Introduction
In this paper, we develop a deformation theory of higher categories, which provides a generalization of the deformation theory in the context of stable ( ∞ , 1 ) (\infty,1) ( ∞ , 1 ) -categories to the setting of stable ( ∞ , n ) (\infty,n) ( ∞ , n ) -categories for n > 1 n > 1 n > 1 . Informally, an ( ∞ , n ) (\infty,n) ( ∞ , n ) -category is a higher category which possesses i i i -morphisms between ( i − 1 ) (i-1) ( i − 1 ) -morphisms for every i ≥ 1 i \ge1 i ≥ 1 (0 0 0 -morphisms are objects) such that i i i -morphisms are invertible for i > n i > n i > n . The ( ∞ , n ) (\infty,n) ( ∞ , n ) -category theory is extensively studied from various respects; see e.g. [2 ] , [14 ] , [22 ] , [26 ] , [29 ] . To explain the purpose of this paper, let us briefly review the deformation theory of stable ( ∞ , 1 ) (\infty,1) ( ∞ , 1 ) -categories (that is, stable ∞ \infty ∞ -categories) with an emphasis on its difficulties. Let D \mathcal{D} D be a k k k -linear stable ∞ \infty ∞ -category over a field k k k (or a k k k -linear dg category), and let R R R be an Artin algebra equipped with a homomorphism R → k R \to k R → k . The deformation theory of D \mathcal{D} D studies deformations (or liftings) ( D R , D R ⊗ R k ≃ D ) (\mathcal{D}_{R}, \mathcal{D}_{R} \otimes_{R} k \simeq\mathcal{D}) ( D R , D R ⊗ R k ≃ D ) of D \mathcal{D} D to R R R , where D R \mathcal{D}_{R} D R is an R R R -linear stable ∞ \infty ∞ -category. These are analyzed by means of algebraic invariants such as Hochschild cohomology H H ∙ ( D / k ) \mathrm{HH}^{\bullet}(\mathcal{D}/k) HH ∙ ( D / k ) . It is well known that unlike the deformation theory of associative algebras [12 ] , [13 ] and more general E n E_{n} E n -algebras (algebras over little n n n -disk operads) [8 ] , the deformation theory of plain k k k -linear stable ∞ \infty ∞ -categories is not well-behaved. Indeed, the deformation problems of plain k k k -linear stable ∞ \infty ∞ -categories are in general not formal moduli problems, and they often cannot be described in terms of Hochschild cohomology: a typical example is the derived ∞ \infty ∞ -category of coherent sheaves on Spec k [ x ] / ( x 2 ) \operatorname{Spec} k[x]/(x^{2}) Spec k [ x ] / ( x 2 ) . Consequently, a good deformation theory essentially requires refined data. In [20 ] , we developed a deformation theory of k k k -linear stable ∞ \infty ∞ -categories equipped with left complete t t t -structures and showed that this deformation theory works well: the deformation problems are formal moduli problems, and they are described by means of Hochschild cohomology of the underlying stable ∞ \infty ∞ -categories. We turn to the setting of higher categories, that is, stable ( ∞ , n ) (\infty,n) ( ∞ , n ) -categories. The above difficulties persist in this context. Even worse, it was shown in [7 ] , using an argument of [25 X] , that the deformation theory of plain stable ( ∞ , n ) (\infty,n) ( ∞ , n ) -categories is in general only an ( n + 1 ) (n+1) ( n + 1 ) -pro formaed formal moduli problem. Roughly speaking, this means that the larger n n n is, the further the deformation problem is from being a formal moduli problem, and hence the larger the a priori discrepancy between the deformation problem and the Hochschild cohomology. We are therefore led to the following question:
What is an appropriate framework in which the deformation problem for ( ∞ , n ) (\infty,n) ( ∞ , n ) -categories becomes well-behaved?
To address this question, we introduce the notion of a t t t -structured stable presentable ( ∞ , n ) (\infty,n) ( ∞ , n ) -category. We then develop a deformation theory of complete t t t -structured k k k -linear ( ∞ , n ) (\infty,n) ( ∞ , n ) -categories by establishing a categorified Koszul duality for them.
Categorified Koszul duality and t t t -structured stable presentable higher categories. We now consider the categorified Koszul duality from the viewpoint of deformation theory. If A A A is an ordinary associative k k k -algebra (more generally, a dg algebra satisfying suitable conditions), the deformation theory of A A A -modules can be understood via a classical Koszul duality (cf. [25 X, Section 5.2] ): for an Artin local k k k -algebra R R R with R → k R \to k R → k and its Koszul dual D 1 ( R ) = End R ( k ) \mathbb{D}_1(R) = \operatorname{End}_R(k) D 1 ( R ) = End R ( k ) , Koszul duality gives an equivalence between right R R R -modules and left D 1 ( R ) \mathbb{D}_1(R) D 1 ( R ) -modules in a suitable formulation. The Koszul duality for k k k -linear stable ∞ \infty ∞ -categories generalizes this equivalence to the following one
right LMod R ⊗ -module objects in Pr k t ± ≃ left LMod D 2 ( R ) ⊗ -module objects in Pr k t ± \text{right }\operatorname{LMod}^{\otimes}_R\text{-module objects in }\operatorname{Pr}^{t\pm}_k \simeq\text{left }\operatorname{LMod}^{\otimes}_{\mathbb{D}_2(R)}\text{-module objects in }\operatorname{Pr}^{t\pm}_k right LMod R ⊗ -module objects in Pr k t ± ≃ left LMod D 2 ( R ) ⊗ -module objects in Pr k t ±
where Pr k t ± \operatorname{Pr}^{t\pm}_k Pr k t ± is the ∞ \infty ∞ -category of k k k -linear stable presentable ∞ \infty ∞ -categories equipped with both left and right complete t t t -structures, and LMod R ⊗ \operatorname{LMod}^{\otimes}_R LMod R ⊗ (resp. LMod D 2 ( R ) ⊗ \operatorname{LMod}^{\otimes}_{\mathbb{D}_2(R)} LMod D 2 ( R ) ⊗ ) is the monoidal ∞ \infty ∞ -category of left R R R -modules (resp. left D 2 ( R ) \mathbb{D}_2(R) D 2 ( R ) -modules) with the right complete t t t -structure. Here D 2 ( R ) \mathbb{D}_2(R) D 2 ( R ) is the E 2 \mathbb{E}_2 E 2 -Koszul dual of R R R . The t t t -structures play the role of filtrations (or topologies) in classical Koszul duality, and t t t -structures are essential for realizing the categorified Koszul duality (see [20 ] for details). We turn to k k k -linear stable ( ∞ , n ) (\infty,n) ( ∞ , n ) -categories. We use the notion of a presentable ( ∞ , n ) (\infty,n) ( ∞ , n ) -category introduced in [30 ] , see also [1 ] for a useful reformulation. Let us focus on ( ∞ , 2 ) (\infty,2) ( ∞ , 2 ) -categories. Roughly speaking, a k k k -linear (stable) presentable ( ∞ , 2 ) (\infty,2) ( ∞ , 2 ) -category is an ∞ \infty ∞ -category D \mathcal{D} D admitting small colimits and equipped with a Pr k \operatorname{Pr}_k Pr k -module structure Pr k × D → D \operatorname{Pr}_k \times\mathcal{D} \to\mathcal{D} Pr k × D → D satisfying a certain finiteness condition, where Pr k \operatorname{Pr}_k Pr k is the symmetric monoidal ∞ \infty ∞ -category of k k k -linear presentable ∞ \infty ∞ -categories. A complete t t t -structured k k k -linear (stable) presentable ( ∞ , 2 ) (\infty,2) ( ∞ , 2 ) -category is an ∞ \infty ∞ -category D \mathcal{D} D admitting small colimits and equipped with a Pr k t ± \operatorname{Pr}^{t\pm}_k Pr k t ± -module structure Pr k t ± × D → D \operatorname{Pr}^{t\pm}_k \times\mathcal{D} \to\mathcal{D} Pr k t ± × D → D satisfying a certain finiteness condition. The notion of a (complete) t t t -structured k k k -linear (stable) presentable ( ∞ , n ) (\infty,n) ( ∞ , n ) -category for arbitrary n n n is defined along the same lines. We prove a higher categorified Koszul duality (i.e., Koszul duality for ( ∞ , n ) (\infty,n) ( ∞ , n ) -categories), stated informally as follows (see Theorem 4.19 ):
Theorem 1.1. Let n ≥ 1 n \ge1 n ≥ 1 and let R R R be an almost Artin E n + 1 \mathbb{E}_{n+1} E n + 1 -algebra. Let D n + 1 ( R ) \mathbb{D}_{n+1}(R) D n + 1 ( R ) denote the E n + 1 \mathbb{E}_{n+1} E n + 1 -Koszul dual. Then there exists an equivalence
right ( n − 1 ) Pr R t ± -module objects in n Pr k t ± ≃ left ( n − 1 ) Pr D n + 1 ( R ) t ± -module objects in n Pr k t ± \text{right }(n-1)\operatorname{Pr}^{t\pm}_R\text{-module objects in }n\operatorname{Pr}^{t\pm}_k \simeq\text{left }(n-1)\operatorname{Pr}^{t\pm}_{\mathbb{D}_{n+1}(R)}\text{-module objects in }n\operatorname{Pr}^{t\pm}_k right ( n − 1 ) Pr R t ± -module objects in n Pr k t ± ≃ left ( n − 1 ) Pr D n + 1 ( R ) t ± -module objects in n Pr k t ± Roughly, for a complete t t t -structured stable presentable ( ∞ , n ) (\infty,n) ( ∞ , n ) -category C \mathcal{C} C , a (right) ( n − 1 ) Pr R t ± (n-1)\operatorname{Pr}^{t\pm}_R ( n − 1 ) Pr R t ± -module structure on C \mathcal{C} C provides a (right) R R R -linear structure on C \mathcal{C} C . Similarly, a (left) ( n − 1 ) Pr D n + 1 ( R ) t ± (n-1)\operatorname{Pr}^{t\pm}_{\mathbb{D}_{n+1}(R)} ( n − 1 ) Pr D n + 1 ( R ) t ± -module structure gives a (left) D n + 1 ( R ) \mathbb{D}_{n+1}(R) D n + 1 ( R ) -linear structure. Thus, Theorem 1.1 gives an equivalence between right R R R -linear complete t t t -structured presentable ( ∞ , n ) (\infty,n) ( ∞ , n ) -categories and left D n + 1 ( R ) \mathbb{D}_{n+1}(R) D n + 1 ( R ) -linear complete t t t -structured presentable ( ∞ , n ) (\infty,n) ( ∞ , n ) -categories. The symbol n Pr k t ± n\operatorname{Pr}^{t\pm}_k n Pr k t ± denotes the ∞ \infty ∞ -category of complete t t t -structured k k k -linear presentable ( ∞ , n ) (\infty,n) ( ∞ , n ) -categories, and ( n − 1 ) Pr R t ± (n-1)\operatorname{Pr}^{t\pm}_R ( n − 1 ) Pr R t ± and ( n − 1 ) Pr D n + 1 ( R ) t ± (n-1)\operatorname{Pr}^{t\pm}_{\mathbb{D}_{n+1}(R)} ( n − 1 ) Pr D n + 1 ( R ) t ± are higher module categories: for example, 0 Pr R t ± 0\operatorname{Pr}^{t\pm}_R 0 Pr R t ± is the LMod R \operatorname{LMod}_R LMod R with t t t -structure determined by the full subcategory of connective objects, 1 Pr R t ± = LMod LMod R ( Pr k t ± ) 1\operatorname{Pr}^{t\pm}_R = \operatorname{LMod}_{\operatorname{LMod}_R}(\operatorname{Pr}^{t\pm}_k) 1 Pr R t ± = LMod LMod R ( Pr k t ± ) , 2 Pr R t ± = LMod 1 Pr R t ± ( 2 Pr k t ± ) 2\operatorname{Pr}^{t\pm}_R = \operatorname{LMod}_{1\operatorname{Pr}^{t\pm}_R}(2\operatorname{Pr}^{t\pm}_k) 2 Pr R t ± = LMod 1 Pr R t ± ( 2 Pr k t ± ) and so on. Here LMod X ( P ) \operatorname{LMod}_X(P) LMod X ( P ) denotes the ∞ \infty ∞ -category of left X X X -module object in P P P , and 2 Pr k t ± 2\operatorname{Pr}^{t\pm}_k 2 Pr k t ± is the symmetric monoidal category of complete t t t -structured k k k -linear presentable ( ∞ , 2 ) (\infty,2) ( ∞ , 2 ) -categories (see Section 3.4.1). When n = 1 n=1 n = 1 , the result is the one above.
Deformation theory. We will give an application of this categorified Koszul duality to the deformation theory of ( ∞ , n ) (\infty,n) ( ∞ , n ) -categorical context. Let D \mathcal{D} D be a complete t t t -structured k k k -linear presentable ( ∞ , n ) (\infty,n) ( ∞ , n ) -category, that is, an object of n Pr k t ± n\operatorname{Pr}^{t\pm}_k n Pr k t ± . For an Artin E n + 1 \mathbb{E}_{n+1} E n + 1 -algebra R R R , a deformation of D \mathcal{D} D to R R R is a right ( n − 1 ) Pr R t ± (n-1)\operatorname{Pr}^{t\pm}_R ( n − 1 ) Pr R t ± -module D R \mathcal{D}_R D R in n Pr k t ± n\operatorname{Pr}^{t\pm}_k n Pr k t ± , together with an equivalence ϕ : D R ⊗ ( n − 1 ) Pr R t ± ( n − 1 ) Pr k t ± ≃ D \phi: \mathcal{D}_R \otimes_{(n-1)\operatorname{Pr}^{t\pm}_R} (n-1)\operatorname{Pr}^{t\pm}_k \simeq\mathcal{D} ϕ : D R ⊗ ( n − 1 ) Pr R t ± ( n − 1 ) Pr k t ± ≃ D , where the tensor product is taken along the functor ( n − 1 ) Pr R t ± → ( n − 1 ) Pr k t ± (n-1)\operatorname{Pr}^{t\pm}_R \to(n-1)\operatorname{Pr}^{t\pm}_k ( n − 1 ) Pr R t ± → ( n − 1 ) Pr k t ± induced by R → k R \to k R → k . We prove the following (see Theorem 7.1, Theorem 7.3 for details):
Theorem 1.2. Suppose that D \mathcal{D} D has Hochschild cohomology H H ∙ ( D / k ) \mathcal{HH}^{\bullet}(\mathcal{D}/k) HH ∙ ( D / k ) (see Section 6). It is then naturally an E n + 1 \mathbb{E}_{n+1} E n + 1 -algebra (cf. the higher Deligne conjecture). Let Def D ( R ) \operatorname{Def}_{\mathcal{D}}(R) Def D ( R ) be the space of deformations of D \mathcal{D} D to R R R . The functor Def D \operatorname{Def}_{\mathcal{D}} Def D from the ∞ \infty ∞ -category of Artin (dg) E n + 1 \mathbb{E}_{n+1} E n + 1 -algebras to that of spaces is
a formal E n + 1 \mathrm{E}_{n+1} E n + 1 -moduli problem determined by Hochschild cohomology H H ∙ ( D / k ) \mathcal{HH}^{\bullet}(\mathcal{D}/k) HH ∙ ( D / k ) (see Theorem 7.3 for the formulation).
For any Artin E n + 1 \mathrm{E}_{n+1} E n + 1 -algebra R R R , there exists a canonical equivalence of spaces
Deform D ( R ) ≃ Map Alg n + 1 ( Mod k ) ( D n + 1 ( R ) , H H ∙ ( D / k ) ) \operatorname{Deform}_{\mathcal{D}}(R) \simeq\operatorname{Map}_{\operatorname{Alg}_{n+1}(\operatorname{Mod}_{k})}\left(\mathbb{D}_{n+1}(R),\mathcal{HH}^{\bullet}(\mathcal{D}/k)\right) Deform D ( R ) ≃ Map Alg n + 1 ( Mod k ) ( D n + 1 ( R ) , HH ∙ ( D / k ) )
where the right-hand side is the mapping space of E n + 1 \mathrm{E}_{n+1} E n + 1 -algebras.
In characteristic zero, if we restrict the deformation problem to commutative Artin (dg) algebras, it becomes a formal moduli problem determined by the underlying dg Lie algebra H H ∙ ( D / k ) [ n ] \mathcal{HH}^{\bullet}(\mathcal{D}/k)[n] HH ∙ ( D / k ) [ n ] . We also deduce the obstruction theory in this context (see Theorem 7.4 ):
Theorem 1.3. Let i ≥ 1 i \ge1 i ≥ 1 . Let p : R ′ → R p:R' \to R p : R ′ → R be an elementary extension of R R R by an ( i − 1 ) (i-1) ( i − 1 ) -shifted k k k -vector space V [ i − 1 ] V[i-1] V [ i − 1 ] (see § 7.2 ). Let ( D R , ϕ ) (\mathcal{D}_{R},\phi) ( D R , ϕ ) be a deformation of D \mathcal{D} D to R R R . Let H H i + n + 1 ( D / k ) \mathcal{HH}^{i+n+1}(\mathcal{D}/k) HH i + n + 1 ( D / k ) denote H i + n + 1 ( H H ∙ ( D / k ) ) H^{i+n+1}(\mathcal{HH}^{\bullet}(\mathcal{D}/k)) H i + n + 1 ( HH ∙ ( D / k )) . There exists a canonically defined element Obs ( D R , p ) ∈ H H i + n + 1 ( D / k ) ⊗ k V \operatorname{Obs}(\mathcal{D}_{R},p)\in\mathcal{HH}^{i+n+1}(\mathcal{D}/k)\otimes_{k}V Obs ( D R , p ) ∈ HH i + n + 1 ( D / k ) ⊗ k V such that Obs ( D R , p ) \operatorname{Obs}(\mathcal{D}_{R},p) Obs ( D R , p ) vanishes if and only if there exists D R ′ ∈ π 0 ( Deform D ( R ′ ) ) \mathcal{D}_{R'}\in\pi_{0}(\operatorname{Deform}_{\mathcal{D}}(R')) D R ′ ∈ π 0 ( Deform D ( R ′ )) lying over D R \mathcal{D}_{R} D R along the map p p p .
Hochschild cohomology. The Hochschild cohomology H H ∙ ( D / k ) \mathcal{HH}^{\bullet}(\mathcal{D}/k) HH ∙ ( D / k ) appearing in Theorems 1.2 and 1.3 is defined from the t t t -structured presentable ( ∞ , n ) (\infty,n) ( ∞ , n ) -category D \mathcal{D} D , and it is defined as a higher version of Hochschild cohomology. It depends a priori on the t t t -structure on D \mathcal{D} D . We also note that, as the first assumption of Theorem 1.2 suggests, the existence of the Hochschild cohomology of a presentable ( ∞ , n ) (\infty,n) ( ∞ , n ) -category is a nontrivial problem even without t t t -structures. We prove that the Hochschild cohomology of t t t -structured k k k -linear presentable ( ∞ , n ) (\infty,n) ( ∞ , n ) -categories exists in several important cases. Moreover, in these cases it depends only on the underlying (non-t t t -structured) k k k -linear presentable ( ∞ , n ) (\infty,n) ( ∞ , n ) -category. We discuss Hochschild cohomology in detail in § 6 .
