalence. Similarly, D p o t ( − ) \mathcal{D}^{\mathrm{pot}}(-) D pot ( − ) satisfies Nisnevich excision by Proposition 7.2, so arrow (1) is also an equivalence. Thus, in order to prove that i ∗ i_* i ∗ is an equivalence, it suffices to show that ( i V i ) ∗ × ( i W i ) ∗ ( i X i − 1 ) ∗ (i_{V_i})_*\times_{(i_{W_i})_*}(i_{X_{i-1}})_* ( i V i ) ∗ × ( i W i ) ∗ ( i X i − 1 ) ∗ is an equivalence.
We proceed by induction on i i i . For the case i = 1 i=1 i = 1 , the Nisnevich square (52) shows that X 1 X_1 X 1 is quasi-fundamental, since the u i u_i u i are isomorphisms away from X i − 1 X_{i-1} X i − 1 by the definition of Nisnevich square. Hence, X 1 → B G 1 X_1\to BG_1 X 1 → B G 1 is a quasi-projective morphism for G 1 G_1 G 1 a linearly reductive embeddable group. In particular, X 1 ≃ U 1 / G 1 X_1\simeq U_1/G_1 X 1 ≃ U 1 / G 1 for some G 1 G_1 G 1 -quasi-projective scheme U 1 U_1 U 1 , so Theorem 6.7 implies that
( i X 1 ) ∗ : D p o t ( Z 1 ) → ∼ D Z 1 p o t ( X 1 ) = : D Z p o t ( X 1 ) . (i_{X_1})_*:\mathcal{D}^{\mathrm{pot}}(Z_1)\xrightarrow{\sim}\mathcal{D}^{\mathrm{pot}}_{Z_1}(X_1)=:\mathcal{D}^{\mathrm{pot}}_Z(X_1). ( i X 1 ) ∗ : D pot ( Z 1 ) ∼ D Z 1 pot ( X 1 ) =: D Z pot ( X 1 ) .
Now let n ≥ i > 1 n\ge i>1 n ≥ i > 1 , and assume as our induction hypothesis that
( i X i − 1 ) ∗ : D p o t ( Z × X X i − 1 ) → D Z p o t ( X i − 1 ) (i_{X_{i-1}})_*:\mathcal{D}^{\mathrm{pot}}(Z\times_X X_{i-1})\to\mathcal{D}^{\mathrm{pot}}_Z(X_{i-1}) ( i X i − 1 ) ∗ : D pot ( Z × X X i − 1 ) → D Z pot ( X i − 1 )
is an equivalence. We now observe that, by the assumption that X X X is linearly scalloped, V i V_i V i admits a quasi-projective morphism V i → B G i V_i\to BG_i V i → B G i for some linearly reductive embeddable group scheme G i G_i G i , hence has the form U i / G i U_i/G_i U i / G i for some G i G_i G i -quasi-projective scheme U i U_i U i , as above. In the same vein, the map W i → V i W_i \to V_i W i → V i is an open immersion,53 so W i W_i W i also admits a quasi-projective morphism to B G i BG_i B G i , meaning it is also the quotient of a G i G_i G i -quasi-projective scheme. Thus, Theorem 6.7 implies that both ( i V i ) ∗ : D p o t ( Z × X V i ) → D Z p o t ( V i ) (i_{V_i})_*: \mathop{\mathrm{D}}\nolimits^{\mathrm{pot}}(Z \times_X V_i) \to\mathop{\mathrm{D}}\nolimits_Z^{\mathrm{pot}}(V_i) ( i V i ) ∗ : D pot ( Z × X V i ) → D Z pot ( V i ) and ( i W i ) ∗ : D p o t ( Z × X W i ) → D Z p o t ( W i ) (i_{W_i})_*: \mathop{\mathrm{D}}\nolimits^{\mathrm{pot}}(Z \times_X W_i) \to\mathop{\mathrm{D}}\nolimits_Z^{\mathrm{pot}}(W_i) ( i W i ) ∗ : D pot ( Z × X W i ) → D Z pot ( W i ) are equivalences. These two equivalences together with the induction hypothesis imply that ( i V i ) ∗ × ( i W i ) ∗ ( i X i − 1 ) ∗ (i_{V_i})_* \times_{(i_{W_i})_*} (i_{X_{i-1}})_* ( i V i ) ∗ × ( i W i ) ∗ ( i X i − 1 ) ∗ is an equivalence.
Taking i = n i=n i = n , we therefore obtain the desired equivalence, i ∗ : D p o t ( Z ) → D Z p o t ( X ) i_*: \mathop{\mathrm{D}}\nolimits^{\mathrm{pot}}(Z) \to\mathop{\mathrm{D}}\nolimits_Z^{\mathrm{pot}}(X) i ∗ : D pot ( Z ) → D Z pot ( X ) . □