Application to sheaves of categories. We give an application of our deformation theory to sheaves of categories on stacks and their higher generalization. For a scheme X \mathcal{X} X over k k k (or, more generally, for an algebraic stack satisfying suitable conditions), the notion of a sheaf of categories on X \mathcal{X} X is obtained by categorifying the notion of a quasi-coherent sheaf. Informally, a sheaf of categories M \mathcal{M} M assigns an R R R -linear presentable ∞ \infty ∞ -category Γ ( U , M ) \Gamma(U,\mathcal{M}) Γ ( U , M ) to each U = Spec R → X U=\operatorname{Spec}R\to\mathcal{X} U = Spec R → X (see [10 ] , [31 ] ). Sheaves of categories on X \mathcal{X} X constitute an ( ∞ , 2 ) (\infty,2) ( ∞ , 2 ) -category, and by 1-affineness of X \mathcal{X} X [10 ] (under a suitable condition) this ( ∞ , 2 ) (\infty,2) ( ∞ , 2 ) -category is given by the module category Mod QC ( X ) ( Pr k ) \operatorname{Mod}_{\operatorname{QC}(\mathcal{X})}(\operatorname{Pr}_{k}) Mod QC ( X ) ( Pr k ) . Here QC ( X ) \operatorname{QC}(\mathcal{X}) QC ( X ) denotes the monoidal ∞ \infty ∞ -category of quasi-coherent sheaves (complexes) on X \mathcal{X} X . It has a canonical enhancement to a complete t t t -structured presentable ( ∞ , 2 ) (\infty,2) ( ∞ , 2 ) -category Mod QC ( X ) ( Pr k t ± ) \operatorname{Mod}_{\operatorname{QC}(\mathcal{X})}(\operatorname{Pr}_{k}^{t\pm}) Mod QC ( X ) ( Pr k t ± ) (see § 6.3 ). This can be generalized to the context of sheaves of ( ∞ , n ) (\infty,n) ( ∞ , n ) -categories, and we define a complete t t t -structured presentable ( ∞ , n + 1 ) (\infty,n+1) ( ∞ , n + 1 ) -category n Mod QC ( X ) [ Pr k t ± ] n\operatorname{Mod}_{\operatorname{QC}(\mathcal{X})}[\operatorname{Pr}_{k}^{t\pm}] n Mod QC ( X ) [ Pr k t ± ] accordingly (see § 6.7 ). This is a higher categorical generalization of Mod QC ( X ) ( Pr k t ± ) \operatorname{Mod}_{\operatorname{QC}(\mathcal{X})}(\operatorname{Pr}_{k}^{t\pm}) Mod QC ( X ) ( Pr k t ± ) in the sense that 1 Mod QC ( X ) [ Pr k t ± ] = Mod QC ( X ) ( Pr k t ± ) 1\operatorname{Mod}_{\operatorname{QC}(\mathcal{X})}[\operatorname{Pr}_{k}^{t\pm}]=\operatorname{Mod}_{\operatorname{QC}(\mathcal{X})}(\operatorname{Pr}_{k}^{t\pm}) 1 Mod QC ( X ) [ Pr k t ± ] = Mod QC ( X ) ( Pr k t ± ) . We prove the following (see Proposition 6.10 , Proposition 6.19 , Theorem 7.5 for details), which provides a deformation-theoretic interpretation of the higher Hochschild cohomology of QC ( X ) \operatorname{QC}(\mathcal{X}) QC ( X ) . Note that n Mod QC ( X ) [ Pr k t ± ] n\operatorname{Mod}_{\operatorname{QC}(\mathcal{X})}[\operatorname{Pr}_{k}^{t\pm}] n Mod QC ( X ) [ Pr k t ± ] is an ( ∞ , n + 1 ) (\infty,n+1) ( ∞ , n + 1 ) -category, so Theorems 1.2 and 1.3 apply with n n n replaced by n + 1 n+1 n + 1 :
Corollary 1.4. Let X \mathcal{X} X be a perfect stack over k k k . Consider the deformation functor Deform n Mod QC ( X ) [ Pr k t ± ] \operatorname{Deform}_{n\operatorname{Mod}_{\operatorname{QC}(\mathcal{X})}[\operatorname{Pr}_{k}^{t\pm}]} Deform n Mod QC ( X ) [ Pr k t ± ] . Then Deform n Mod QC ( X ) [ Pr k t ± ] \operatorname{Deform}_{n\operatorname{Mod}_{\operatorname{QC}(\mathcal{X})}[\operatorname{Pr}_{k}^{t\pm}]} Deform n Mod QC ( X ) [ Pr k t ± ] is an E n + 2 \mathrm{E}_{n+2} E n + 2 -formal moduli problem determined by Hochschild cohomology H H ∙ ( n Mod QC ( X ) [ Pr k t ± ] / k ) \mathcal{HH}^{\bullet}(n\operatorname{Mod}_{\operatorname{QC}(\mathcal{X})}[\operatorname{Pr}_{k}^{t\pm}]/k) HH ∙ ( n Mod QC ( X ) [ Pr k t ± ] / k ) . Moreover, we have
H H ∙ ( n Mod QC ( X ) [ Pr k t ± ] / k ) \mathcal{HH}^{\bullet}(n\operatorname{Mod}_{\operatorname{QC}(\mathcal{X})}[\operatorname{Pr}_{k}^{t\pm}]/k) HH ∙ ( n Mod QC ( X ) [ Pr k t ± ] / k ) is equivalent to the Hom complex Hom QC ( Map ( S n , X ) ) ( O Δ , O Δ ) \operatorname{Hom}_{\operatorname{QC}(\operatorname{Map}(S^{n},\mathcal{X}))}(\mathcal{O}_{\Delta},\mathcal{O}_{\Delta}) Hom QC ( Map ( S n , X )) ( O Δ , O Δ ) in Mod k \operatorname{Mod}_{k} Mod k . Here Map ( S n , X ) \operatorname{Map}(S^{n},\mathcal{X}) Map ( S n , X ) is the derived free loop space, that is, the mapping stack from the sphere S n S^{n} S n to X \mathcal{X} X in the framework of derived geometry, and O Δ \mathcal{O}_{\Delta} O Δ is the pushforward of O X \mathcal{O}_{\mathcal{X}} O X along the morphism X → Map ( S n , X ) \mathcal{X}\to\operatorname{Map}(S^{n},\mathcal{X}) X → Map ( S n , X ) given by constant loops.
We further assume that X \mathcal{X} X is a scheme separated of finite type over k k k of characteristic zero. Then there is a Koszul duality type isomorphism
H H ∙ ( n Mod QC ( X ) [ Pr k t ± ] / k ) ≃ Γ ( X , Sym O X ∙ ( T X / k [ − n − 1 ] ) ) \mathcal{HH}^{\bullet}(n\operatorname{Mod}_{\operatorname{QC}(\mathcal{X})}[\operatorname{Pr}_{k}^{t\pm}]/k)\simeq\Gamma\left(\mathcal{X},\operatorname{Sym}^{\bullet}_{\mathcal{O}_{\mathcal{X}}}\left(\mathbb{T}_{\mathcal{X}/k}[-n-1]\right)\right) HH ∙ ( n Mod QC ( X ) [ Pr k t ± ] / k ) ≃ Γ ( X , Sym O X ∙ ( T X / k [ − n − 1 ] ) ) where the right-hand side denotes the derived global sections.
The isomorphism classes of deformations of n Mod QC ( X ) [ Pr k t ± ] n\operatorname{Mod}_{\operatorname{QC}(\mathcal{X})}[\operatorname{Pr}_{k}^{t\pm}] n Mod QC ( X ) [ Pr k t ± ] to the trivial square zero extension k ⊕ k [ j ] k\oplus k[j] k ⊕ k [ j ] ( j ≥ 0 ) (j\ge0) ( j ≥ 0 ) (i.e. π 0 ( Deform n Mod QC ( X ) [ Pr k t ± ] ( k ⊕ k [ j ] ) ) \pi_{0}(\operatorname{Deform}_{n\operatorname{Mod}_{\operatorname{QC}(\mathcal{X})}[\operatorname{Pr}_{k}^{t\pm}]}(k\oplus k[j])) π 0 ( Deform n Mod QC ( X ) [ Pr k t ± ] ( k ⊕ k [ j ])) ) is isomorphic to
H j + n + 2 ( Hom QC ( Map ( S n , X ) ) ( O Δ , O Δ ) ) . H^{j+n+2}\left(\operatorname{Hom}_{\operatorname{QC}(\operatorname{Map}(S^{n},\mathcal{X}))}(\mathcal{O}_{\Delta},\mathcal{O}_{\Delta})\right). H j + n + 2 ( Hom QC ( Map ( S n , X )) ( O Δ , O Δ ) ) .
if p : R ′ → R is an elementary extension of R by a shifted k -vector space V [ i − 1 ] and ( n Mod QC ( X ) [ Pr k t ± ] R , ϕ ) is a deformation of n Mod QC ( X ) [ Pr k t ± ] R to R , then there exists an obstruction class Obs ( n Mod QC ( X ) [ Pr k t ± ] R , p ) in H i + n + 2 ( Hom QC ( Map ( S n , X ) ) ( O Δ , O Δ ) ) ⊗ k V for a deformation to R ′ . (1) \text{if }p:R'\to R\text{ is an elementary extension of }R\text{ by a shifted }k\text{-vector space }V[i-1]\text{ and }(^{n}\operatorname{Mod}_{\operatorname{QC}(\mathcal{X})}[\operatorname{Pr}^{\mathrm{t}\pm}_{k}]_{R},\phi)\text{ is a deformation of }{}^{n}\operatorname{Mod}_{\operatorname{QC}(\mathcal{X})}[\operatorname{Pr}^{\mathrm{t}\pm}_{k}]_{R}\text{ to }R,\text{ then there exists an obstruction class }\operatorname{Obs}({}^{n}\operatorname{Mod}_{\operatorname{QC}(\mathcal{X})}[\operatorname{Pr}^{\mathrm{t}\pm}_{k}]_{R},p)\text{ in }H^{i+n+2}(\operatorname{Hom}_{\operatorname{QC}(\operatorname{Map}(S^{n},\mathcal{X}))}(\mathcal{O}_{\Delta},\mathcal{O}_{\Delta}))\otimes_{k}V\text{ for a deformation to }R'. \tag*{(1)} if p : R ′ → R is an elementary extension of R by a shifted k -vector space V [ i − 1 ] and ( n Mod QC ( X ) [ Pr k t ± ] R , ϕ ) is a deformation of n Mod QC ( X ) [ Pr k t ± ] R to R , then there exists an obstruction class Obs ( n Mod QC ( X ) [ Pr k t ± ] R , p ) in H i + n + 2 ( Hom QC ( Map ( S n , X )) ( O Δ , O Δ )) ⊗ k V for a deformation to R ′ . (1)
We conclude the introduction with some directions for future research. First, our framework applies to many t t t -structured presentable ( ∞ , n ) (\infty,n) ( ∞ , n ) -categories beyond the examples of this paper (i.e., sheaves of categories). For example, one can consider the category of (a suitable class of) k k k -linear ∞ \infty ∞ -categories with t t t -structures whose Hom ( C 1 , C 2 ) \operatorname{Hom}(C_1,C_2) Hom ( C 1 , C 2 ) from C 1 C_1 C 1 to C 2 C_2 C 2 is the ∞ \infty ∞ -category of right t t t -exact k k k -linear functors, which can be defined as k k k -linear ∞ \infty ∞ -categories with a t t t -structure. After an appropriate procedure (right completions and left completions of t t t -structures on Hom \operatorname{Hom} Hom ), one obtain an Pr k t ± \operatorname{Pr}^{\mathrm{t}\pm}_{k} Pr k t ± -enriched category, which gives rise to a t t t -structured k k k -linear presentable ( ∞ , 2 ) (\infty,2) ( ∞ , 2 ) -category. One level higher, we can also consider the Morita category whose objects are (a suitable class of) k k k -linear monoidal ∞ \infty ∞ -categories with t t t -structures, and whose mapping object from C 1 ⊗ C_1^{\otimes} C 1 ⊗ to C 2 ⊗ C_2^{\otimes} C 2 ⊗ is the ∞ \infty ∞ -category of C 1 ⊗ C_1^{\otimes} C 1 ⊗ -C 2 ⊗ C_2^{\otimes} C 2 ⊗ -bimodule objects in Pr k t ± \operatorname{Pr}^{\mathrm{t}\pm}_{k} Pr k t ± . Such a Morita category should give rise to a complete t t t -structured k k k -linear presentable ( ∞ , 3 ) (\infty,3) ( ∞ , 3 ) -category, provided that the construction of Morita categories of algebras is carried out in the enriched (k k k -linear) setting. We expect that such a construction will become available in the near future, and that it will provide a large supply of examples to which our deformation theory applies. Second, in Corollary 1.4, the Hochschild cohomology H H ∙ ( n Mod QC ( X ) [ Pr k t ± ] / k ) HH^\bullet({}^{n}\operatorname{Mod}_{\operatorname{QC}(\mathcal{X})}[\operatorname{Pr}^{\mathrm{t}\pm}_{k}]/k) H H ∙ ( n Mod QC ( X ) [ Pr k t ± ] / k ) is naturally an E n + 2 E_{n+2} E n + 2 -algebra, and is equivalent, as a k k k -module spectrum, to H : = Hom QC ( Map ( S n , X ) ) ( O Δ , O Δ ) H:=\operatorname{Hom}_{\operatorname{QC}(\operatorname{Map}(S^{n},\mathcal{X}))}(\mathcal{O}_{\Delta},\mathcal{O}_{\Delta}) H := Hom QC ( Map ( S n , X )) ( O Δ , O Δ ) . On the other hand, Toën [32 ] constructed an E n + 1 E_{n+1} E n + 1 -monoidal structure on QC ( Map ( S n , X ) ) \operatorname{QC}(\operatorname{Map}(S^{n},\mathcal{X})) QC ( Map ( S n , X )) by a cobordism-theoretic construction using cospans, which endows H H H with another E n + 2 E_{n+2} E n + 2 -algebra structure. We expect that these two structures coincide (see Remark 6.20). This would relate the deformation theory of categorified sheaves to topological field theory and to deformation quantization of shifted Poisson structures. Third, it would also be interesting to compare the deformation theory of an E n + 1 E_{n+1} E n + 1 -algebra A A A (see Francis [8 ] ) with that of the t t t -structured ( ∞ , n + 1 ) (\infty,n+1) ( ∞ , n + 1 ) -category n Pr A t ± {}^{n}\operatorname{Pr}^{\mathrm{t}\pm}_{A} n Pr A t ± : a deformation of A A A induces that of n Pr A t ± {}^{n}\operatorname{Pr}^{\mathrm{t}\pm}_{A} n Pr A t ± .
This paper is organized as follows. In § 2 we fix some notation and terminology and review notions and some results. In § 3 we introduce t t t -structured stable presentable ∞ \infty ∞ -categories after reviewing generalities on t t t -structures and presentable ( ∞ , n ) (\infty,n) ( ∞ , n ) -categories. In § 4 and § 5 , we prove the categorified Koszul duality equivalence, that is, Theorem 1.1 . In § 6 , we study Hochschild cohomology of t t t -structured presentable ( ∞ , n ) (\infty,n) ( ∞ , n ) -categories. § 7 discusses the deformation theory. We prove Theorem 1.2 and Theorem 1.3 .
Acknowledgements. The author is supported by JSPS KAKENHI grant.
Preliminaries
Let k k k denote a fixed field, which serves as the base field.
Notation
Notation and Terminology. We use the language of ( ∞ , 1 ) (\infty,1) ( ∞ , 1 ) -categories. We use the theory of quasi-categories as a model of ( ∞ , 1 ) (\infty,1) ( ∞ , 1 ) -categories. Our main references are [23 ] and [24 ] . Following [23 ] , we shall refer to quasi-categories as ∞ \infty ∞ -categories. To an ordinary category, we can assign an ∞ \infty ∞ -category by taking its nerve, and therefore when we treat ordinary categories, we often omit the nerve N ( − ) \mathrm{N}(-) N ( − ) and directly regard them as ∞ \infty ∞ -categories. We use presentable higher categories, i.e., presentable ( ∞ , n ) (\infty,n) ( ∞ , n ) -categories in the sense of Stefani [30 ] , which we briefly review in § 3.2 . This is formulated within the theory of ( ∞ , 1 ) (\infty,1) ( ∞ , 1 ) -categories, so the reader is not required to have a background of models of ( ∞ , n ) (\infty,n) ( ∞ , n ) -categories for n > 1 n>1 n > 1 .
Here is a list of some of the notation and conventions we frequently use:
≃ \simeq ≃ : categorical equivalence or homotopy equivalence. The symbol ≃ \simeq ≃ is used to indicate isomorphisms and equivalences in categories and ∞ \infty ∞ -categories.
C o p C^{\mathrm{op}} C op : the opposite ∞ \infty ∞ -category of an ∞ \infty ∞ -category.
C ≃ C^{\simeq} C ≃ : the largest Kan subcomplex of an ∞ \infty ∞ -category of C C C .
Map C ( C , C ′ ) \operatorname{Map}_{C}(C,C') Map C ( C , C ′ ) : the mapping space from an object C ∈ C C\in C C ∈ C to C ′ ∈ C C'\in C C ′ ∈ C where C C C is an ∞ \infty ∞ -category.
Fun ( A , B ) \operatorname{Fun}(A,B) Fun ( A , B ) : the function complex for simplicial sets A A A and B B B . If A A A and B B B are ∞ \infty ∞ -categories, we regard Fun ( A , B ) \operatorname{Fun}(A,B) Fun ( A , B ) as the functor category.
C a t ∞ \mathrm{Cat}_{\infty} Cat ∞ : the ∞ \infty ∞ -category of small ∞ \infty ∞ -categories. The word “small” means that it belongs to a fixed universe U \mathcal{U} U . We write C a t ^ ∞ \widehat{\mathrm{Cat}}_{\infty} Cat ∞ for the ∞ \infty ∞ -category of ∞ \infty ∞ -categories in an enlarged universe V \mathcal{V} V (U ∈ V \mathcal{U} \in\mathcal{V} U ∈ V ). We write C a t ^ † \widehat{\mathrm{Cat}}_{\dagger} Cat † for the (nonfull) subcategory of C a t ^ ∞ \widehat{\mathrm{Cat}}_{\infty} Cat ∞ which consists of ∞ \infty ∞ -categories having small colimits, and whose 1-morphisms are spanned by those functors which preserve small colimits.
S \mathcal{S} S : the ∞ \infty ∞ -category of U \mathcal{U} U -small spaces. We write S ^ \widehat{\mathcal{S}} S for the V \mathcal{V} V -small spaces.
Ind ( C ) \operatorname{Ind}(\mathcal{C}) Ind ( C ) : the Ind-category of C \mathcal{C} C .
We will use the theory of higher algebras. We refer the reader to [24 ] . Here is a list of (some) of the notation about ∞ \infty ∞ -operads and algebras over them that we will use:
Alg n ( M ) \operatorname{Alg}_{n}(\mathcal{M}) Alg n ( M ) : For a symmetric monoidal ∞ \infty ∞ -category M ⊗ \mathcal{M}^{\otimes} M ⊗ , we write Alg n ( M ) \operatorname{Alg}_{n}(\mathcal{M}) Alg n ( M ) for the ∞ \infty ∞ -category of algebra objects over the ∞ \infty ∞ -operad of little n n n -cubes E n ⊗ \mathbb{E}_{n}^{\otimes} E n ⊗ in M ⊗ \mathcal{M}^{\otimes} M ⊗ (cf. [24 5.1] ). We refer to an object of Alg n ( M ) \operatorname{Alg}_{n}(\mathcal{M}) Alg n ( M ) as an E n \mathbb{E}_{n} E n -algebra in M \mathcal{M} M . If we denote by A s ⊗ \mathbb{A}s^{\otimes} A s ⊗ the associative operad ([24 4.1.1] ), there is the standard equivalence A s ⊗ ≃ E 1 ⊗ \mathbb{A}s^{\otimes} \simeq\mathbb{E}_{1}^{\otimes} A s ⊗ ≃ E 1 ⊗ of ∞ \infty ∞ -operads. We usually identify Alg 1 ( M ) \operatorname{Alg}_{1}(\mathcal{M}) Alg 1 ( M ) with the ∞ \infty ∞ -category Alg A s ( M ) \operatorname{Alg}_{\mathbb{A}s}(\mathcal{M}) Alg A s ( M ) , that is, the ∞ \infty ∞ -category of associative algebras in M \mathcal{M} M .
Alg n + ( M ) = Alg n ( M ) / 1 \operatorname{Alg}_{n}^{+}(\mathcal{M}) = \operatorname{Alg}_{n}(\mathcal{M})_{/1} Alg n + ( M ) = Alg n ( M ) /1 : the ∞ \infty ∞ -category of augmented algebras. Here 1 1 1 is a unit algebra.
CAlg ( M ) \operatorname{CAlg}(\mathcal{M}) CAlg ( M ) : the ∞ \infty ∞ -category of commutative algebra objects in a symmetric monoidal ∞ \infty ∞ -category M ⊗ \mathcal{M}^{\otimes} M ⊗ . Instead of Alg E ∞ ( M ) \operatorname{Alg}_{\mathbb{E}_{\infty}}(\mathcal{M}) Alg E ∞ ( M ) or Alg c o m m ( M ⊗ ) \operatorname{Alg}_{\mathrm{comm}}(\mathcal{M}^{\otimes}) Alg comm ( M ⊗ ) we write CAlg ( M ) \operatorname{CAlg}(\mathcal{M}) CAlg ( M ) for the ∞ \infty ∞ -category of algebra objects over the E ∞ \mathbb{E}_{\infty} E ∞ -operad or the commutative operad in M ⊗ \mathcal{M}^{\otimes} M ⊗ .
Mod R ( M ) \operatorname{Mod}_{R}(\mathcal{M}) Mod R ( M ) : the ∞ \infty ∞ -category of R R R -module objects, where M ⊗ \mathcal{M}^{\otimes} M ⊗ is a symmetric monoidal ∞ \infty ∞ -category and R ∈ CAlg ( M ) R \in\operatorname{CAlg}(\mathcal{M}) R ∈ CAlg ( M ) .
Mod R \operatorname{Mod}_{R} Mod R : When M ⊗ \mathcal{M}^{\otimes} M ⊗ is the symmetric monoidal ∞ \infty ∞ -category S p ⊗ \mathrm{Sp}^{\otimes} Sp ⊗ of spectra, we write Mod R ⊗ \operatorname{Mod}_{R}^{\otimes} Mod R ⊗ for Mod R ⊗ ( S p ) \operatorname{Mod}_{R}^{\otimes}(\mathrm{Sp}) Mod R ⊗ ( Sp ) . We denote by Mod R \operatorname{Mod}_{R} Mod R the underlying category, that is, Mod R ( S p ) \operatorname{Mod}_{R}(\mathrm{Sp}) Mod R ( Sp ) . When R R R is the Eilenberg-Maclane ring spectrum H A HA H A , we write Mod A \operatorname{Mod}_{A} Mod A for Mod H A \operatorname{Mod}_{HA} Mod H A . For example, Mod k \operatorname{Mod}_{k} Mod k is equivalent to the derived ∞ \infty ∞ -category of k k k -modules.
CAlg k \operatorname{CAlg}_{k} CAlg k : CAlg k = CAlg ( Mod k ) \operatorname{CAlg}_{k} = \operatorname{CAlg}(\operatorname{Mod}_{k}) CAlg k = CAlg ( Mod k ) .
LMod ( M ) \operatorname{LMod}(\mathcal{M}) LMod ( M ) : the ∞ \infty ∞ -category Alg L M ( M ) \operatorname{Alg}_{\mathbb{L}\mathbb{M}}(\mathcal{M}) Alg LM ( M ) of L M ⊗ \mathbb{L}\mathbb{M}^{\otimes} L M ⊗ -algebras over L M ⊗ \mathbb{L}\mathbb{M}^{\otimes} L M ⊗ : the ∞ \infty ∞ -operad defined in [24 4.2.1.7] . We may consider an object of LMod ( M ) \operatorname{LMod}(\mathcal{M}) LMod ( M ) to be a pair of ( A , M ) (A,M) ( A , M ) such that A A A is an associative algebra object in M \mathcal{M} M and a left A A A -module M M M . There is a Cartesian fibration LMod ( M ) → Alg A s ( M ) ≃ Alg 1 ( M ) \operatorname{LMod}(\mathcal{M}) \to\operatorname{Alg}_{\mathbb{A}s}(\mathcal{M}) \simeq\operatorname{Alg}_{1}(\mathcal{M}) LMod ( M ) → Alg A s ( M ) ≃ Alg 1 ( M ) which sends ( A , M ) (A,M) ( A , M ) to A A A . Another canonical morphism is the forgetful functor LMod ( M ) → M \operatorname{LMod}(\mathcal{M}) \to\mathcal{M} LMod ( M ) → M which is informally given by ( A , M ) ↦ M (A,M) \mapsto M ( A , M ) ↦ M .
LMod A ( M ) \operatorname{LMod}_{A}(\mathcal{M}) LMod A ( M ) , LMod A \operatorname{LMod}_{A} LMod A : the ∞ \infty ∞ -category of left A A A -module objects in M \mathcal{M} M . For A ∈ Alg 1 ( M ) A \in\operatorname{Alg}_{1}(\mathcal{M}) A ∈ Alg 1 ( M ) , we define LMod A ( M ) \operatorname{LMod}_{A}(\mathcal{M}) LMod A ( M ) to be the fiber of LMod ( M ) → Alg 1 ( M ) \operatorname{LMod}(\mathcal{M}) \to\operatorname{Alg}_{1}(\mathcal{M}) LMod ( M ) → Alg 1 ( M ) over A A A in C a t ∞ \mathrm{Cat}_{\infty} Cat ∞ . If M \mathcal{M} M is the ∞ \infty ∞ -category of spectra (or Mod k \operatorname{Mod}_{k} Mod k ), we write LMod A \operatorname{LMod}_{A} LMod A for LMod A ( M ) \operatorname{LMod}_{A}(\mathcal{M}) LMod A ( M ) .
RMod ( M ) \operatorname{RMod}(\mathcal{M}) RMod ( M ) : the right module version of LMod ( M ) \operatorname{LMod}(\mathcal{M}) LMod ( M ) .
RMod B ( M ) \operatorname{RMod}_{B}(\mathcal{M}) RMod B ( M ) , RMod A \operatorname{RMod}_{A} RMod A : the ∞ \infty ∞ -category of right B B B -module objects in M \mathcal{M} M . RMod A \operatorname{RMod}_{A} RMod A is the right module version of LMod A \operatorname{LMod}_{A} LMod A .
P r \mathrm{Pr} Pr : ∞ \infty ∞ -category of presentable ∞ \infty ∞ -categories such that the mapping space consists of functors which preserve small colimits.
P r S t \mathrm{Pr}_{\mathrm{St}} Pr St , P r k \mathrm{Pr}_{k} Pr k : P r S t \mathrm{Pr}_{\mathrm{St}} Pr St is the ∞ \infty ∞ -category of presentable stable ∞ \infty ∞ -categories. A morphism C → D \mathcal{C} \to\mathcal{D} C → D in P r S t \mathrm{Pr}_{\mathrm{St}} Pr St is a functor that preserves small colimits. It has a closed symmetric monoidal structure whose unit is the stable ∞ \infty ∞ -category S p \mathrm{Sp} Sp of spectra (see [23 Definition 5.5.3.1] , [24 4.8] ). We write P r k \mathrm{Pr}_{k} Pr k for Mod Mod k ( P r S t ) \operatorname{Mod}_{\operatorname{Mod}_{k}}(\mathrm{Pr}_{\mathrm{St}}) Mod Mod k ( Pr St ) and write ⊗ k \otimes_{k} ⊗ k for the tensor products.
Hom objects
Let M \mathcal{M} M be an ∞ \infty ∞ -category. Let A ⊗ \mathcal{A}^{\otimes} A ⊗ be a monoidal ∞ \infty ∞ -category. Suppose that M \mathcal{M} M is a left A ⊗ \mathcal{A}^{\otimes} A ⊗ -module, that is, it has a left A ⊗ \mathcal{A}^{\otimes} A ⊗ -module action in C a t ^ ∞ \widehat{\mathrm{Cat}}_{\infty} Cat ∞ . To be precise, we let C a t ^ ∞ × Γ L M ⊗ → L M ⊗ \widehat{\mathrm{Cat}}_{\infty} \times_{\Gamma} \mathbb{L}\mathbb{M}^{\otimes} \to\mathbb{L}\mathbb{M}^{\otimes} Cat ∞ × Γ L M ⊗ → L M ⊗ be the base change of the Cartesian symmetric monoidal ∞ \infty ∞ -category C a t ^ ∞ ⊗ → Γ \widehat{\mathrm{Cat}}_{\infty}^{\otimes} \to\Gamma Cat ∞ ⊗ → Γ , and we consider the ∞ \infty ∞ -category LMod ( C a t ^ ∞ ) \operatorname{LMod}(\widehat{\mathrm{Cat}}_{\infty}) LMod ( Cat ∞ ) of L M ⊗ \mathbb{L}\mathbb{M}^{\otimes} L M ⊗ -algebra objects (see [24 4.2.1] ). This ∞ \infty ∞ -category LMod ( C a t ^ ∞ ) \operatorname{LMod}(\widehat{\mathrm{Cat}}_{\infty}) LMod ( Cat ∞ ) classifies pairs ( A ⊗ , M ) (\mathcal{A}^{\otimes},\mathcal{M}) ( A ⊗ , M ) and admits a functor LMod ( C a t ^ ∞ ) → Alg 1 ( C a t ^ ∞ ) \operatorname{LMod}(\widehat{\mathrm{Cat}}_{\infty}) \to\operatorname{Alg}_{1}(\widehat{\mathrm{Cat}}_{\infty}) LMod ( Cat ∞ ) → Alg 1 ( Cat ∞ ) which carries ( A ⊗ , M ) (\mathcal{A}^{\otimes},\mathcal{M}) ( A ⊗ , M ) to A ⊗ \mathcal{A}^{\otimes} A ⊗ . We define LMod A ⊗ ( C a t ^ ∞ ) \operatorname{LMod}_{\mathcal{A}^{\otimes}}(\widehat{\mathrm{Cat}}_{\infty}) LMod A ⊗ ( Cat ∞ ) to be LMod ( C a t ^ ∞ ) × Alg 1 ( C a t ^ ∞ ) { A ⊗ } \operatorname{LMod}(\widehat{\mathrm{Cat}}_{\infty}) \times_{\operatorname{Alg}_{1}(\widehat{\mathrm{Cat}}_{\infty})} \{\mathcal{A}^{\otimes}\} LMod ( Cat ∞ ) × Alg 1 ( Cat ∞ ) { A ⊗ } and refer to it as the ∞ \infty ∞ -category of left A ⊗ \mathcal{A}^{\otimes} A ⊗ -modules. In this context, the Hom object from M \mathcal{M} M to M ′ \mathcal{M}' M ′ (defined) in A \mathcal{A} A is an object H o m ( M , M ′ ) \mathcal{H}om(\mathcal{M},\mathcal{M}') H o m ( M , M ′ ) of A \mathcal{A} A together with a morphism f : H o m ( M , M ′ ) ⊗ M → M ′ f:\mathcal{H}om(\mathcal{M},\mathcal{M}') \otimes\mathcal{M} \to\mathcal{M}' f : H o m ( M , M ′ ) ⊗ M → M ′ , which has the universal property described as follows: For any E ∈ A E \in\mathcal{A} E ∈ A , the composition with f f f induces an equivalence of mapping spaces
Map A ( E , H o m ( M , M ′ ) ) → Map M ( E ⊗ M , M ′ ) \operatorname{Map}_{\mathcal{A}}(E,\mathcal{H}om(\mathcal{M},\mathcal{M}')) \to\operatorname{Map}_{\mathcal{M}}(E \otimes\mathcal{M},\mathcal{M}') Map A ( E , H o m ( M , M ′ )) → Map M ( E ⊗ M , M ′ )
(in [24 ] , Hom objects are called morphism objects).
The endomorphism algebra object of M ∈ M M \in\mathcal{M} M ∈ M (defined) in A \mathcal{A} A is an associative algebra End ( M ) \operatorname{End}(M) End ( M ) in A \mathcal{A} A (i.e., an object of Alg 1 ( A ) \operatorname{Alg}_{1}(\mathcal{A}) Alg 1 ( A ) ) together with a left module action of End ( M ) \operatorname{End}(M) End ( M ) on M M M , which has the universal property described informally as follows: For any associative algebra R R R in A \mathcal{A} A and a left module action R ⊗ M → M R \otimes M \to M R ⊗ M → M on M M M , there exists an essentially unique morphism f : R → End ( M ) f:R \to\operatorname{End}(M) f : R → End ( M ) such that the left module action of R R R is induced by the restriction of the action of End ( M ) \operatorname{End}(M) End ( M ) along f f f . To be precise, End ( M ) \operatorname{End}(M) End ( M ) is defined to be an object of Alg 1 ( A ⊗ ) \operatorname{Alg}_{1}(\mathcal{A}^{\otimes}) Alg 1 ( A ⊗ ) which represents the functor Alg 1 ( A ⊗ ) op → S \operatorname{Alg}_{1}(\mathcal{A}^{\otimes})^{\operatorname{op}} \to\mathcal{S} Alg 1 ( A ⊗ ) op → S corresponding to the right fibration
LMod ( M ) × M { M } → Alg 1 ( A ⊗ ) . \operatorname{LMod}(\mathcal{M}) \times_{\mathcal{M}} \{M\} \to\operatorname{Alg}_{1}(\mathcal{A}^{\otimes}). LMod ( M ) × M { M } → Alg 1 ( A ⊗ ) .
Here LMod ( M ) → M \operatorname{LMod}(\mathcal{M}) \to\mathcal{M} LMod ( M ) → M is the forgetful functor.
The more specialized case is that M ⊗ \mathcal{M}^{\otimes} M ⊗ is a monoidal ∞ \infty ∞ -category and the underlying ∞ \infty ∞ -category M \mathcal{M} M is regarded as a left M ⊗ \mathcal{M}^{\otimes} M ⊗ -module via the tensor product functor. In this case, Hom objects are referred to as internal Hom objects.
Koszul duality of algebras
Let M ⊗ \mathcal{M}^{\otimes} M ⊗ be a symmetric monoidal ∞ \infty ∞ -category which has small colimits and limits such that the tensor product preserves sifted colimits. We write 1 1 1 for a unit object. We review the Koszul duality of algebras in M ⊗ \mathcal{M}^{\otimes} M ⊗ in the formalism of pairings (see [24 ] , [25 X] for more details). Let Alg 1 + ( M ) = Alg 1 ( M ) / 1 \operatorname{Alg}_{1}^{+}(\mathcal{M}) = \operatorname{Alg}_{1}(\mathcal{M})/1 Alg 1 + ( M ) = Alg 1 ( M ) /1 be the ∞ \infty ∞ -category of augmented E 1 \mathbf{E}_{1} E 1 -algebra objects (associative algebra objects) in M ⊗ \mathcal{M}^{\otimes} M ⊗ . For a pair ( ϵ A : A → 1 , ϵ B : B → 1 ) ∈ Alg 1 + ( M ) × Alg 1 + ( M ) (\epsilon_A:A \to1,\epsilon_B:B \to1) \in\operatorname{Alg}_{1}^{+}(\mathcal{M}) \times\operatorname{Alg}_{1}^{+}(\mathcal{M}) ( ϵ A : A → 1 , ϵ B : B → 1 ) ∈ Alg 1 + ( M ) × Alg 1 + ( M ) we put
Pair ( ϵ A , ϵ B ) : = Map Alg 1 ( M ) ( A ⊗ B , 1 ) × Map ( A , 1 ) × Map ( B , 1 ) { ( ϵ A , ϵ B ) } . \operatorname{Pair}(\epsilon_A,\epsilon_B) := \operatorname{Map}_{\operatorname{Alg}_{1}(\mathcal{M})}(A \otimes B,1) \times_{\operatorname{Map}(A,1)\times\operatorname{Map}(B,1)} \{(\epsilon_A,\epsilon_B)\}. Pair ( ϵ A , ϵ B ) := Map Alg 1 ( M ) ( A ⊗ B , 1 ) × Map ( A , 1 ) × Map ( B , 1 ) {( ϵ A , ϵ B )} .
Fix ϵ B : B → 1 \epsilon_B:B \to1 ϵ B : B → 1 and consider the functor u ϵ B : Alg 1 + ( M ) op → S u_{\epsilon_B}:\operatorname{Alg}_{1}^{+}(\mathcal{M})^{\operatorname{op}} \to\mathcal{S} u ϵ B : Alg 1 + ( M ) op → S given informally by [ ϵ A : A → 1 ] ↦ Pair ( ϵ A , ϵ B ) [\epsilon_A:A \to1] \mapsto\operatorname{Pair}(\epsilon_A,\epsilon_B) [ ϵ A : A → 1 ] ↦ Pair ( ϵ A , ϵ B ) . The Koszul dual D M ⊗ ( B ) ∈ Alg 1 + ( M ) \mathbb{D}_{\mathcal{M}^{\otimes}}(B) \in\operatorname{Alg}_{1}^{+}(\mathcal{M}) D M ⊗ ( B ) ∈ Alg 1 + ( M ) of B B B (we abuse notation since D M ⊗ ( B ) \mathbb{D}_{\mathcal{M}^{\otimes}}(B) D M ⊗ ( B ) depends on the augmentation ϵ B : B → 1 \epsilon_B:B \to1 ϵ B : B → 1 ) is defined by the following universal property that D M ⊗ ( B ) \mathbb{D}_{\mathcal{M}^{\otimes}}(B) D M ⊗ ( B ) represents the functor u ϵ B u_{\epsilon_B} u ϵ B . Under a mild condition on M ⊗ \mathcal{M}^{\otimes} M ⊗ , there exist D M ⊗ ( B ) \mathbb{D}_{\mathcal{M}^{\otimes}}(B) D M ⊗ ( B ) and functorial equivalences
Map Alg 1 + ( M ) ( A , D M ⊗ ( B ) ) ≃ Pair ( ϵ A , ϵ B ) \operatorname{Map}_{\operatorname{Alg}_{1}^{+}(\mathcal{M})}(A,\mathbb{D}_{\mathcal{M}^{\otimes}}(B)) \simeq\operatorname{Pair}(\epsilon_A,\epsilon_B) Map Alg 1 + ( M ) ( A , D M ⊗ ( B )) ≃ Pair ( ϵ A , ϵ B )
for ϵ A : A → 1 \epsilon_A:A \to1 ϵ A : A → 1 . We shall refer to the pairing B ⊗ D M ⊗ ( B ) → 1 B \otimes\mathbb{D}_{\mathcal{M}^{\otimes}}(B) \to1 B ⊗ D M ⊗ ( B ) → 1 corresponding to the identity of D M ⊗ ( B ) \mathbb{D}_{\mathcal{M}^{\otimes}}(B) D M ⊗ ( B ) as the universal pairing for B → k B \to k B → k . In that case, there is a functor
D M ⊗ : Alg 1 ( M ) ⟶ Alg 1 ( M ) op \mathbb{D}_{\mathcal{M}^{\otimes}}:\operatorname{Alg}_{1}(\mathcal{M}) \longrightarrow\operatorname{Alg}_{1}(\mathcal{M})^{\operatorname{op}} D M ⊗ : Alg 1 ( M ) ⟶ Alg 1 ( M ) op
which carries A → 1 A \to1 A → 1 to D M ⊗ ( A ) → 1 \mathbb{D}_{\mathcal{M}^{\otimes}}(A) \to1 D M ⊗ ( A ) → 1 .
The Koszul duals can be described in terms of bar constructions and endomorphism algebras. Let Bar : Alg 1 + ( M ) → M \operatorname{Bar}:\operatorname{Alg}_{1}^{+}(\mathcal{M}) \to\mathcal{M} Bar : Alg 1 + ( M ) → M be the functor given by [ B → 1 ] ↦ 1 ⊗ B 1 [B \to1] \mapsto1 \otimes_B 1 [ B → 1 ] ↦ 1 ⊗ B 1 , which we refer to as the bar construction of augmented algebras. Then 1 ⊗ B 1 1 \otimes_B 1 1 ⊗ B 1 admits a structure of a coaugmented coalgebra in a suitable way, and Bar \operatorname{Bar} Bar is promoted to Bar : Alg 1 + ( M ) → Alg 1 + ( ( M o p ) op ) \operatorname{Bar}:\operatorname{Alg}_{1}^{+}(\mathcal{M}) \to\operatorname{Alg}_{1}^{+}((\mathcal{M}^{\mathrm{op}})^{\operatorname{op}}) Bar : Alg 1 + ( M ) → Alg 1 + (( M op ) op ) (notice that the target is the ∞ \infty ∞ -category of coaugmented coalgebras). According to [25 X, Example 4.4.20] , for B ∈ Alg 1 + ( M ) B \in\operatorname{Alg}_{1}^{+}(\mathcal{M}) B ∈ Alg 1 + ( M ) , the Koszul dual D M ⊗ ( B ) \mathbb{D}_{\mathcal{M}^{\otimes}}(B) D M ⊗ ( B ) is equivalent to the dual Bar ( B ) ∨ \operatorname{Bar}(B)^{\vee} Bar ( B ) ∨ in M \mathcal{M} M , that is, the endomorphism algebra End LMod B ( M ) ( 1 ) ∈ Alg 1 ( M ) \operatorname{End}_{\operatorname{LMod}_{B}(\mathcal{M})}(1) \in\operatorname{Alg}_{1}(\mathcal{M}) End LMod B ( M ) ( 1 ) ∈ Alg 1 ( M ) .
We recall the Koszul duality between E n \mathbf{E}_{n} E n -algebras. Set M ⊗ = Mod k \mathcal{M}^{\otimes}=\operatorname{Mod}_{k} M ⊗ = Mod k : in this case Koszul duals exist (cf. [25 X, 3.1.5, 4.3.8, 4.4.8] ). Let n ≥ 1 n \ge1 n ≥ 1 be a natural number and let Alg n ( Mod k ) \operatorname{Alg}_{n}(\operatorname{Mod}_{k}) Alg n ( Mod k ) be the ∞ \infty ∞ -category of E n \mathbf{E}_{n} E n -algebras in Mod k \operatorname{Mod}_{k} Mod k . We write Alg n + ( Mod k ) = Alg n ( Mod k ) / k \operatorname{Alg}_{n}^{+}(\operatorname{Mod}_{k}) = \operatorname{Alg}_{n}(\operatorname{Mod}_{k})/k Alg n + ( Mod k ) = Alg n ( Mod k ) / k for the ∞ \infty ∞ -category Alg n ( Mod k ) / k \operatorname{Alg}_{n}(\operatorname{Mod}_{k})/k Alg n ( Mod k ) / k of augmented E n \mathbf{E}_{n} E n -algebras. We consider the Koszul duals of augmented E n \mathbf{E}_{n} E n -algebras. We take M ⊗ \mathcal{M}^{\otimes} M ⊗ to be Alg n − 1 ⊗ ( Mod k ) \operatorname{Alg}_{n-1}^{\otimes}(\operatorname{Mod}_{k}) Alg n − 1 ⊗ ( Mod k ) . There exists the Koszul duality functor
D n : Alg n + ( Mod k ) ≃ Alg 1 + ( Alg n − 1 ( Mod k ) ) ⟶ Alg 1 + ( Alg n − 1 ( Mod k ) ) op ≃ Alg n + ( Mod k ) op \mathbb{D}_{n}:\operatorname{Alg}_{n}^{+}(\operatorname{Mod}_{k}) \simeq\operatorname{Alg}_{1}^{+}(\operatorname{Alg}_{n-1}(\operatorname{Mod}_{k})) \longrightarrow\operatorname{Alg}_{1}^{+}(\operatorname{Alg}_{n-1}(\operatorname{Mod}_{k}))^{\operatorname{op}} \simeq\operatorname{Alg}_{n}^{+}(\operatorname{Mod}_{k})^{\operatorname{op}} D n : Alg n + ( Mod k ) ≃ Alg 1 + ( Alg n − 1 ( Mod k )) ⟶ Alg 1 + ( Alg n − 1 ( Mod k ) ) op ≃ Alg n + ( Mod k ) op
(see e.g. [25 X, 4.4.8] ). Let ϵ B : B → k \epsilon_B:B \to k ϵ B : B → k be an augmented E n \mathbf{E}_{n} E n -algebra. As mentioned above, D n ( B ) \mathbb{D}_{n}(B) D n ( B ) is determined by the following universal property. For an augmented E n \mathbf{E}_{n} E n -algebra ϵ C : C → k \epsilon_C:C \to k ϵ C : C → k , we write
Pair ( ϵ B , ϵ C ) = Map Alg n ( Mod k ) ( B ⊗ k C , k ) × Map ( B , k ) × Map ( C , k ) { ( ϵ B , ϵ C ) } . \operatorname{Pair}(\epsilon_B,\epsilon_C) = \operatorname{Map}_{\operatorname{Alg}_{n}(\operatorname{Mod}_{k})}(B \otimes_k C,k) \times_{\operatorname{Map}(B,k)\times\operatorname{Map}(C,k)} \{(\epsilon_B,\epsilon_C)\}. Pair ( ϵ B , ϵ C ) = Map Alg n ( Mod k ) ( B ⊗ k C , k ) × Map ( B , k ) × Map ( C , k ) {( ϵ B , ϵ C )} .
Then the functor Alg n + ( Mod k ) op → S \operatorname{Alg}_{n}^{+}(\operatorname{Mod}_{k})^{\operatorname{op}} \to\mathcal{S} Alg n + ( Mod k ) op → S given by [ ϵ C : C → k ] ↦ Pair ( ϵ B , ϵ C ) [\epsilon_C:C \to k] \mapsto\operatorname{Pair}(\epsilon_B,\epsilon_C) [ ϵ C : C → k ] ↦ Pair ( ϵ B , ϵ C ) is representable by D n ( B ) \mathbb{D}_{n}(B) D n ( B ) . There is a universal (Koszul dual) pairing B ⊗ k D n ( B ) → k B \otimes_k \mathbb{D}_{n}(B) \to k B ⊗ k D n ( B ) → k corresponding to the identity map id ∈ Map Alg n + ( Mod k ) ( D n ( B ) , D n ( B ) ) \operatorname{id} \in\operatorname{Map}_{\operatorname{Alg}_{n}^{+}(\operatorname{Mod}_{k})}(\mathbb{D}_{n}(B),\mathbb{D}_{n}(B)) id ∈ Map Alg n + ( Mod k ) ( D n ( B ) , D n ( B )) . We refer to D n ( B ) \mathbb{D}_{n}(B) D n ( B ) as the E n \mathbf{E}_{n} E n -Koszul dual of B B B .
t t t -Structured Presentable Higher Categories
Generalities on t t t -structures
Unless otherwise stated, we adopt the cohomological index. Let us recall the notion of t t t -structures on stable ∞ \infty ∞ -categories. Let C \mathcal{C} C be a stable ∞ \infty ∞ -category. Recall that the homotopy category h ( C ) h(\mathcal{C}) h ( C ) has the structure of a triangulated category. A t t t -structure on C \mathcal{C} C is a pair of full subcategories ( C ≤ 0 , C ≥ 0 ) (\mathcal{C}^{\leq0},\mathcal{C}^{\geq0}) ( C ≤ 0 , C ≥ 0 ) of C \mathcal{C} C such that ( h ( C ≤ 0 ) , h ( C ≥ 0 ) ) (h(\mathcal{C}^{\leq0}),h(\mathcal{C}^{\geq0})) ( h ( C ≤ 0 ) , h ( C ≥ 0 )) is a t t t -structure on the triangulated category h ( C ) h(\mathcal{C}) h ( C ) (see e.g. [24 ] , [20 Section 2.1] ). We shall write C ≤ n \mathcal{C}^{\leq n} C ≤ n and C ≥ n \mathcal{C}^{\geq n} C ≥ n for C ≤ 0 [ − n ] \mathcal{C}^{\leq0}[-n] C ≤ 0 [ − n ] and C ≥ 0 [ − n ] \mathcal{C}^{\geq0}[-n] C ≥ 0 [ − n ] , respectively.
Suppose that ( C ≤ 0 , C ≥ 0 ) (\mathcal{C}^{\leq0},\mathcal{C}^{\geq0}) ( C ≤ 0 , C ≥ 0 ) is a t t t -structure. The inclusion C ≤ n ↪ C \mathcal{C}^{\leq n}\hookrightarrow\mathcal{C} C ≤ n ↪ C has a right adjoint τ ≤ n : C ⟶ C ≤ n \tau^{\leq n}:\mathcal{C}\longrightarrow\mathcal{C}^{\leq n} τ ≤ n : C ⟶ C ≤ n and the inclusion C ≥ n ↪ C \mathcal{C}^{\geq n}\hookrightarrow\mathcal{C} C ≥ n ↪ C has a left adjoint τ ≥ n : C ⟶ C ≥ n \tau^{\geq n}:\mathcal{C}\longrightarrow\mathcal{C}^{\geq n} τ ≥ n : C ⟶ C ≥ n . There exists a canonical equivalence τ ≥ 0 ∘ τ ≤ 0 ≃ τ ≤ 0 ∘ τ ≥ 0 \tau^{\geq0}\circ\tau^{\leq0}\simeq\tau^{\leq0}\circ\tau^{\geq0} τ ≥ 0 ∘ τ ≤ 0 ≃ τ ≤ 0 ∘ τ ≥ 0 . We often use the notation ( C ≤ 0 , C ≥ 0 ) (\mathcal{C}^{\leq0},\mathcal{C}^{\geq0}) ( C ≤ 0 , C ≥ 0 ) to denote a t t t -structure on C \mathcal{C} C . Note that C ≥ 0 \mathcal{C}^{\geq0} C ≥ 0 is uniquely determined by C ≤ 0 \mathcal{C}^{\leq0} C ≤ 0 : the full subcategory C ≥ 0 \mathcal{C}^{\geq0} C ≥ 0 is spanned by those objects C ′ C' C ′ such that Map C ( C , C ′ ) \operatorname{Map}_{\mathcal{C}}(C,C') Map C ( C , C ′ ) is contractible for any C ∈ C ≤ − 1 C\in\mathcal{C}^{\leq-1} C ∈ C ≤ − 1 . Dually, C ≤ 0 \mathcal{C}^{\leq0} C ≤ 0 is uniquely determined by C ≥ 0 \mathcal{C}^{\geq0} C ≥ 0 . By ( C , C ≤ 0 ) (\mathcal{C},\mathcal{C}^{\leq0}) ( C , C ≤ 0 ) we mean a stable ∞ \infty ∞ -category endowed with the t t t -structure determined by a full subcategory C ≤ 0 \mathcal{C}^{\leq0} C ≤ 0 .
Definition 3.1. Let ( C ≤ 0 , C ≥ 0 ) (\mathcal{C}^{\leq0},\mathcal{C}^{\geq0}) ( C ≤ 0 , C ≥ 0 ) be a t t t -structure on a stable ∞ \infty ∞ -category C \mathcal{C} C .
For any n ∈ Z n\in\mathbb{Z} n ∈ Z , the left adjoint functor τ ≥ n : C → C ≥ n \tau^{\geq n}:\mathcal{C}\to\mathcal{C}^{\geq n} τ ≥ n : C → C ≥ n to the inclusion C ≥ n ⊂ C \mathcal{C}^{\geq n}\subset\mathcal{C} C ≥ n ⊂ C induces a morphism C → τ ≥ n ( C ) \mathcal{C}\to\tau^{\geq n}(C) C → τ ≥ n ( C ) , determined by the unit map. For all n ∈ Z n\in\mathbb{Z} n ∈ Z , this yields the sequence
C ⟶ ⋯ ⟶ τ ≥ n − 1 ( C ) ⟶ τ ≥ n ( C ) ⟶ τ ≥ n + 1 ( C ) ⟶ ⋯ C\longrightarrow\cdots\longrightarrow\tau^{\geq n-1}(C)\longrightarrow\tau^{\geq n}(C)\longrightarrow\tau^{\geq n+1}(C)\longrightarrow\cdots C ⟶ ⋯ ⟶ τ ≥ n − 1 ( C ) ⟶ τ ≥ n ( C ) ⟶ τ ≥ n + 1 ( C ) ⟶ ⋯
in C \mathcal{C} C . The t t t -structure is left complete if, for any C ∈ C C\in\mathcal{C} C ∈ C , the induced morphism C → lim ← n ∈ Z τ ≥ n ( C ) C\to\lim\limits_{\leftarrow n\in\mathbb{Z}}\tau^{\geq n}(C) C → ← n ∈ Z lim τ ≥ n ( C ) is an equivalence.
The right adjoint functor τ ≤ n : C → C ≤ n \tau^{\leq n}:\mathcal{C}\to\mathcal{C}^{\leq n} τ ≤ n : C → C ≤ n to the inclusion C ≤ n ⊂ C \mathcal{C}^{\leq n}\subset\mathcal{C} C ≤ n ⊂ C induces a morphism τ ≤ n ( C ) → C \tau^{\leq n}(C)\to C τ ≤ n ( C ) → C , determined by the counit map. For all n ∈ Z n\in\mathbb{Z} n ∈ Z , this yields the sequence
⋯ ⟶ τ ≤ n − 1 ( C ) ⟶ τ ≤ n ( C ) ⟶ τ ≤ n + 1 ( C ) ⟶ ⋯ ⟶ C \cdots\longrightarrow\tau^{\leq n-1}(C)\longrightarrow\tau^{\leq n}(C)\longrightarrow\tau^{\leq n+1}(C)\longrightarrow\cdots\longrightarrow C ⋯ ⟶ τ ≤ n − 1 ( C ) ⟶ τ ≤ n ( C ) ⟶ τ ≤ n + 1 ( C ) ⟶ ⋯ ⟶ C
in C \mathcal{C} C . The t t t -structure is right complete if, for any C ∈ C C\in\mathcal{C} C ∈ C , the induced morphism lim ⟶ n ∈ Z τ ≤ n ( C ) → C \lim\limits_{\longrightarrow n\in\mathbb{Z}}\tau^{\leq n}(C)\to C ⟶ n ∈ Z lim τ ≤ n ( C ) → C is an equivalence.
Suppose that C \mathcal{C} C is presentable. The t t t -structure ( C ≤ 0 , C ≥ 0 ) (\mathcal{C}^{\leq0},\mathcal{C}^{\geq0}) ( C ≤ 0 , C ≥ 0 ) is accessible if the full subcategory C ≤ 0 \mathcal{C}^{\leq0} C ≤ 0 is presentable.
Suppose that C \mathcal{C} C has a monoidal structure. If the tensor product functor C × C → C \mathcal{C}\times\mathcal{C}\to\mathcal{C} C × C → C sends C ≤ 0 × C ≤ 0 \mathcal{C}^{\leq0}\times\mathcal{C}^{\leq0} C ≤ 0 × C ≤ 0 to C ≤ 0 \mathcal{C}^{\leq0} C ≤ 0 and the unit object lies in C ≤ 0 \mathcal{C}^{\leq0} C ≤ 0 , we say that the t t t -structure is compatible with the monoidal structure.
Example 3.2. Let A A A be an associative connective ring spectrum, and let LMod A \operatorname{LMod}_{A} LMod A be the stable ∞ \infty ∞ -category of left A A A -module spectra. Define LMod A ≤ 0 \operatorname{LMod}_{A}^{\leq0} LMod A ≤ 0 as the full subcategory of LMod A \operatorname{LMod}_{A} LMod A spanned by objects M M M such that π n ( M ) = 0 \pi_n(M)=0 π n ( M ) = 0 for all n < 0 n<0 n < 0 , and LMod A ≥ 0 \operatorname{LMod}_{A}^{\geq0} LMod A ≥ 0 as the full subcategory spanned by objects M M M such that π n ( M ) = 0 \pi_n(M)=0 π n ( M ) = 0 for all n > 0 n>0 n > 0 . Then ( LMod A ≤ 0 , LMod A ≥ 0 ) (\operatorname{LMod}_{A}^{\leq0},\operatorname{LMod}_{A}^{\geq0}) ( LMod A ≤ 0 , LMod A ≥ 0 ) defines an accessible, left complete, and right complete t t t -structure (see [24 7.1.1.13] ). The full subcategory LMod A ≤ 0 \operatorname{LMod}_{A}^{\leq0} LMod A ≤ 0 is also characterized as the smallest full subcategory which contains A A A and is closed under small colimits. Moreover, the t t t -structure is compatible with filtered colimits, as LMod A ≥ 0 ⊂ LMod A \operatorname{LMod}_{A}^{\geq0}\subset\operatorname{LMod}_{A} LMod A ≥ 0 ⊂ LMod A is closed under filtered colimits. This t t t -structure is compatible with the monoidal structure when A A A is an E 2 \mathbb{E}_2 E 2 -algebra. The heart LMod A ♡ \operatorname{LMod}_{A}^{\heartsuit} LMod A ♡ is equivalent to the abelian category of left π 0 ( A ) \pi_0(A) π 0 ( A ) -modules.
Remark 3.3. Example 3.2 is generalized to the nonconnective case as follows. For A ∈ Alg 1 ( Sp ) A\in\operatorname{Alg}_1(\operatorname{Sp}) A ∈ Alg 1 ( Sp ) we define LMod A ≤ 0 \operatorname{LMod}_{A}^{\leq0} LMod A ≤ 0 to be the smallest full subcategory which contains A A A and is closed under colimits and extensions. By [24 1.4.4.11] , LMod A ≤ 0 \operatorname{LMod}_{A}^{\leq0} LMod A ≤ 0 determines an accessible t t t -structure on LMod A \operatorname{LMod}_{A} LMod A such that LMod A ≥ 0 \operatorname{LMod}_{A}^{\geq0} LMod A ≥ 0 consists of those objects M M M such that π 0 ( Map LMod A ( A [ n ] , M ) ) = π n ( M ) = H − n ( M ) = 0 \pi_0(\operatorname{Map}_{\operatorname{LMod}_{A}}(A[n],M))=\pi_n(M)=H^{-n}(M)=0 π 0 ( Map LMod A ( A [ n ] , M )) = π n ( M ) = H − n ( M ) = 0 for n > 0 n>0 n > 0 . The full subcategory LMod A ≥ 0 \operatorname{LMod}_{A}^{\geq0} LMod A ≥ 0 is closed under filtered colimits, and the t t t -structure is right complete. We shall refer to this t t t -structure as the standard t t t -structure. When A A A is an E 2 \mathbb{E}_2 E 2 -algebra, the t t t -structure is compatible with the monoidal structure on LMod A \operatorname{LMod}_{A} LMod A . If A A A is not connective, the resulting t t t -structure is generally not left complete.
Example 3.4. Let C \mathcal{C} C be a small stable ∞ \infty ∞ -category. Let ( C ≤ 0 , C ≥ 0 ) (\mathcal{C}^{\leq0},\mathcal{C}^{\geq0}) ( C ≤ 0 , C ≥ 0 ) be a t t t -structure on C \mathcal{C} C . Then ( Ind ( C ≤ 0 ) , Ind ( C ≥ 0 ) ) (\operatorname{Ind}(\mathcal{C}^{\leq0}),\operatorname{Ind}(\mathcal{C}^{\geq0})) ( Ind ( C ≤ 0 ) , Ind ( C ≥ 0 )) defines a t t t -structure on Ind ( C ) \operatorname{Ind}(\mathcal{C}) Ind ( C ) . By the construction, the t t t -structure is compatible with filtered colimits. If the t t t -structure on C \mathcal{C} C is right bounded, then ( Ind ( C ≤ 0 ) , Ind ( C ≥ 0 ) ) (\operatorname{Ind}(\mathcal{C}^{\leq0}),\operatorname{Ind}(\mathcal{C}^{\geq0})) ( Ind ( C ≤ 0 ) , Ind ( C ≥ 0 )) is right complete (see [25 VIII, Lemma 5.4.1] ).
For instance, suppose that R R R is a noetherian connective E 1 \mathrm{E}_1 E 1 -algebra over a field k k k , that is, an object of Alg 1 ( Mod k ≤ 0 ) \operatorname{Alg}_1(\operatorname{Mod}^{\leq0}_k) Alg 1 ( Mod k ≤ 0 ) . We let LCoh ( R ) \operatorname{LCoh}(R) LCoh ( R ) denote the full subcategory of LMod R \operatorname{LMod}_R LMod R spanned by those objects M M M such that H n ( M ) = 0 H^n(M)=0 H n ( M ) = 0 for ∣ n ∣ ≫ 0 |n|\gg0 ∣ n ∣ ≫ 0 , and each H n ( M ) H^n(M) H n ( M ) is a finitely generated H 0 ( R ) H^0(R) H 0 ( R ) -module. This small stable ∞ \infty ∞ -category LCoh ( R ) \operatorname{LCoh}(R) LCoh ( R ) inherits a bounded t t t -structure from that of LMod R \operatorname{LMod}_R LMod R in Example 3.2. Then ( Ind ( LCoh ( R ) ≤ 0 ) , Ind ( LCoh ( R ) ≥ 0 ) ) (\operatorname{Ind}(\operatorname{LCoh}(R)^{\leq0}),\operatorname{Ind}(\operatorname{LCoh}(R)^{\geq0})) ( Ind ( LCoh ( R ) ≤ 0 ) , Ind ( LCoh ( R ) ≥ 0 )) is a right complete accessible t t t -structure. This is generally not left complete.
Let Pr t \operatorname{Pr}^{\mathrm{t}} Pr t be the ∞ \infty ∞ -category of presentable stable ∞ \infty ∞ -categories equipped with accessible t t t -structures. An object of Pr t \operatorname{Pr}^{\mathrm{t}} Pr t is a pair ( C , C ≤ 0 ) (\mathcal{C},\mathcal{C}^{\leq0}) ( C , C ≤ 0 ) where C \mathcal{C} C is a presentable stable ∞ \infty ∞ -category, and C ≤ 0 \mathcal{C}^{\leq0} C ≤ 0 is a full subcategory inducing a t t t -structure, such that C ≤ 0 \mathcal{C}^{\leq0} C ≤ 0 itself is presentable. A morphism ( C , C ≤ 0 ) → ( D , D ≤ 0 ) (\mathcal{C},\mathcal{C}^{\leq0})\to(\mathcal{D},\mathcal{D}^{\leq0}) ( C , C ≤ 0 ) → ( D , D ≤ 0 ) is a colimit-preserving functor F : C → D F:\mathcal{C}\to\mathcal{D} F : C → D which is right t t t -exact (i.e., F ( C ≤ 0 ) ⊂ D ≤ 0 F(\mathcal{C}^{\leq0})\subset\mathcal{D}^{\leq0} F ( C ≤ 0 ) ⊂ D ≤ 0 ). Let Pr t + \operatorname{Pr}^{\mathrm{t}+} Pr t + be the full subcategory of Pr t \operatorname{Pr}^{\mathrm{t}} Pr t , spanned by presentable stable ∞ \infty ∞ -categories equipped with accessible right complete t t t -structures. Let Pr t ± \operatorname{Pr}^{\mathrm{t}\pm} Pr t ± be the full subcategory of Pr t \operatorname{Pr}^{\mathrm{t}} Pr t , spanned by presentable stable ∞ \infty ∞ -categories equipped with accessible left complete and right complete t t t -structures. There exists an adjoint pair
I + : Pr t + ⇄ Pr t : R ^ I^+:\operatorname{Pr}^{\mathrm{t}+}\rightleftarrows\operatorname{Pr}^{\mathrm{t}}:\widehat{R} I + : Pr t + ⇄ Pr t : R
such that I + I^+ I + is the inclusion, and the right adjoint R ^ \widehat{R} R carries ( C , C ≤ 0 ) (\mathcal{C},\mathcal{C}^{\leq0}) ( C , C ≤ 0 ) to the right completion R ^ ( C , C ≤ 0 ) \widehat{R}(\mathcal{C},\mathcal{C}^{\leq0}) R ( C , C ≤ 0 ) , which is ( colim n → ∞ C ≤ n , C ≤ 0 ) (\operatorname*{colim}_{n\to\infty}\mathcal{C}^{\leq n},\mathcal{C}^{\leq0}) ( colim n → ∞ C ≤ n , C ≤ 0 ) where colim n → ∞ C ≤ n \operatorname*{colim}_{n\to\infty}\mathcal{C}^{\leq n} colim n → ∞ C ≤ n is the colimit of the sequence of inclusions
C ≤ 0 → C ≤ 1 → C ≤ 2 → ⋯ → C ≤ n → ⋯ \mathcal{C}^{\leq0}\to\mathcal{C}^{\leq1}\to\mathcal{C}^{\leq2}\to\cdots\to\mathcal{C}^{\leq n}\to\cdots C ≤ 0 → C ≤ 1 → C ≤ 2 → ⋯ → C ≤ n → ⋯
in Pr S t \operatorname{Pr}_{\mathrm{St}} Pr St (cf. [23 5.5.3.4, 5.5.3.18] ).
The left completion L ^ ( ( C , C ≤ 0 ) ) \widehat{L}((\mathcal{C},\mathcal{C}^{\leq0})) L (( C , C ≤ 0 )) of ( C , C ≤ 0 ) (\mathcal{C},\mathcal{C}^{\leq0}) ( C , C ≤ 0 ) is defined as lim n → − ∞ C ≥ n \lim_{n\to-\infty}\mathcal{C}^{\geq n} lim n → − ∞ C ≥ n in Pr S t \operatorname{Pr}_{\mathrm{St}} Pr St of the sequence
⋯ → C ≥ n − 1 → C ≥ n → C ≥ n + 1 → ⋯ \cdots\to\mathcal{C}^{\geq n-1}\to\mathcal{C}^{\geq n}\to\mathcal{C}^{\geq n+1}\to\cdots ⋯ → C ≥ n − 1 → C ≥ n → C ≥ n + 1 → ⋯
induced by truncations. We usually write L ^ ( C ) \widehat{L}(\mathcal{C}) L ( C ) for L ^ ( ( C , C ≤ 0 ) ) \widehat{L}((\mathcal{C},\mathcal{C}^{\leq0})) L (( C , C ≤ 0 )) . If ( C , C ≤ 0 ) ∈ Pr t + (\mathcal{C},\mathcal{C}^{\leq0})\in\operatorname{Pr}^{\mathrm{t}+} ( C , C ≤ 0 ) ∈ Pr t + , then the left completion L ^ ( C ) \widehat{L}(\mathcal{C}) L ( C ) is also accessible and right complete (see [25 VIII, 4.6.13] ). There exists a localization adjoint pair
L ^ : Pr t + ⇄ Pr t ± : I ± \widehat{L}:\operatorname{Pr}^{\mathrm{t}+}\rightleftarrows\operatorname{Pr}^{\mathrm{t}\pm}:I^\pm L : Pr t + ⇄ Pr t ± : I ±
such that I ± I^\pm I ± is the inclusion, and L ^ \widehat{L} L is the left adjoint given by the left completion.
Symmetric monoidal structures. Next, we recall the symmetric monoidal structure on Pr t \operatorname{Pr}^{\mathrm{t}} Pr t (see [25 VIII, 4.6.1] ). Let ( B , B ≤ 0 ) (\mathcal{B},\mathcal{B}^{\leq0}) ( B , B ≤ 0 ) and ( C , C ≤ 0 ) (\mathcal{C},\mathcal{C}^{\leq0}) ( C , C ≤ 0 ) be presentable ∞ \infty ∞ -categories endowed with accessible t t t -structures. The tensor product of ( B , B ≤ 0 ) (\mathcal{B},\mathcal{B}^{\leq0}) ( B , B ≤ 0 ) and ( C , C ≤ 0 ) (\mathcal{C},\mathcal{C}^{\leq0}) ( C , C ≤ 0 ) is defined as B ⊗ C \mathcal{B}\otimes\mathcal{C} B ⊗ C (in Pr S t \operatorname{Pr}_{\mathrm{St}} Pr St ) together with the smallest full subcategory ( B ⊗ C ) ≤ 0 (\mathcal{B}\otimes\mathcal{C})^{\leq0} ( B ⊗ C ) ≤ 0 which contains the essential image of B ≤ 0 × C ≤ 0 → B ⊗ C \mathcal{B}^{\leq0}\times\mathcal{C}^{\leq0}\to\mathcal{B}\otimes\mathcal{C} B ≤ 0 × C ≤ 0 → B ⊗ C and is closed under small colimits and extensions. By [24 1.4.5.11] , ( B ⊗ C ) ≤ 0 (\mathcal{B}\otimes\mathcal{C})^{\leq0} ( B ⊗ C ) ≤ 0 defines an accessible t t t -structure. The universal property of ( B ⊗ C , ( B ⊗ C ) ≤ 0 ) (\mathcal{B}\otimes\mathcal{C},(\mathcal{B}\otimes\mathcal{C})^{\leq0}) ( B ⊗ C , ( B ⊗ C ) ≤ 0 ) is as follows: For any ( D , D ≤ 0 ) ∈ Pr t (\mathcal{D},\mathcal{D}^{\leq0})\in\operatorname{Pr}^{\mathrm{t}} ( D , D ≤ 0 ) ∈ Pr t the composition with the canonical functor B × C → B ⊗ C \mathcal{B}\times\mathcal{C}\to\mathcal{B}\otimes\mathcal{C} B × C → B ⊗ C gives an equivalence
Fun Pr t ( ( B ⊗ C , ( B ⊗ C ) ≤ 0 ) , ( D , D ≤ 0 ) ) → Fun ≤ ( B × C , D ) \operatorname{Fun}_{\operatorname{Pr}^{\mathrm{t}}}((\mathcal{B}\otimes\mathcal{C},(\mathcal{B}\otimes\mathcal{C})^{\leq0}),(\mathcal{D},\mathcal{D}^{\leq0}))\to\operatorname{Fun}^{\leq}(\mathcal{B}\times\mathcal{C},\mathcal{D}) Fun Pr t (( B ⊗ C , ( B ⊗ C ) ≤ 0 ) , ( D , D ≤ 0 )) → Fun ≤ ( B × C , D )
where Fun ≤ ( B × C , D ) \operatorname{Fun}^{\leq}(\mathcal{B}\times\mathcal{C},\mathcal{D}) Fun ≤ ( B × C , D ) is the full subcategory of Fun ( B × C , D ) \operatorname{Fun}(\mathcal{B}\times\mathcal{C},\mathcal{D}) Fun ( B × C , D ) , which consists of those functors F F F that preserve small colimits separately in each variable, such that the essential image of B ≤ 0 × C ≤ 0 \mathcal{B}^{\leq0}\times\mathcal{C}^{\leq0} B ≤ 0 × C ≤ 0 is contained in D ≤ 0 \mathcal{D}^{\leq0} D ≤ 0 .
If ( B , B ≤ 0 ) (\mathcal{B},\mathcal{B}^{\leq0}) ( B , B ≤ 0 ) and ( C , C ≤ 0 ) (\mathcal{C},\mathcal{C}^{\leq0}) ( C , C ≤ 0 ) are right complete, then the tensor product ( B ⊗ C , ( B ⊗ C ) ≤ 0 ) (\mathcal{B}\otimes\mathcal{C},(\mathcal{B}\otimes\mathcal{C})^{\leq0}) ( B ⊗ C , ( B ⊗ C ) ≤ 0 ) in Pr t \operatorname{Pr}^{\mathrm{t}} Pr t is also right complete (see [25 VIII, 4.6.11] ). Thus, Pr t + \operatorname{Pr}^{\mathrm{t}+} Pr t + inherits a symmetric monoidal structure from that of Pr t \operatorname{Pr}^{\mathrm{t}} Pr t . The left adjoint I + I^+ I + is promoted to a symmetric monoidal functor. The ∞ \infty ∞ -category Pr t ± \operatorname{Pr}^{\mathrm{t}\pm} Pr t ± also has a symmetric monoidal structure. Let ( B , B ≤ 0 ) (\mathcal{B},\mathcal{B}^{\leq0}) ( B , B ≤ 0 ) and ( C , C ≤ 0 ) (\mathcal{C},\mathcal{C}^{\leq0}) ( C , C ≤ 0 ) be objects of Pr t ± \operatorname{Pr}^{\mathrm{t}\pm} Pr t ± . Explicitly, the tensor product of ( B , B ≤ 0 ) (\mathcal{B},\mathcal{B}^{\leq0}) ( B , B ≤ 0 ) and ( C , C ≤ 0 ) (\mathcal{C},\mathcal{C}^{\leq0}) ( C , C ≤ 0 ) in Pr t ± \operatorname{Pr}^{\mathrm{t}\pm} Pr t ± is given by
L ^ ( ( B ⊗ C , ( B ⊗ C ) ≤ 0 ) ) . \widehat{L}((\mathcal{B}\otimes\mathcal{C},(\mathcal{B}\otimes\mathcal{C})^{\leq0})). L (( B ⊗ C , ( B ⊗ C ) ≤ 0 )) .
We usually write L ^ ( B ⊗ C ) \widehat{L}(\mathcal{B}\otimes\mathcal{C}) L ( B ⊗ C ) or B ⊗ ^ C \mathcal{B}\widehat{\otimes}\mathcal{C} B ⊗ C for it. The functor L ^ : Pr t + → Pr t ± \widehat{L}:\operatorname{Pr}^{\mathrm{t}+}\to\operatorname{Pr}^{\mathrm{t}\pm} L : Pr t + → Pr t ± is promoted to a symmetric monoidal functor. Since L ^ \widehat{L} L is a left adjoint functor, L ^ \widehat{L} L commutes with the relative tensor products. The inclusion I ± I^\pm I ± is not a symmetric monoidal functor but a lax symmetric monoidal functor.
Remark 3.5. The symmetric monoidal Pr t \operatorname{Pr}^{\mathrm{t}} Pr t admits internal Hom objects: For P , Q ∈ Pr t \mathcal{P},\mathcal{Q}\in\operatorname{Pr}^{\mathrm{t}} P , Q ∈ Pr t , an internal Hom object Hom Pr t ( P , Q ) \operatorname{Hom}_{\operatorname{Pr}^{\mathrm{t}}}(\mathcal{P},\mathcal{Q}) Hom Pr t ( P , Q ) from P \mathcal{P} P to Q \mathcal{Q} Q in Pr t \operatorname{Pr}^{\mathrm{t}} Pr t is given by the pair ( Hom Pr ( P , Q ) , Hom Pr ( P , Q ) ≤ 0 ) (\operatorname{Hom}_{\operatorname{Pr}}(\mathcal{P},\mathcal{Q}),\operatorname{Hom}_{\operatorname{Pr}}(\mathcal{P},\mathcal{Q})^{\leq0}) ( Hom Pr ( P , Q ) , Hom Pr ( P , Q ) ≤ 0 ) such that Hom Pr ( P , Q ) ∈ Pr \operatorname{Hom}_{\operatorname{Pr}}(\mathcal{P},\mathcal{Q})\in\operatorname{Pr} Hom Pr ( P , Q ) ∈ Pr is the internal Hom object from the underlying object P \mathcal{P} P to Q \mathcal{Q} Q , and Hom Pr ( P , Q ) ≤ 0 \operatorname{Hom}_{\Pr}(\mathcal{P},\mathcal{Q})^{\leq0} Hom P r ( P , Q ) ≤ 0 is the full subcategory, spanned by right t t t -exact functors. Thus, Pr t + \Pr^{t+} Pr t + also admits internal Hom objects obtained by taking right completions of internal Hom objects in Pr t \Pr^t Pr t .
Notation 3.6. Let Pr k t + \Pr_k^{t+} Pr k t + denote Mod Mod k ( Pr t + ) \operatorname{Mod}_{\operatorname{Mod}_k}(\Pr^{t+}) Mod Mod k ( Pr t + ) where we regard Mod k \operatorname{Mod}_k Mod k as an object of CAlg ( Pr t + ) \operatorname{CAlg}(\Pr^{t+}) CAlg ( Pr t + ) by taking Mod k ≤ 0 \operatorname{Mod}_k^{\leq0} Mod k ≤ 0 to be the full subcategory spanned by connective k k k -module spectra. We write Pr k t ± \Pr_k^{t\pm} Pr k t ± for Mod Mod k ( Pr t ± ) \operatorname{Mod}_{\operatorname{Mod}_k}(\Pr^{t\pm}) Mod Mod k ( Pr t ± ) .
Presentable higher categories
Let Cat ^ † \widehat{\operatorname{Cat}}_{\dagger} Cat † denote the subcategory of the ∞ \infty ∞ -category of V \mathsf{V} V -small ∞ \infty ∞ -categories such that objects are spanned by those ∞ \infty ∞ -categories which admit U \mathsf{U} U -small colimits, and 1-morphisms are spanned by functors which preserve U \mathsf{U} U -small colimits. According to [24 4.8.4.2] , Cat ^ † \widehat{\operatorname{Cat}}_{\dagger} Cat † is a V \mathsf{V} V -presentable ∞ \infty ∞ -category.
For instance, an U \mathsf{U} U -presentable ∞ \infty ∞ -category belongs to Cat ^ † \widehat{\operatorname{Cat}}_{\dagger} Cat † . The ∞ \infty ∞ -category Pr \Pr Pr of presentable ∞ \infty ∞ -categories has U \mathsf{U} U -small colimits so that Pr \Pr Pr belongs to Cat ^ † \widehat{\operatorname{Cat}}_{\dagger} Cat † . In addition, by [30 Proposition 5.1.4] , it is U \mathsf{U} U -compactly generated ∞ \infty ∞ -category, and
Pr ⊂ Cat ^ † \Pr\subset\widehat{\operatorname{Cat}}_{\dagger} Pr ⊂ Cat †
is characterized as the full subcategory of κ 0 \kappa_0 κ 0 -compact objects where κ 0 \kappa_0 κ 0 is the cardinal of the set of U \mathsf{U} U -small sets. (κ 0 \kappa_0 κ 0 is the cardinality of U \mathsf{U} U )
The ∞ \infty ∞ -category Cat ^ † \widehat{\operatorname{Cat}}_{\dagger} Cat † has a symmetric monoidal structure (see [24 4.8.1.3] ). For D , D ′ ∈ Cat ^ † \mathcal{D},\mathcal{D}' \in\widehat{\operatorname{Cat}}_{\dagger} D , D ′ ∈ Cat † , there is D ⊗ D ′ ∈ Cat ^ † \mathcal{D} \otimes\mathcal{D}' \in\widehat{\operatorname{Cat}}_{\dagger} D ⊗ D ′ ∈ Cat † with a functor u : D × D ′ → D ⊗ D ′ u:\mathcal{D}\times\mathcal{D}'\to\mathcal{D}\otimes\mathcal{D}' u : D × D ′ → D ⊗ D ′ which preserves U \mathsf{U} U -small colimits separately in each variable and further satisfies the universal property that for any E ∈ Cat ^ † \mathcal{E} \in\widehat{\operatorname{Cat}}_{\dagger} E ∈ Cat † , the composition with u u u induces an equivalence Map Cat ^ † ( D ⊗ D ′ , E ) → Map Cat ^ † e ( D × D ′ , E ) \operatorname{Map}_{\widehat{\operatorname{Cat}}_{\dagger}}(\mathcal{D}\otimes\mathcal{D}',\mathcal{E}) \to\operatorname{Map}^{e}_{\widehat{\operatorname{Cat}}_{\dagger}}(\mathcal{D}\times\mathcal{D}',\mathcal{E}) Map Cat † ( D ⊗ D ′ , E ) → Map Cat † e ( D × D ′ , E ) of mapping spaces. Here Map Cat ^ † e ( D × D ′ , E ) \operatorname{Map}^{e}_{\widehat{\operatorname{Cat}}_{\dagger}}(\mathcal{D}\times\mathcal{D}',\mathcal{E}) Map Cat † e ( D × D ′ , E ) is the full subcategory of Map Cat ^ † ( D × D ′ , E ) \operatorname{Map}_{\widehat{\operatorname{Cat}}_{\dagger}}(\mathcal{D}\times\mathcal{D}',\mathcal{E}) Map Cat † ( D × D ′ , E ) , which consists of functors preserving small colimits separately in each variable. Moreover, Pr \Pr Pr is closed under tensor products, and the tensor product functor ⊗ : Pr × Pr → Pr \otimes:\Pr\times\Pr\to\Pr ⊗ : Pr × Pr → Pr preserves U \mathsf{U} U -small colimits separately in each variable. Thus, Pr \Pr Pr is a commutative algebra object of Cat ^ † \widehat{\operatorname{Cat}}_{\dagger} Cat † .
We consider Mod Pr ( Cat ^ † ) \operatorname{Mod}_{\Pr}(\widehat{\operatorname{Cat}}_{\dagger}) Mod P r ( Cat † ) and define 2 Pr 2\Pr 2 Pr to be the full subcategory of Mod Pr ( Cat ^ † ) \operatorname{Mod}_{\Pr}(\widehat{\operatorname{Cat}}_{\dagger}) Mod P r ( Cat † ) , which consists of κ 0 \kappa_0 κ 0 -compact objects. The ∞ \infty ∞ -category 2 Pr 2\Pr 2 Pr itself is closed under U \mathsf{U} U -small colimits and is generated by Pr ⊗ Fun ( Δ 1 , S ) \Pr\otimes\operatorname{Fun}(\Delta^1,\mathcal{S}) Pr ⊗ Fun ( Δ 1 , S ) under small colimits. In particular, 2 Pr 2\Pr 2 Pr is closed under taking tensor products in Mod Pr ( Cat ^ † ) \operatorname{Mod}_{\Pr}(\widehat{\operatorname{Cat}}_{\dagger}) Mod P r ( Cat † ) so that 2 Pr 2\Pr 2 Pr belongs to CAlg ( Cat ^ † ) \operatorname{CAlg}(\widehat{\operatorname{Cat}}_{\dagger}) CAlg ( Cat † ) .
More generally, this argument is extended to the following:
Construction 3.7. For any C ∈ CAlg ( Cat ^ † ) \mathcal{C}\in\operatorname{CAlg}(\widehat{\operatorname{Cat}}_{\dagger}) C ∈ CAlg ( Cat † ) the full subcategory Mod C ( Cat ^ † ) κ 0 \operatorname{Mod}_{\mathcal{C}}(\widehat{\operatorname{Cat}}_{\dagger})^{\kappa_0} Mod C ( Cat † ) κ 0 of Mod C ( Cat ^ † ) \operatorname{Mod}_{\mathcal{C}}(\widehat{\operatorname{Cat}}_{\dagger}) Mod C ( Cat † ) , which consists of κ 0 \kappa_0 κ 0 -compact objects, belongs to CAlg ( Cat ^ † ) \operatorname{CAlg}(\widehat{\operatorname{Cat}}_{\dagger}) CAlg ( Cat † ) . This assignment C ↦ Mod C ( Cat ^ † ) κ 0 \mathcal{C}\mapsto\operatorname{Mod}_{\mathcal{C}}(\widehat{\operatorname{Cat}}_{\dagger})^{\kappa_0} C ↦ Mod C ( Cat † ) κ 0 determines a functor
M : CAlg ( Cat ^ † ) ⟶ CAlg ( Cat ^ † ) \mathsf{M}:\operatorname{CAlg}(\widehat{\operatorname{Cat}}_{\dagger})\longrightarrow\operatorname{CAlg}(\widehat{\operatorname{Cat}}_{\dagger}) M : CAlg ( Cat † ) ⟶ CAlg ( Cat † )
such that any morphism f : C → C ′ f:\mathcal{C}\to\mathcal{C}' f : C → C ′ defines the base change functor ( − ) ⊗ C C ′ : Mod C ( Cat ^ † ) → Mod C ′ ( Cat ^ † ) (-)\otimes_{\mathcal{C}}\mathcal{C}':\operatorname{Mod}_{\mathcal{C}}(\widehat{\operatorname{Cat}}_{\dagger})\to\operatorname{Mod}_{\mathcal{C}'}(\widehat{\operatorname{Cat}}_{\dagger}) ( − ) ⊗ C C ′ : Mod C ( Cat † ) → Mod C ′ ( Cat † ) whose restriction induces M ( f ) : M ( C ) = Mod C ( Cat ^ † ) κ 0 → M ( C ′ ) = Mod C ′ ( Cat ^ † ) κ 0 \mathsf{M}(f):\mathsf{M}(\mathcal{C})=\operatorname{Mod}_{\mathcal{C}}(\widehat{\operatorname{Cat}}_{\dagger})^{\kappa_0}\to\mathsf{M}(\mathcal{C}')=\operatorname{Mod}_{\mathcal{C}'}(\widehat{\operatorname{Cat}}_{\dagger})^{\kappa_0} M ( f ) : M ( C ) = Mod C ( Cat † ) κ 0 → M ( C ′ ) = Mod C ′ ( Cat † ) κ 0 . For n ≥ 2 n\geq2 n ≥ 2 we define n C n\mathcal{C} n C to be M n − 1 ( C ) \mathsf{M}^{n-1}(\mathcal{C}) M n − 1 ( C ) . By convention, 1 C 1\mathcal{C} 1 C is C \mathcal{C} C .
Using this construction we define n Pr n\Pr n Pr to be the full subcategory Mod ( n − 1 ) Pr ( Cat ^ † ) κ 0 \operatorname{Mod}_{(n-1)\Pr}(\widehat{\operatorname{Cat}}_{\dagger})^{\kappa_0} Mod ( n − 1 ) P r ( Cat † ) κ 0 of Mod ( n − 1 ) Pr ( Cat ^ † ) \operatorname{Mod}_{(n-1)\Pr}(\widehat{\operatorname{Cat}}_{\dagger}) Mod ( n − 1 ) P r ( Cat † ) , which consists of κ 0 \kappa_0 κ 0 -compact objects. That is, n Pr = M n − 1 ( Pr ) n\Pr=\mathsf{M}^{n-1}(\Pr) n Pr = M n − 1 ( Pr ) . By convention, if n = 1 n=1 n = 1 , 1 Pr 1\Pr 1 Pr is Pr \Pr Pr . We refer to an object of n Pr n\Pr n Pr as a presentable ( ∞ , n ) (\infty,n) ( ∞ , n ) -category.
Let Pr s t \Pr_{\mathrm{st}} Pr st denote the ∞ \infty ∞ -category of stable presentable ∞ \infty ∞ -categories. We often identify it with Mod S ( Pr ) \operatorname{Mod}_{\mathbb{S}}(\Pr) Mod S ( Pr ) . For n ≥ 2 n\geq2 n ≥ 2 , we define n Pr s t n\Pr_{\mathrm{st}} n Pr st to be M n − 1 ( Pr s t ) \mathsf{M}^{n-1}(\Pr_{\mathrm{st}}) M n − 1 ( Pr st ) . We put 1 Pr s t = Pr s t 1\Pr_{\mathrm{st}}=\Pr_{\mathrm{st}} 1 Pr st = Pr st . We refer to an object of n Pr s t n\Pr_{\mathrm{st}} n Pr st as a stable presentable ( ∞ , n ) (\infty,n) ( ∞ , n ) -category. By convention, we set 0 Pr s t = S p 0\Pr_{\mathrm{st}}=\mathsf{Sp} 0 Pr st = Sp
t t t -structured presentable higher categories
Let Pr t + \Pr^{t+} Pr t + be the ∞ \infty ∞ -category of stable presentable ∞ \infty ∞ -categories equipped with right complete and accessible t t t -structures. A 1-morphism of Pr t + \Pr^{t+} Pr t + is a functor that preserves small colimits and is right t t t -exact.
Colimits and limits
Let
Sp ( C ) = C ⊗ Sp c o n n Sp ≃ colim Σ C ⇒ C ⇒ Σ C ⇒ ⋯ \operatorname{Sp}(\mathcal{C})=\mathcal{C}\otimes_{\operatorname{Sp}^{\mathrm{conn}}}\operatorname{Sp}\simeq\operatorname{colim}_{\Sigma\mathcal{C}\Rightarrow\mathcal{C}\Rightarrow\Sigma\mathcal{C}\Rightarrow\cdots} Sp ( C ) = C ⊗ Sp conn Sp ≃ colim Σ C ⇒ C ⇒ Σ C ⇒ ⋯
be the stabilization of C \mathcal{C} C , where Σ C \Sigma\mathcal{C} Σ C is the pushout ∗ ⊔ C ∗ \ast\sqcup_{\mathcal{C}}\ast ∗ ⊔ C ∗ for C ∈ Pr \mathcal{C}\in\Pr C ∈ Pr , and the right-hand side is a colimit in Pr s t \Pr_{\mathrm{st}} Pr st (equivalently, in Pr \Pr Pr ). Consider its full subcategory Sp ( C ) ≤ 0 \operatorname{Sp}(\mathcal{C})^{\leq0} Sp ( C ) ≤ 0 that is the smallest one, which contains the image of C \mathcal{C} C and is closed under extensions and small colimits. According to [24 1.4.3.4] , Sp ( C ) ≤ 0 \operatorname{Sp}(\mathcal{C})^{\leq0} Sp ( C ) ≤ 0 is presentable, and the pair ( Sp ( C ) , Sp ( C ) ≤ 0 ) (\operatorname{Sp}(\mathcal{C}),\operatorname{Sp}(\mathcal{C})^{\leq0}) ( Sp ( C ) , Sp ( C ) ≤ 0 ) constitutes an accessible t t t -structure. As proved below, Sp ( C ) ≤ 0 \operatorname{Sp}(\mathcal{C})^{\leq0} Sp ( C ) ≤ 0 is a right complete t t t -structure on Sp ( C ) \operatorname{Sp}(\mathcal{C}) Sp ( C ) .
Proposition 3.8. Let C : Pr t + → Mod S p c o n n ( Pr ) \mathcal{C}\colon\operatorname{Pr}^{\mathrm{t+}}\to\operatorname{Mod}_{\mathrm{Sp}^{\mathrm{conn}}}(\operatorname{Pr}) C : Pr t + → Mod Sp conn ( Pr ) be the functor given by ( D , D ≤ 0 ) ↦ D ≤ 0 (\mathcal{D},\mathcal{D}^{\leq0})\mapsto\mathcal{D}^{\leq0} ( D , D ≤ 0 ) ↦ D ≤ 0 . Then there is a left adjoint S : Mod S p c o n n ( Pr ) → Pr t + \mathcal{S}\colon\operatorname{Mod}_{\mathrm{Sp}^{\mathrm{conn}}}(\operatorname{Pr})\to\operatorname{Pr}^{\mathrm{t+}} S : Mod Sp conn ( Pr ) → Pr t + given by C ↦ ( Sp ( C ) , Sp ( C ) ≤ 0 ) \mathcal{C}\mapsto(\operatorname{Sp}(\mathcal{C}),\operatorname{Sp}(\mathcal{C})^{\leq0}) C ↦ ( Sp ( C ) , Sp ( C ) ≤ 0 ) , which constitutes an adjunction
S : Mod S p c o n n ( Pr ) ⇄ Pr t + : C . \mathcal{S}\colon\operatorname{Mod}_{\mathrm{Sp}^{\mathrm{conn}}}(\operatorname{Pr}) \rightleftarrows\operatorname{Pr}^{\mathrm{t+}}\colon\mathcal{C}. S : Mod Sp conn ( Pr ) ⇄ Pr t + : C . Moreover, C \mathcal{C} C is fully faithful.
Proof. Consider the inclusions C → Sp ( C ) ≤ 0 → Sp ( C ) \mathcal{C}\to\operatorname{Sp}(\mathcal{C})^{\leq0}\to\operatorname{Sp}(\mathcal{C}) C → Sp ( C ) ≤ 0 → Sp ( C ) . For ( D , D ≤ 0 ) ∈ Pr t + (\mathcal{D},\mathcal{D}^{\leq0})\in\operatorname{Pr}^{\mathrm{t+}} ( D , D ≤ 0 ) ∈ Pr t + , the composition induces
Map Pr t + ( ( Sp ( C ) , Sp ( C ) ≤ 0 ) , ( D , D ≤ 0 ) ) → f Map Pr ( Sp ( C ) ≤ 0 , D ≤ 0 ) → g Map Pr ( C , D ≤ 0 ) . \operatorname{Map}_{\operatorname{Pr}^{\mathrm{t+}}}((\operatorname{Sp}(\mathcal{C}),\operatorname{Sp}(\mathcal{C})^{\leq0}),(\mathcal{D},\mathcal{D}^{\leq0}))\xrightarrow{f}\operatorname{Map}_{\operatorname{Pr}}(\operatorname{Sp}(\mathcal{C})^{\leq0},\mathcal{D}^{\leq0})\xrightarrow{g}\operatorname{Map}_{\operatorname{Pr}}(\mathcal{C},\mathcal{D}^{\leq0}). Map Pr t + (( Sp ( C ) , Sp ( C ) ≤ 0 ) , ( D , D ≤ 0 )) f Map Pr ( Sp ( C ) ≤ 0 , D ≤ 0 ) g Map Pr ( C , D ≤ 0 ) .
Using the universal property of Sp ( C ) \operatorname{Sp}(\mathcal{C}) Sp ( C ) , that is an equivalence Map Pr S t ( Sp ( C ) , D ) ≃ Map Pr ( C , D ) \operatorname{Map}_{\operatorname{Pr}^{\mathrm{St}}}(\operatorname{Sp}(\mathcal{C}),\mathcal{D})\simeq\operatorname{Map}_{\operatorname{Pr}}(\mathcal{C},\mathcal{D}) Map Pr St ( Sp ( C ) , D ) ≃ Map Pr ( C , D ) induced by C → Sp ( C ) \mathcal{C}\to\operatorname{Sp}(\mathcal{C}) C → Sp ( C ) , we see that the composite g ∘ f g\circ f g ∘ f is an equivalence. Thus, to see that S \mathcal{S} S is a left adjoint to C \mathcal{C} C , it suffices to prove that ( Sp ( C ) , Sp ( C ) ≤ 0 ) (\operatorname{Sp}(\mathcal{C}),\operatorname{Sp}(\mathcal{C})^{\leq0}) ( Sp ( C ) , Sp ( C ) ≤ 0 ) belongs to Pr t + \operatorname{Pr}^{\mathrm{t+}} Pr t + . (This also implies that both f f f and g g g are equivalences.) The counit map colim Σ D ≤ 0 → D \operatorname*{colim}_{\Sigma}\mathcal{D}^{\leq0}\to\mathcal{D} colim Σ D ≤ 0 → D , and the unit map is C → Sp ( C ) ≤ 0 \mathcal{C}\to\operatorname{Sp}(\mathcal{C})^{\leq0} C → Sp ( C ) ≤ 0 We consider the sequence of stabilizations
colim Σ C = colim ( C → Σ C → Σ ⋯ ) → colim Σ Sp ( C ) ≤ 0 = colim ( Sp ( C ) ≤ 0 → Σ Sp ( C ) ≤ 0 → Σ ⋯ ) → colim Σ Sp ( C ) \operatorname*{colim}_{\Sigma}\mathcal{C}=\operatorname*{colim}(\mathcal{C}\xrightarrow{\Sigma}\mathcal{C}\xrightarrow{\Sigma}\cdots)\to\operatorname*{colim}_{\Sigma}\operatorname{Sp}(\mathcal{C})^{\leq0}=\operatorname*{colim}(\operatorname{Sp}(\mathcal{C})^{\leq0}\xrightarrow{\Sigma}\operatorname{Sp}(\mathcal{C})^{\leq0}\xrightarrow{\Sigma}\cdots)\to\operatorname*{colim}_{\Sigma}\operatorname{Sp}(\mathcal{C}) Σ colim C = colim ( C Σ C Σ ⋯ ) → Σ colim Sp ( C ) ≤ 0 = colim ( Sp ( C ) ≤ 0 Σ Sp ( C ) ≤ 0 Σ ⋯ ) → Σ colim Sp ( C )
induced by C → Sp ( C ) ≤ 0 → Sp ( C ) \mathcal{C}\to\operatorname{Sp}(\mathcal{C})^{\leq0}\to\operatorname{Sp}(\mathcal{C}) C → Sp ( C ) ≤ 0 → Sp ( C ) . This can naturally be identified with Sp ( C ) → ϕ R ^ ( Sp ( C ) ) = R ^ ( Sp ( C ) , Sp ( C ) ≤ 0 ) → ψ Sp ( C ) \operatorname{Sp}(\mathcal{C})\xrightarrow{\phi}\widehat{\mathcal{R}}(\operatorname{Sp}(\mathcal{C}))=\widehat{\mathcal{R}}(\operatorname{Sp}(\mathcal{C}),\operatorname{Sp}(\mathcal{C})^{\leq0})\xrightarrow{\psi}\operatorname{Sp}(\mathcal{C}) Sp ( C ) ϕ R ( Sp ( C )) = R ( Sp ( C ) , Sp ( C ) ≤ 0 ) ψ Sp ( C ) . The composite ψ ∘ ϕ \psi\circ\phi ψ ∘ ϕ is the canonical functor Sp ( C ) → Sp ( Sp ( C ) ) \operatorname{Sp}(\mathcal{C})\to\operatorname{Sp}(\operatorname{Sp}(\mathcal{C})) Sp ( C ) → Sp ( Sp ( C )) so that it is an equivalence. By construction, the composite ϕ ∘ ψ \phi\circ\psi ϕ ∘ ψ induces an identity functor R ^ ( Sp ( C ) ) ⊃ Sp ( C ) ≤ 0 → ≃ Sp ( C ) ≤ 0 ⊂ R ^ ( Sp ( C ) ) \widehat{\mathcal{R}}(\operatorname{Sp}(\mathcal{C}))\supset\operatorname{Sp}(\mathcal{C})^{\leq0}\xrightarrow{\simeq}\operatorname{Sp}(\mathcal{C})^{\leq0}\subset\widehat{\mathcal{R}}(\operatorname{Sp}(\mathcal{C})) R ( Sp ( C )) ⊃ Sp ( C ) ≤ 0 ≃ Sp ( C ) ≤ 0 ⊂ R ( Sp ( C )) , so that ϕ ∘ ψ \phi\circ\psi ϕ ∘ ψ is an equivalence since Sp ( C ) ≤ 0 \operatorname{Sp}(\mathcal{C})^{\leq0} Sp ( C ) ≤ 0 determines a right complete t t t -structure on R ^ ( Sp ( C ) ) \widehat{\mathcal{R}}(\operatorname{Sp}(\mathcal{C})) R ( Sp ( C )) . Thus, ϕ \phi ϕ and ψ \psi ψ are mutually inverse equivalences Sp ( C ) ≃ R ^ ( Sp ( C ) ) \operatorname{Sp}(\mathcal{C})\simeq\widehat{\mathcal{R}}(\operatorname{Sp}(\mathcal{C})) Sp ( C ) ≃ R ( Sp ( C )) . Thus, Sp ( C ) ≤ 0 \operatorname{Sp}(\mathcal{C})^{\leq0} Sp ( C ) ≤ 0 determines a right complete t t t -structure.
For any ( D , D ≤ 0 ) ∈ Pr t + (\mathcal{D},\mathcal{D}^{\leq0})\in\operatorname{Pr}^{\mathrm{t+}} ( D , D ≤ 0 ) ∈ Pr t + , the canonical map colim Σ D ≤ 0 → D \operatorname*{colim}_{\Sigma}\mathcal{D}^{\leq0}\to\mathcal{D} colim Σ D ≤ 0 → D is an equivalence. It follows that S ∘ C \mathcal{S}\circ\mathcal{C} S ∘ C is equivalent to the identity functor. Thus, C \mathcal{C} C is fully faithful. □ \square □
Corollary 3.10. The ∞ \infty ∞ -category Pr t + \operatorname{Pr}^{\mathrm{t+}} Pr t + has small colimits. The forgetful functor Pr t + → Pr S t \operatorname{Pr}^{\mathrm{t+}}\to\operatorname{Pr}^{\mathrm{St}} Pr t + → Pr St preserves small colimits. Moreover, Pr t + \operatorname{Pr}^{\mathrm{t+}} Pr t + belongs to CAlg ( Cat ^ ∞ ) \operatorname{CAlg}(\widehat{\operatorname{Cat}}_{\infty}) CAlg ( Cat ∞ ) .
Proof. Since Pr \operatorname{Pr} Pr has small colimits, Mod S p c o n n ( Pr ) \operatorname{Mod}_{\mathrm{Sp}^{\mathrm{conn}}}(\operatorname{Pr}) Mod Sp conn ( Pr ) also has small colimits (cf. [24 4.2.3.5] ). If f : I → Pr t + f\colon I\to\operatorname{Pr}^{\mathrm{t+}} f : I → Pr t + is a diagram indexed by a small ∞ \infty ∞ -category I I I , we can take a colimit C ∞ \mathcal{C}_{\infty} C ∞ of the composite I → Pr t + → C Mod S p c o n n ( Pr ) I\to\operatorname{Pr}^{\mathrm{t+}}\xrightarrow{\mathcal{C}}\operatorname{Mod}_{\mathrm{Sp}^{\mathrm{conn}}}(\operatorname{Pr}) I → Pr t + C Mod Sp conn ( Pr ) . By Proposition 3.8, we see that S ( C ∞ ) \mathcal{S}(\mathcal{C}_{\infty}) S ( C ∞ ) is a colimit of f f f . Note that if we write ( C i , C i ≤ 0 ) = f ( i ) (\mathcal{C}_i,\mathcal{C}_i^{\leq0})=f(i) ( C i , C i ≤ 0 ) = f ( i ) , the colimit of f f f is ( Sp ( colim i ∈ I C i ≤ 0 ) , Sp ( colim i ∈ I C i ≤ 0 ) ≤ 0 ) (\operatorname{Sp}(\operatorname*{colim}_{i\in I}\mathcal{C}_i^{\leq0}),\operatorname{Sp}(\operatorname*{colim}_{i\in I}\mathcal{C}_i^{\leq0})^{\leq0}) ( Sp ( colim i ∈ I C i ≤ 0 ) , Sp ( colim i ∈ I C i ≤ 0 ) ≤ 0 ) where colim i ∈ I C i ≤ 0 \operatorname*{colim}_{i\in I}\mathcal{C}_i^{\leq0} colim i ∈ I C i ≤ 0 is a colimit taken in Mod S p c o n n ( Pr ) \operatorname{Mod}_{\mathrm{Sp}^{\mathrm{conn}}}(\operatorname{Pr}) Mod Sp conn ( Pr ) . We note the equivalences
Sp ( colim i ∈ I C i ≤ 0 ) ≃ ( colim i ∈ I C i ≤ 0 ) ⊗ S p c o n n S p ≃ colim i ∈ I ( C i ≤ 0 ⊗ S p c o n n S p ) ≃ colim i ∈ I Sp ( C i ≤ 0 ) ≃ colim i ∈ I C i . \operatorname{Sp}\left(\operatorname*{colim}_{i\in I}\mathcal{C}_i^{\leq0}\right)\simeq\left(\operatorname*{colim}_{i\in I}\mathcal{C}_i^{\leq0}\right)\otimes_{\mathrm{Sp}^{\mathrm{conn}}}\mathrm{Sp}\simeq\operatorname*{colim}_{i\in I}\left(\mathcal{C}_i^{\leq0}\otimes_{\mathrm{Sp}^{\mathrm{conn}}}\mathrm{Sp}\right)\simeq\operatorname*{colim}_{i\in I}\operatorname{Sp}(\mathcal{C}_i^{\leq0})\simeq\operatorname*{colim}_{i\in I}\mathcal{C}_i. Sp ( i ∈ I colim C i ≤ 0 ) ≃ ( i ∈ I colim C i ≤ 0 ) ⊗ Sp conn Sp ≃ i ∈ I colim ( C i ≤ 0 ⊗ Sp conn Sp ) ≃ i ∈ I colim Sp ( C i ≤ 0 ) ≃ i ∈ I colim C i .
Thus, the underlying presentable stable ∞ \infty ∞ -category of the colimit of f f f is colim i ∈ I C i \operatorname*{colim}_{i\in I}\mathcal{C}_i colim i ∈ I C i , and Sp ( colim i ∈ I C i ≤ 0 ) ≤ 0 \operatorname{Sp}(\operatorname*{colim}_{i\in I}\mathcal{C}_i^{\leq0})^{\leq0} Sp ( colim i ∈ I C i ≤ 0 ) ≤ 0 is the smallest full subcategory of colim i ∈ I C i \operatorname*{colim}_{i\in I}\mathcal{C}_i colim i ∈ I C i which contains the image of colim i ∈ I C i ≤ 0 \operatorname*{colim}_{i\in I}\mathcal{C}_i^{\leq0} colim i ∈ I C i ≤ 0 and is closed under extensions and small colimits. In particular, the forgetful functor Pr t + → Pr S t \operatorname{Pr}^{\mathrm{t+}}\to\operatorname{Pr}^{\mathrm{St}} Pr t + → Pr St preserves small colimits.
For the last assertion, it suffices to show that the tensor product functor ⊗ : Pr t + × Pr t + → Pr t + \otimes\colon\operatorname{Pr}^{\mathrm{t+}}\times\operatorname{Pr}^{\mathrm{t+}}\to\operatorname{Pr}^{\mathrm{t+}} ⊗ : Pr t + × Pr t + → Pr t + preserves small colimits separately in each variable. The tensor product Pr \operatorname{Pr} Pr preserves small colimits separately in each variable. For ( D , D ≤ 0 ) ∈ P r t + (\mathcal{D},\mathcal{D}^{\leq0}) \in\mathrm{Pr}^{t+} ( D , D ≤ 0 ) ∈ Pr t + , we have a commutative diagram in P r \mathrm{Pr} Pr :
( colim i ∈ I C i ≤ 0 ) ⊗ D ≤ 0 → ( colim i ∈ I C i ) ⊗ D ≃ ↓ ↓ ≃ colim i ∈ I ( C i ≤ 0 ⊗ D ≤ 0 ) → colim i ∈ I ( C i ⊗ D ) \begin{CD}
(\operatorname*{colim}_{i\in I}\mathcal{C}_{i}^{\leq0})\otimes\mathcal{D}^{\leq0} @>>> (\operatorname*{colim}_{i\in I}\mathcal{C}_{i})\otimes\mathcal{D} \\
@V{\simeq}VV @VV{\simeq}V \\
\operatorname*{colim}_{i\in I}(\mathcal{C}_{i}^{\leq0}\otimes\mathcal{D}^{\leq0}) @>>> \operatorname*{colim}_{i\in I}(\mathcal{C}_{i}\otimes\mathcal{D})
\end{CD} ( i ∈ I colim C i ≤ 0 ) ⊗ D ≤ 0 ≃ ↓ ⏐ i ∈ I colim ( C i ≤ 0 ⊗ D ≤ 0 ) ( i ∈ I colim C i ) ⊗ D ↓ ⏐ ≃ i ∈ I colim ( C i ⊗ D )
Consequently, the tensor product in P r t + \mathrm{Pr}^{t+} Pr t + preserves small colimits separately in each variable. □
Corollary 3.11. The ∞ \infty ∞ -category P r t ± \mathrm{Pr}^{t\pm} Pr t ± has small colimits. Moreover, P r t ± \mathrm{Pr}^{t\pm} Pr t ± belongs to C A l g ( C a t ^ † ) \mathrm{CAlg}(\widehat{\mathrm{Cat}}_{\dagger}) CAlg ( Cat † ) .
Proof. Use Corollary 3.10 and the left completion functor L ^ : P r t + → P r t ± \widehat{L}:\mathrm{Pr}^{t+}\to\mathrm{Pr}^{t\pm} L : Pr t + → Pr t ± . (Note that L ^ \widehat{L} L is symmetric monoidal and preserves small colimits.) □
Proposition 3.12. The following holds.
The ∞ \infty ∞ -category P r t \mathrm{Pr}^{t} Pr t has small limits. The forgetful functor P r t → P r S t \mathrm{Pr}^{t}\to\mathrm{Pr}_{\mathrm{St}} Pr t → Pr St preserves small limits.
The ∞ \infty ∞ -category P r t + \mathrm{Pr}^{t+} Pr t + has small limits.
Proof. A small limit in P r \mathrm{Pr} Pr can be computed as a small limit in C a t ^ ∞ \widehat{\mathrm{Cat}}_{\infty} Cat ∞ , so that lim i ∈ I ( C i , C i ≤ 0 ) \lim_{i\in I}(\mathcal{C}_{i},\mathcal{C}_{i}^{\leq0}) lim i ∈ I ( C i , C i ≤ 0 ) of a diagram D : I → P r t ⊂ Fun ( Δ 1 , P r ) D:I\to\mathrm{Pr}^{t}\subset\operatorname{Fun}(\Delta^{1},\mathrm{Pr}) D : I → Pr t ⊂ Fun ( Δ 1 , Pr ) is lim i ∈ I C i ≤ 0 → lim i ∈ I C i \lim_{i\in I}\mathcal{C}_{i}^{\leq0}\to\lim_{i\in I}\mathcal{C}_{i} lim i ∈ I C i ≤ 0 → lim i ∈ I C i where lim i ∈ I C i ≤ 0 \lim_{i\in I}\mathcal{C}_{i}^{\leq0} lim i ∈ I C i ≤ 0 and lim i ∈ I C i \lim_{i\in I}\mathcal{C}_{i} lim i ∈ I C i are limits in C a t ^ ∞ \widehat{\mathrm{Cat}}_{\infty} Cat ∞ (equivalently, limits in P r \mathrm{Pr} Pr ). Thus the induced functor C : lim i ∈ I C i ≤ 0 → lim i ∈ I C i \mathcal{C}:\lim_{i\in I}\mathcal{C}_{i}^{\leq0}\to\lim_{i\in I}\mathcal{C}_{i} C : lim i ∈ I C i ≤ 0 → lim i ∈ I C i is fully faithful. Since each C i → C j \mathcal{C}_{i}\to\mathcal{C}_{j} C i → C j is a right t t t -exact colimit-preserving functor between stable ∞ \infty ∞ -categories, it follows that the essential image of C \mathcal{C} C is closed under small colimits and extensions. Thus, a limit of D → Fun ( Δ 1 , P r ) D\to\operatorname{Fun}(\Delta^{1},\mathrm{Pr}) D → Fun ( Δ 1 , Pr ) lies in P r t \mathrm{Pr}^{t} Pr t . Thus, P r t \mathrm{Pr}^{t} Pr t has small limits, and the forgetful functor P r t → P r S t \mathrm{Pr}^{t}\to\mathrm{Pr}_{\mathrm{St}} Pr t → Pr St preserves small limits. This proves (1). The right completion functor P r t → P r t + \mathrm{Pr}^{t}\to\mathrm{Pr}^{t+} Pr t → Pr t + is right adjoint so that it preserves small limits. This proves (2). □
We define t t t -structured presentable ( ∞ , n ) (\infty,n) ( ∞ , n ) -categories.
Definition 3.13. For a positive integer n ≥ 1 n\geq1 n ≥ 1 and P r t + ∈ C A l g ( C a t ^ † ) \mathrm{Pr}^{t+}\in\mathrm{CAlg}(\widehat{\mathrm{Cat}}_{\dagger}) Pr t + ∈ CAlg ( Cat † ) (see Corollary 3.10 ), we define n P r t + ∈ C A l g ( C a t ^ † ) n\mathrm{Pr}^{t+}\in\mathrm{CAlg}(\widehat{\mathrm{Cat}}_{\dagger}) n Pr t + ∈ CAlg ( Cat † ) (see Construction 3.7). By convention, 1 P r t + = P r t + 1\mathrm{Pr}^{t+}=\mathrm{Pr}^{t+} 1 Pr t + = Pr t + and 0 P r t + = ( S p , S p c o n n ) 0\mathrm{Pr}^{t+}=(\mathrm{Sp},\mathrm{Sp}^{\mathrm{conn}}) 0 Pr t + = ( Sp , Sp conn ) . We refer to an object of n P r t + n\mathrm{Pr}^{t+} n Pr t + as a t t t -structured presentable ( ∞ , n ) (\infty,n) ( ∞ , n ) -category.
For a positive integer n ≥ 1 n\geq1 n ≥ 1 and P r t ± ∈ C A l g ( C a t ^ † ) \mathrm{Pr}^{t\pm}\in\mathrm{CAlg}(\widehat{\mathrm{Cat}}_{\dagger}) Pr t ± ∈ CAlg ( Cat † ) (see Corollary 3.11), we define n P r t ± ∈ C A l g ( C a t ^ † ) n\mathrm{Pr}^{t\pm}\in\mathrm{CAlg}(\widehat{\mathrm{Cat}}_{\dagger}) n Pr t ± ∈ CAlg ( Cat † ) (see Construction 3.7). By convention, 1 P r t ± = P r t ± 1\mathrm{Pr}^{t\pm}=\mathrm{Pr}^{t\pm} 1 Pr t ± = Pr t ± and 0 P r t ± = ( S p , S p c o n n ) 0\mathrm{Pr}^{t\pm}=(\mathrm{Sp},\mathrm{Sp}^{\mathrm{conn}}) 0 Pr t ± = ( Sp , Sp conn ) . We refer to an object of n P r t ± n\mathrm{Pr}^{t\pm} n Pr t ± as a complete t t t -structured presentable ( ∞ , n ) (\infty,n) ( ∞ , n ) -category.
Remark 3.14. In this paper, we will not consider P r t \mathrm{Pr}^{t} Pr t -modules and their higher variants as the main objects of interest because they do not appear in categorified Koszul duality. Thus, we simply call objects of n P r t + n\mathrm{Pr}^{t+} n Pr t + t t t -structured presentable ( ∞ , n ) (\infty,n) ( ∞ , n ) -categories.
Completions of t t t -structured presentable categories.
Definition 3.15. We consider the symmetric monoidal localization functor L ^ : P r t + → P r t ± \widehat{L}:\mathrm{Pr}^{t+}\to\mathrm{Pr}^{t\pm} L : Pr t + → Pr t ± , whose right adjoint is the lax symmetric monoidal inclusion P r t ± ⊂ P r t + \mathrm{Pr}^{t\pm}\subset\mathrm{Pr}^{t+} Pr t ± ⊂ Pr t + . The base change M o d P r t + ( C a t ^ † ) → M o d P r t ± ( C a t ^ † ) \mathrm{Mod}_{\mathrm{Pr}^{t+}}(\widehat{\mathrm{Cat}}_{\dagger})\to\mathrm{Mod}_{\mathrm{Pr}^{t\pm}}(\widehat{\mathrm{Cat}}_{\dagger}) Mod Pr t + ( Cat † ) → Mod Pr t ± ( Cat † ) induces a symmetric monoidal functor 2 P r t + → 2 P r t ± 2\mathrm{Pr}^{t+}\to2\mathrm{Pr}^{t\pm} 2 Pr t + → 2 Pr t ± . Inductively, for any n ≥ 2 n\geq2 n ≥ 2 , a symmetric monoidal functor ( n − 1 ) P r t + → ( n − 1 ) P r t ± (n-1)\mathrm{Pr}^{t+}\to(n-1)\mathrm{Pr}^{t\pm} ( n − 1 ) Pr t + → ( n − 1 ) Pr t ± induces the base change functor ( − ) ⊗ ( n − 1 ) P r t + ( n − 1 ) P r t ± : n P r t + ⟶ n P r t ± (-)\otimes_{(n-1)\mathrm{Pr}^{t+}}(n-1)\mathrm{Pr}^{t\pm}:n\mathrm{Pr}^{t+}\longrightarrow n\mathrm{Pr}^{t\pm} ( − ) ⊗ ( n − 1 ) Pr t + ( n − 1 ) Pr t ± : n Pr t + ⟶ n Pr t ± .
For D ∈ n P r t + \mathcal{D}\in n\mathrm{Pr}^{t+} D ∈ n Pr t + , we define the completion of D \mathcal{D} D as D ⊗ ( n − 1 ) P r t + P r t ± \mathcal{D}\otimes_{(n-1)\mathrm{Pr}^{t+}}\mathrm{Pr}^{t\pm} D ⊗ ( n − 1 ) Pr t + Pr t ± . For n ≥ 2 n\geq2 n ≥ 2 , we write n L ^ = ( − ) ⊗ ( n − 1 ) P r t + ( n − 1 ) P r t ± n\widehat{L}=(-)\otimes_{(n-1)\mathrm{Pr}^{t+}}(n-1)\mathrm{Pr}^{t\pm} n L = ( − ) ⊗ ( n − 1 ) Pr t + ( n − 1 ) Pr t ± and refer to it as the completion functor.
Let S S S be the collection of all 1 1 1 -morphisms f f f of P r t + \mathrm{Pr}^{t+} Pr t + such that L ^ ( f ) \widehat{L}(f) L ( f ) is an equivalence. Note that S S S is multiplicative in the sense that for any f ∈ S f\in S f ∈ S and any object C ∈ P r t + \mathcal{C}\in\mathrm{Pr}^{t+} C ∈ Pr t + , f ⊗ id C f\otimes\operatorname{id}_{\mathcal{C}} f ⊗ id C also belongs to S S S . According to [23 5.2.7.12] and [24 1.3.4.1, 4.1.3.4] , P r t ± \mathrm{Pr}^{t\pm} Pr t ± is the symmetric monoidal ∞ \infty ∞ -category obtained from P r t + \mathrm{Pr}^{t+} Pr t + by inverting S S S : For any symmetric monoidal ∞ \infty ∞ -category D ∞ \mathcal{D}^{\infty} D ∞ , the functor Map ( P r t ± , D ∞ ) → Map ( P r t + , D ∞ ) \operatorname{Map}(\mathrm{Pr}^{t\pm},\mathcal{D}^{\infty})\to\operatorname{Map}(\mathrm{Pr}^{t+},\mathcal{D}^{\infty}) Map ( Pr t ± , D ∞ ) → Map ( Pr t + , D ∞ ) induced by the composition with L ^ \widehat{L} L is fully faithful (here Map means the mapping space between symmetric monoidal ∞ \infty ∞ -categories). The essential image consists of those symmetric monoidal functors which carry S S S to equivalences in D \mathcal{D} D . The underlying ∞ \infty ∞ -category P r t ± \mathrm{Pr}^{\mathrm{t}\pm} Pr t ± is obtained from P r t + \mathrm{Pr}^{\mathrm{t}+} Pr t + by inverting S S S .
Proposition 3.16. Let ι : Mod P r t ± ( C a t ^ † ) → Mod P r t + ( C a t ^ † ) \iota: \operatorname{Mod}_{\mathrm{Pr}^{\mathrm{t}\pm}}(\widehat{\mathrm{Cat}}_{\dagger}) \to\operatorname{Mod}_{\mathrm{Pr}^{\mathrm{t}+}}(\widehat{\mathrm{Cat}}_{\dagger}) ι : Mod Pr t ± ( Cat † ) → Mod Pr t + ( Cat † ) be the restriction functor along L ^ : P r t + → P r t ± \widehat{L} : \mathrm{Pr}^{\mathrm{t}+} \to\mathrm{Pr}^{\mathrm{t}\pm} L : Pr t + → Pr t ± . Then ι \iota ι is fully faithful.
Proof. Since C a t ^ † \widehat{\mathrm{Cat}}_{\dagger} Cat † has geometric realizations, there exists a left adjoint ℓ : Mod P r t + ( C a t ^ † ) → Mod P r t ± ( C a t ^ † ) \ell: \operatorname{Mod}_{\mathrm{Pr}^{\mathrm{t}+}}(\widehat{\mathrm{Cat}}_{\dagger}) \to\operatorname{Mod}_{\mathrm{Pr}^{\mathrm{t}\pm}}(\widehat{\mathrm{Cat}}_{\dagger}) ℓ : Mod Pr t + ( Cat † ) → Mod Pr t ± ( Cat † ) of ι \iota ι given by the base change along L ^ : P r t + → P r t ± \widehat{L} : \mathrm{Pr}^{\mathrm{t}+} \to\mathrm{Pr}^{\mathrm{t}\pm} L : Pr t + → Pr t ± . To prove that ι \iota ι is fully faithful, it suffices to show that for any M ∈ Mod P r t ± ( C a t ^ † ) \mathcal{M} \in\operatorname{Mod}_{\mathrm{Pr}^{\mathrm{t}\pm}}(\widehat{\mathrm{Cat}}_{\dagger}) M ∈ Mod Pr t ± ( Cat † ) the counit map ℓ ( ι ( M ) ) → M \ell(\iota(\mathcal{M})) \to\mathcal{M} ℓ ( ι ( M )) → M is an equivalence. Note that both ℓ \ell ℓ and ι \iota ι preserve (large) colimits. In addition, M \mathcal{M} M can be described as a colimit of a free P r t ± \mathrm{Pr}^{\mathrm{t}\pm} Pr t ± -modules. Thus, we may and will assume that E = C ⊗ P r t ± \mathcal{E} = \mathcal{C} \otimes\mathrm{Pr}^{\mathrm{t}\pm} E = C ⊗ Pr t ± with some C ∈ C a t ^ † \mathcal{C} \in\widehat{\mathrm{Cat}}_{\dagger} C ∈ Cat † . The counit map ℓ ( ι ( C ⊗ P r t ± ) ) ≃ ℓ ( C ⊗ P r t + [ S − 1 ] ) ≃ C ⊗ ( P r t + [ S − 1 ] ) [ S ‾ − 1 ] → C ⊗ P r t + [ S − 1 ] \ell(\iota(\mathcal{C} \otimes\mathrm{Pr}^{\mathrm{t}\pm})) \simeq\ell(\mathcal{C} \otimes\mathrm{Pr}^{\mathrm{t}+}[S^{-1}]) \simeq\mathcal{C} \otimes(\mathrm{Pr}^{\mathrm{t}+}[S^{-1}])[\overline{S}^{-1}] \to\mathcal{C} \otimes\mathrm{Pr}^{\mathrm{t}+}[S^{-1}] ℓ ( ι ( C ⊗ Pr t ± )) ≃ ℓ ( C ⊗ Pr t + [ S − 1 ]) ≃ C ⊗ ( Pr t + [ S − 1 ]) [ S − 1 ] → C ⊗ Pr t + [ S − 1 ] is an equivalence since ( P r t + [ S − 1 ] ) [ S ‾ − 1 ] → P r t + [ S − 1 ] (\mathrm{Pr}^{\mathrm{t}+}[S^{-1}])[\overline{S}^{-1}] \to\mathrm{Pr}^{\mathrm{t}+}[S^{-1}] ( Pr t + [ S − 1 ]) [ S − 1 ] → Pr t + [ S − 1 ] is an equivalence where [ S ‾ − 1 ] [\overline{S}^{-1}] [ S − 1 ] indicates the localization defined by inverting the image S ‾ \overline{S} S of S S S in P r t + [ S − 1 ] \mathrm{Pr}^{\mathrm{t}+}[S^{-1}] Pr t + [ S − 1 ] . □ \square □
Let P r k t + \mathrm{Pr}^{\mathrm{t}+}_{k} Pr k t + denote \operatorname{Mod}_{(\operatorname{Mod}_{k},\operatorname{Mod}_{\leq0}_{k})}(\mathrm{Pr}^{\mathrm{t}+}) . Let P r k t ± \mathrm{Pr}^{\mathrm{t}\pm}_{k} Pr k t ± denote \operatorname{Mod}_{(\operatorname{Mod}_{k},\operatorname{Mod}_{\leq0}_{k})}(\mathrm{Pr}^{\mathrm{t}\pm}) . The symmetric monoidal functor L ^ : P r t + → P r t ± \widehat{L} : \mathrm{Pr}^{\mathrm{t}+} \to\mathrm{Pr}^{\mathrm{t}\pm} L : Pr t + → Pr t ± induces P r k t + → P r k t ± \mathrm{Pr}^{\mathrm{t}+}_{k} \to\mathrm{Pr}^{\mathrm{t}\pm}_{k} Pr k t + → Pr k t ± which we also denote by L ^ \widehat{L} L .
The same argument shows:
Proposition 3.17. Let ι : Mod P r k t ± ( C a t ^ † ) → Mod P r k t + ( C a t ^ † ) \iota: \operatorname{Mod}_{\mathrm{Pr}^{\mathrm{t}\pm}_{k}}(\widehat{\mathrm{Cat}}_{\dagger}) \to\operatorname{Mod}_{\mathrm{Pr}^{\mathrm{t}+}_{k}}(\widehat{\mathrm{Cat}}_{\dagger}) ι : Mod Pr k t ± ( Cat † ) → Mod Pr k t + ( Cat † ) denote the restriction functor along L ^ : P r k t + → P r k t ± \widehat{L} : \mathrm{Pr}^{\mathrm{t}+}_{k} \to\mathrm{Pr}^{\mathrm{t}\pm}_{k} L : Pr k t + → Pr k t ± . Then ι \iota ι is fully faithful.
In particular, we have the localization adjoint pair
ℓ : Mod P r k t + ( C a t ^ † ) ⇄ Mod P r k t ± ( C a t ^ † ) : ι \ell: \operatorname{Mod}_{\mathrm{Pr}^{\mathrm{t}+}_{k}}(\widehat{\mathrm{Cat}}_{\dagger}) \rightleftarrows\operatorname{Mod}_{\mathrm{Pr}^{\mathrm{t}\pm}_{k}}(\widehat{\mathrm{Cat}}_{\dagger}) : \iota ℓ : Mod Pr k t + ( Cat † ) ⇄ Mod Pr k t ± ( Cat † ) : ι
We also refer to ℓ \ell ℓ as the completion functor.
Remark 3.18. At the time of writing, we do not know whether or not ι \iota ι preserves κ 0 \kappa_{0} κ 0 -compact objects.
Underlying stable ( ∞ , n ) (\infty,n) ( ∞ , n ) -categories
Consider a stable presentable ∞ \infty ∞ -category ( C , C ≤ 0 ) (\mathcal{C},\mathcal{C}^{\leq0}) ( C , C ≤ 0 ) equipped with a t t t -structure. One can consider the underlying stable presentable ∞ \infty ∞ -category C \mathcal{C} C obtained by forgetting the t t t -structure C ≤ 0 \mathcal{C}^{\leq0} C ≤ 0 . We define underlying stable ( ∞ , n ) (\infty,n) ( ∞ , n ) -categories obtained from objects of n P r t + n\mathrm{Pr}^{\mathrm{t}+} n Pr t + .
Recall from Proposition 3.8 that P r t + \mathrm{Pr}^{\mathrm{t}+} Pr t + can be regarded as a full subcategory of Mod S p c o n n ( P r ) \operatorname{Mod}_{\mathrm{Sp}^{\mathrm{conn}}}(\mathrm{Pr}) Mod Sp conn ( Pr ) through the assignment given by ( C , C ≤ 0 ) ↦ C ≤ 0 (\mathcal{C},\mathcal{C}^{\leq0}) \mapsto\mathcal{C}^{\leq0} ( C , C ≤ 0 ) ↦ C ≤ 0 . We also note that the stabilization S p ( C ≤ 0 ) \mathrm{Sp}(\mathcal{C}^{\leq0}) Sp ( C ≤ 0 ) of C ≤ 0 \mathcal{C}^{\leq0} C ≤ 0 is canonically equivalent to C \mathcal{C} C . The tensor product with S p \mathrm{Sp} Sp over S p c o n n \mathrm{Sp}^{\mathrm{conn}} Sp conn induces
S p ⊗ S p c o n n ( − ) : Mod S p c o n n ( P r ) ⟶ Mod S p ( P r ) \mathrm{Sp} \otimes_{\mathrm{Sp}^{\mathrm{conn}}}(-) : \operatorname{Mod}_{\mathrm{Sp}^{\mathrm{conn}}}(\mathrm{Pr}) \longrightarrow\operatorname{Mod}_{\mathrm{Sp}}(\mathrm{Pr}) Sp ⊗ Sp conn ( − ) : Mod Sp conn ( Pr ) ⟶ Mod Sp ( Pr )
which carries P \mathcal{P} P to the stabilization S p ( P ) ≃ S p ⊗ S p c o n n P \mathrm{Sp}(\mathcal{P}) \simeq\mathrm{Sp} \otimes_{\mathrm{Sp}^{\mathrm{conn}}} \mathcal{P} Sp ( P ) ≃ Sp ⊗ Sp conn P . This observation yields a commutative diagram
P r t + ⟶ P r S t ↓ S p ⊗ S p c o n n ( − ) ↓ Mod S p c o n n ( P r ) → S p ⊗ S p c o n n ( − ) Mod S p ( P r ) \begin{aligned}
\mathrm{Pr}^{\mathrm{t}+} &\longrightarrow\mathrm{Pr}_{\mathrm{St}} \\
\downarrow\phantom{\mathrm{Sp} \otimes_{\mathrm{Sp}^{\mathrm{conn}}}(-)} & \qquad\downarrow\\
\operatorname{Mod}_{\mathrm{Sp}^{\mathrm{conn}}}(\mathrm{Pr}) &\xrightarrow{\mathrm{Sp} \otimes_{\mathrm{Sp}^{\mathrm{conn}}}(-)} \operatorname{Mod}_{\mathrm{Sp}}(\mathrm{Pr})
\end{aligned} Pr t + ↓ Sp ⊗ Sp conn ( − ) Mod Sp conn ( Pr ) ⟶ Pr St ↓ Sp ⊗ Sp conn ( − ) Mod Sp ( Pr )
where the vertical arrows are fully faithful embeddings. The upper horizontal arrow is the forgetful functor, which is also regarded as the functor given informally by ( C , C ≤ 0 ) ↦ S p ( C ≤ 0 ) (\mathcal{C},\mathcal{C}^{\leq0}) \mapsto\mathrm{Sp}(\mathcal{C}^{\leq0}) ( C , C ≤ 0 ) ↦ Sp ( C ≤ 0 ) . The forgetful functor P r t + → P r S t \mathrm{Pr}^{\mathrm{t}+} \to\mathrm{Pr}_{\mathrm{St}} Pr t + → Pr St plays the role of the canonical functor S p t + → S p \mathrm{Sp}^{\mathrm{t}+} \to\mathrm{Sp} Sp t + → Sp . This leads to the following definition:
Definition 3.19. For E ∈ 2 P r t + \mathcal{E} \in2\mathrm{Pr}^{\mathrm{t}+} E ∈ 2 Pr t + , we define the underlying stable presentable ( ∞ , 2 ) (\infty,2) ( ∞ , 2 ) -category of E \mathcal{E} E to be P r S t ⊗ P r t + E \mathrm{Pr}_{\mathrm{St}} \otimes_{\mathrm{Pr}^{\mathrm{t}+}} \mathcal{E} Pr St ⊗ Pr t + E . The assignment E ↦ P r S t ⊗ P r t + E \mathcal{E} \mapsto\mathrm{Pr}_{\mathrm{St}} \otimes_{\mathrm{Pr}^{\mathrm{t}+}} \mathcal{E} E ↦ Pr St ⊗ Pr t + E induces a symmetric monoidal functor λ 2 : 2 P r t + → 2 P r S t \lambda_{2} : 2\mathrm{Pr}^{\mathrm{t}+} \to2\mathrm{Pr}_{\mathrm{St}} λ 2 : 2 Pr t + → 2 Pr St .
This construction can be iterated: if a symmetric monoidal functor λ n − 1 : ( n − 1 ) P r t + → ( n − 1 ) P r S t \lambda_{n-1} : (n-1)\mathrm{Pr}^{\mathrm{t}+} \to(n-1)\mathrm{Pr}_{\mathrm{St}} λ n − 1 : ( n − 1 ) Pr t + → ( n − 1 ) Pr St has been defined, we define λ n : n P r t + → n P r S t \lambda_{n} : n\mathrm{Pr}^{\mathrm{t}+} \to n\mathrm{Pr}_{\mathrm{St}} λ n : n Pr t + → n Pr St by λ n ( P ) = ( n − 1 ) P r S t ⊗ ( n − 1 ) P r t + P \lambda_{n}(\mathcal{P}) = (n-1)\mathrm{Pr}_{\mathrm{St}} \otimes_{(n-1)\mathrm{Pr}^{\mathrm{t}+}} \mathcal{P} λ n ( P ) = ( n − 1 ) Pr St ⊗ ( n − 1 ) Pr t + P . For P ∈ n P r t + \mathcal{P} \in n\mathrm{Pr}^{\mathrm{t}+} P ∈ n Pr t + , we refer to λ n ( P ) \lambda_{n}(\mathcal{P}) λ n ( P ) as the underlying stable presentable ( ∞ , n ) (\infty,n) ( ∞ , n ) -category of P \mathcal{P} P .
If we replace P r t + → P r S t \mathrm{Pr}^{t+} \to\mathrm{Pr}_{\mathrm{St}} Pr t + → Pr St with the forgetful functor P r t ± → P r S t \mathrm{Pr}^{t\pm} \to\mathrm{Pr}_{\mathrm{St}} Pr t ± → Pr St , then the same construction defines ( n − 1 ) P r t ± → ( n − 1 ) P r S t (n-1)\mathrm{Pr}^{t\pm} \to(n-1)\mathrm{Pr}_{\mathrm{St}} ( n − 1 ) Pr t ± → ( n − 1 ) Pr St . This also induces
( − ) ⊗ ( n − 1 ) P r t ± ( n − 1 ) P r S t : n P r t ± → n P r S t . (-)\otimes_{(n-1)\mathrm{Pr}^{t\pm}}(n-1)\mathrm{Pr}_{\mathrm{St}}:n\mathrm{Pr}^{t\pm}\to n\mathrm{Pr}_{\mathrm{St}}. ( − ) ⊗ ( n − 1 ) Pr t ± ( n − 1 ) Pr St : n Pr t ± → n Pr St .
For P ∈ n P r t ± \mathcal{P}\in n\mathrm{Pr}^{t\pm} P ∈ n Pr t ± , we refer to P ⊗ ( n − 1 ) P r t ± ( n − 1 ) P r S t \mathcal{P}\otimes_{(n-1)\mathrm{Pr}^{t\pm}}(n-1)\mathrm{Pr}_{\mathrm{St}} P ⊗ ( n − 1 ) Pr t ± ( n − 1 ) Pr St as the underlying stable presentable ( ∞ , n ) (\infty,n) ( ∞ , n ) -category of P \mathcal{P} P .
k k k -linear contexts
Let k k k be a connective commutative ring spectrum. In subsequent sections, we work in k k k -linear contexts. We generalize t t t -structured presentable ( ∞ , n ) (\infty,n) ( ∞ , n ) -categories to the k k k -linear setting. This generalization is formal. We write P r k t + = Mod Mod k ( P r t + ) \mathrm{Pr}^{t+}_{k}=\operatorname{Mod}_{\operatorname{Mod}_{k}}(\mathrm{Pr}^{t+}) Pr k t + = Mod Mod k ( Pr t + ) and P r k t ± = Mod Mod k ( P r t ± ) \mathrm{Pr}^{t\pm}_{k}=\operatorname{Mod}_{\operatorname{Mod}_{k}}(\mathrm{Pr}^{t\pm}) Pr k t ± = Mod Mod k ( Pr t ± ) . Here we abuse notation by writing Mod k \operatorname{Mod}_{k} Mod k for ( Mod k , Mod k ≤ 0 ) ∈ P r t ± (\operatorname{Mod}_{k},\operatorname{Mod}_{k}^{\leq0})\in\mathrm{Pr}^{t\pm} ( Mod k , Mod k ≤ 0 ) ∈ Pr t ± . Using Corollary 3.10 and Corollary 3.11 together with [24 Corollary 4.2.3.5] we see that both P r k t + \mathrm{Pr}^{t+}_{k} Pr k t + and P r k t ± \mathrm{Pr}^{t\pm}_{k} Pr k t ± belong to CAlg ( Cat † ) \operatorname{CAlg}(\operatorname{Cat}^{\dagger}) CAlg ( Cat † ) . For n ≥ 0 n\geq0 n ≥ 0 , we can consider n P r k t + n\mathrm{Pr}^{t+}_{k} n Pr k t + and n P r k t ± n\mathrm{Pr}^{t\pm}_{k} n Pr k t ± . By convention, 1 P r k t + = P r k t + 1\mathrm{Pr}^{t+}_{k}=\mathrm{Pr}^{t+}_{k} 1 Pr k t + = Pr k t + , 1 P r k t ± = P r k t ± 1\mathrm{Pr}^{t\pm}_{k}=\mathrm{Pr}^{t\pm}_{k} 1 Pr k t ± = Pr k t ± , 0 P r k t + = ( Mod k , Mod k ≤ 0 ) 0\mathrm{Pr}^{t+}_{k}=(\operatorname{Mod}_{k},\operatorname{Mod}_{k}^{\leq0}) 0 Pr k t + = ( Mod k , Mod k ≤ 0 ) , and 0 P r k t ± = ( Mod k , Mod k ≤ 0 ) 0\mathrm{Pr}^{t\pm}_{k}=(\operatorname{Mod}_{k},\operatorname{Mod}_{k}^{\leq0}) 0 Pr k t ± = ( Mod k , Mod k ≤ 0 ) . We refer to an object of n P r k t + n\mathrm{Pr}^{t+}_{k} n Pr k t + (resp. n P r k t ± n\mathrm{Pr}^{t\pm}_{k} n Pr k t ± ) as a t t t -structured k k k -linear presentable ( ∞ , n ) (\infty,n) ( ∞ , n ) -category (resp. a complete t t t -structured k k k -linear presentable ( ∞ , n ) (\infty,n) ( ∞ , n ) -category).
There is a completion functor
n L ^ = ( − ) ⊗ ( n − 1 ) P r k t + ( n − 1 ) P r k t ± : n P r k t + ⟶ n P r k t ± . n\widehat{L}=(-)\otimes_{(n-1)\mathrm{Pr}^{t+}_{k}}(n-1)\mathrm{Pr}^{t\pm}_{k}:n\mathrm{Pr}^{t+}_{k}\longrightarrow n\mathrm{Pr}^{t\pm}_{k}. n L = ( − ) ⊗ ( n − 1 ) Pr k t + ( n − 1 ) Pr k t ± : n Pr k t + ⟶ n Pr k t ± .
which we define in the same way.
The underlying k k k -linear presentable ( ∞ , n ) (\infty,n) ( ∞ , n ) -categories are defined by functors
( − ) ⊗ ( n − 1 ) P r k t + ( n − 1 ) P r k : n P r k t + → n P r k , ( − ) ⊗ ( n − 1 ) P r k t ± ( n − 1 ) P r k : n P r k t ± → n P r k (-)\otimes_{(n-1)\mathrm{Pr}^{t+}_{k}}(n-1)\mathrm{Pr}_{k}:n\mathrm{Pr}^{t+}_{k}\to n\mathrm{Pr}_{k},\qquad
(-)\otimes_{(n-1)\mathrm{Pr}^{t\pm}_{k}}(n-1)\mathrm{Pr}_{k}:n\mathrm{Pr}^{t\pm}_{k}\to n\mathrm{Pr}_{k} ( − ) ⊗ ( n − 1 ) Pr k t + ( n − 1 ) Pr k : n Pr k t + → n Pr k , ( − ) ⊗ ( n − 1 ) Pr k t ± ( n − 1 ) Pr k : n Pr k t ± → n Pr k
which we define in the same way. We have the diagram
n P r k t + → n L ^ n P r k t ± ( − ) ⊗ ( n − 1 ) P r k t + ( n − 1 ) P r k ↓ ↓ ( − ) ⊗ ( n − 1 ) P r k t ± ( n − 1 ) P r k n P r k = n P r k \begin{CD}
n\mathrm{Pr}^{t+}_{k} @>{n\widehat{L}}>> n\mathrm{Pr}^{t\pm}_{k}\\
@V{(-)\otimes_{(n-1)\mathrm{Pr}^{t+}_{k}}(n-1)\mathrm{Pr}_{k}}VV @VV{(-)\otimes_{(n-1)\mathrm{Pr}^{t\pm}_{k}}(n-1)\mathrm{Pr}_{k}}V\\
n\mathrm{Pr}_{k} @= n\mathrm{Pr}_{k}
\end{CD} n Pr k t + ( − ) ⊗ ( n − 1 ) Pr k t + ( n − 1 ) Pr k ↓ ⏐ n Pr k n L n Pr k t ± ↓ ⏐ ( − ) ⊗ ( n − 1 ) Pr k t ± ( n − 1 ) Pr k n Pr k
Note that for n ≥ 1 n\geq1 n ≥ 1 this diagram does not commute.
Higher module categories
Let A A A be a commutative ring spectrum, that is, A ∈ CAlg ( S p ) A\in\operatorname{CAlg}(\mathrm{Sp}) A ∈ CAlg ( Sp ) . Let us consider LMod A ( S p ) ≃ Mod A ( S p ) \operatorname{LMod}_{A}(\mathrm{Sp})\simeq\operatorname{Mod}_{A}(\mathrm{Sp}) LMod A ( Sp ) ≃ Mod A ( Sp ) . This ∞ \infty ∞ -category naturally has a symmetric monoidal structure. Moreover, it admits a right complete accessible t t t -structure (cf. Example 3.2, Remark 3.3), which exhibits LMod A ( S p ) \operatorname{LMod}_{A}(\mathrm{Sp}) LMod A ( Sp ) as an object of CAlg ( P r t + ) \operatorname{CAlg}(\mathrm{Pr}^{t+}) CAlg ( Pr t + ) . Next we consider LMod LMod A ( S p ) ( P r t + ) \operatorname{LMod}_{\operatorname{LMod}_{A}(\mathrm{Sp})}(\mathrm{Pr}^{t+}) LMod LMod A ( Sp ) ( Pr t + ) . This category also has a symmetric monoidal structure, which exhibits it as an object of CAlg ( 2 P r t + ) \operatorname{CAlg}(2\mathrm{Pr}^{t+}) CAlg ( 2 Pr t + ) . We write
P r A t + = LMod LMod A ( S p ) ( P r t + ) . \mathrm{Pr}^{t+}_{A}=\operatorname{LMod}_{\operatorname{LMod}_{A}(\mathrm{Sp})}(\mathrm{Pr}^{t+}). Pr A t + = LMod LMod A ( Sp ) ( Pr t + ) .
We consider 2 P r A t + = LMod P r A t + ( 2 P r t + ) 2\mathrm{Pr}^{t+}_{A}=\operatorname{LMod}_{\mathrm{Pr}^{t+}_{A}}(2\mathrm{Pr}^{t+}) 2 Pr A t + = LMod Pr A t + ( 2 Pr t + ) . We can regard it as an object of CAlg ( 3 P r t + ) \operatorname{CAlg}(3\mathrm{Pr}^{t+}) CAlg ( 3 Pr t + ) . In the next step we have
3 P r A t + = LMod LMod LMod A ( S p ) ( P r t + ) ( 2 P r t + ) ( 3 P r t + ) . 3\mathrm{Pr}^{t+}_{A}=\operatorname{LMod}_{\operatorname{LMod}_{\operatorname{LMod}_{A}(\mathrm{Sp})}(\mathrm{Pr}^{t+})}(2\mathrm{Pr}^{t+})(3\mathrm{Pr}^{t+}). 3 Pr A t + = LMod LMod LMod A ( Sp ) ( Pr t + ) ( 2 Pr t + ) ( 3 Pr t + ) .
We can repeat this procedure to obtain m P r A t + ∈ CAlg ( ( m + 1 ) P r t + ) m\mathrm{Pr}^{t+}_{A}\in\operatorname{CAlg}((m+1)\mathrm{Pr}^{t+}) m Pr A t + ∈ CAlg (( m + 1 ) Pr t + ) .
We will work in a slightly more general setting. Let n n n be a positive integer. Let A ∈ Alg n ( S p ) A\in\operatorname{Alg}_{n}(\mathrm{Sp}) A ∈ Alg n ( Sp ) . We will associate m P r A t + m\mathrm{Pr}^{t+}_{A} m Pr A t + , that is an E n − m − 1 E_{n-m-1} E n − m − 1 -monoidal t t t -structured presentable ( ∞ , m + 1 ) (\infty,m+1) ( ∞ , m + 1 ) -category, i.e., an object of Alg n − m − 1 ( ( m + 1 ) P r t + ) \operatorname{Alg}_{n-m-1}((m+1)\mathrm{Pr}^{t+}) Alg n − m − 1 (( m + 1 ) Pr t + ) . The construction gives a functor
θ S p , n ↦ m + : Alg n ( S p ) ⟶ Alg n − m − 1 ( ( m + 1 ) P r t + ) \theta^{+}_{\mathrm{Sp},n\mapsto m}:\operatorname{Alg}_{n}(\mathrm{Sp})\longrightarrow\operatorname{Alg}_{n-m-1}((m+1)\mathrm{Pr}^{t+}) θ Sp , n ↦ m + : Alg n ( Sp ) ⟶ Alg n − m − 1 (( m + 1 ) Pr t + )
which carries A A A to m P r A t + m\mathrm{Pr}^{t+}_{A} m Pr A t + . The construction is given in Construction 3.28.
When A A A is connective, the stable ∞ \infty ∞ -category ( LMod A , LMod A ≤ 0 ) (\operatorname{LMod}_{A},\operatorname{LMod}_{A}^{\leq0}) ( LMod A , LMod A ≤ 0 ) equipped with the standard t t t -structure is both left and right complete so that ( LMod A , LMod A ≤ 0 ) (\operatorname{LMod}_{A},\operatorname{LMod}_{A}^{\leq0}) ( LMod A , LMod A ≤ 0 ) belongs to P r t ± \mathrm{Pr}^{t\pm} Pr t ± . Thus, we can define P r A t ± = LMod LMod A ( S p ) ( P r t ± ) \mathrm{Pr}^{t\pm}_{A}=\operatorname{LMod}_{\operatorname{LMod}_{A}(\mathrm{Sp})}(\mathrm{Pr}^{t\pm}) Pr A t ± = LMod LMod A ( Sp ) ( Pr t ± ) , 2 P r A t ± = LMod LMod LMod A ( S p ) ( P r t ± ) ( 2 P r t ± ) 2\mathrm{Pr}^{t\pm}_{A}=\operatorname{LMod}_{\operatorname{LMod}_{\operatorname{LMod}_{A}(\mathrm{Sp})}(\mathrm{Pr}^{t\pm})}(2\mathrm{Pr}^{t\pm}) 2 Pr A t ± = LMod LMod LMod A ( Sp ) ( Pr t ± ) ( 2 Pr t ± ) and so on. In the same way as in the setting of P r t + \mathrm{Pr}^{t+} Pr t + , we have
θ S p , n ↦ m ± : Alg n ( S p c o n n ) ⟶ Alg n − m − 1 ( ( m + 1 ) P r t ± ) \theta^{\pm}_{\mathrm{Sp},n\mapsto m}:\operatorname{Alg}_{n}(\mathrm{Sp}^{\mathrm{conn}})\longrightarrow\operatorname{Alg}_{n-m-1}((m+1)\mathrm{Pr}^{t\pm}) θ Sp , n ↦ m ± : Alg n ( Sp conn ) ⟶ Alg n − m − 1 (( m + 1 ) Pr t ± )
which carries A A A to m P r A t ± m\mathrm{Pr}^{t\pm}_{A} m Pr A t ± (see Construction 3.30).
We can also consider a version of P r S t \mathrm{Pr}_{\mathrm{St}} Pr St : there is a functor
θ S p , n ↦ m : Alg n ( S p ) ⟶ Alg n − m − 1 ( ( m + 1 ) P r S t ) \theta_{\mathrm{Sp},n\mapsto m}\colon\operatorname{Alg}_{n}(\mathrm{Sp})\longrightarrow\operatorname{Alg}_{n-m-1}((m+1)\mathrm{Pr}_{\mathrm{St}}) θ Sp , n ↦ m : Alg n ( Sp ) ⟶ Alg n − m − 1 (( m + 1 ) Pr St )
which carries A A A to m P r A {}_{m}\mathrm{Pr}_{A} m Pr A (see Construction 3.28). In what follows, we construct these functors and discuss relations between them.
From algebras to higher module categories
Construction 3.20. Let C ⊗ ∈ CAlg ( Cat ^ ∞ ) \mathcal{C}^{\otimes}\in\operatorname{CAlg}(\widehat{\operatorname{Cat}}_{\infty}) C ⊗ ∈ CAlg ( Cat ∞ ) be a symmetric monoidal ∞ \infty ∞ -category which has U \mathbb{U} U -small colimits, and the tensor product functor preserves U \mathbb{U} U -small colimits separately in each variable. To A ∈ Alg 1 ( C ) A\in\operatorname{Alg}_{1}(\mathcal{C}) A ∈ Alg 1 ( C ) we associate the ∞ \infty ∞ -category LMod A ( C ) \operatorname{LMod}_{A}(\mathcal{C}) LMod A ( C ) . Note that LMod A ( C ) \operatorname{LMod}_{A}(\mathcal{C}) LMod A ( C ) has a C ⊗ \mathcal{C}^{\otimes} C ⊗ -module structure. The assignment A ↦ LMod A ( C ) A\mapsto\operatorname{LMod}_{A}(\mathcal{C}) A ↦ LMod A ( C ) is functorial: There exists a functor
θ C : Alg 1 ( C ) ⟶ Mod C ( Cat ^ ∞ ) C / \theta_{\mathcal{C}}\colon\operatorname{Alg}_{1}(\mathcal{C})\longrightarrow\operatorname{Mod}_{\mathcal{C}}(\widehat{\operatorname{Cat}}_{\infty})_{\mathcal{C}/} θ C : Alg 1 ( C ) ⟶ Mod C ( Cat ∞ ) C / which carries A A A to LMod A ( C ) \operatorname{LMod}_{A}(\mathcal{C}) LMod A ( C ) equipped with an essentially unique C \mathcal{C} C -module map C → LMod A ( C ) \mathcal{C}\to\operatorname{LMod}_{A}(\mathcal{C}) C → LMod A ( C ) sending the unit of C \mathcal{C} C to the left A A A -module A A A (see [24 4.8.3] ; right modules are adopted). A morphism A → A ′ A\to A' A → A ′ in Alg 1 ( C ) \operatorname{Alg}_{1}(\mathcal{C}) Alg 1 ( C ) maps to the base change functor LMod A ( C ) → LMod A ′ ( C ) \operatorname{LMod}_{A}(\mathcal{C})\to\operatorname{LMod}_{A'}(\mathcal{C}) LMod A ( C ) → LMod A ′ ( C ) .
According to [24 4.8.5.5, 4.8.5.16] ,
θ C \theta_{\mathcal{C}} θ C is fully faithful,
θ C \theta_{\mathcal{C}} θ C is promoted to a symmetric monoidal functor.
Note that n C n\mathcal{C} n C belongs to CAlg ( Cat ^ ∞ ) \operatorname{CAlg}(\widehat{\operatorname{Cat}}_{\infty}) CAlg ( Cat ∞ ) . We consider the sequence of the symmetric monoidal functors
Alg 1 ( n C ) → θ n C Mod n C ( Cat ^ ∞ ) n C / ⟶ Mod n C ( Cat ^ ∞ ) \operatorname{Alg}_{1}(n\mathcal{C})\xrightarrow{\theta_{n\mathcal{C}}}\operatorname{Mod}_{n\mathcal{C}}(\widehat{\operatorname{Cat}}_{\infty})_{n\mathcal{C}/}\longrightarrow\operatorname{Mod}_{n\mathcal{C}}(\widehat{\operatorname{Cat}}_{\infty}) Alg 1 ( n C ) θ n C Mod n C ( Cat ∞ ) n C / ⟶ Mod n C ( Cat ∞ ) such that the right one is the forgetful functor. By abuse of notation, we usually write θ n C \theta_{n\mathcal{C}} θ n C for the composite. This composite carries A ∈ Alg 1 ( n C ) A\in\operatorname{Alg}_{1}(n\mathcal{C}) A ∈ Alg 1 ( n C ) to LMod A ( n C ) \operatorname{LMod}_{A}(n\mathcal{C}) LMod A ( n C ) . LMod A ( n C ) \operatorname{LMod}_{A}(n\mathcal{C}) LMod A ( n C ) is dualizable, and its dual is given by RMod A ( n C ) = LMod A o p p ( n C ) \operatorname{RMod}_{A}(n\mathcal{C})=\operatorname{LMod}_{A^{\mathrm{opp}}}(n\mathcal{C}) RMod A ( n C ) = LMod A opp ( n C ) . Here A o p p A^{\mathrm{opp}} A opp means the opposite algebra of A A A . This implies that LMod A ( n C ) \operatorname{LMod}_{A}(n\mathcal{C}) LMod A ( n C ) is κ 0 \kappa_{0} κ 0 -compact (cf. [30 Proposition 5.1.10] ). Thus, we have symmetric monoidal functors
θ n C : Alg 1 ( n C ) ⟶ ( ( n + 1 ) C ) n C / ⟶ ( n + 1 ) C . \theta_{n\mathcal{C}}\colon\operatorname{Alg}_{1}(n\mathcal{C})\longrightarrow((n+1)\mathcal{C})_{n\mathcal{C}/}\longrightarrow(n+1)\mathcal{C}. θ n C : Alg 1 ( n C ) ⟶ (( n + 1 ) C ) n C / ⟶ ( n + 1 ) C . such that the left one is fully faithful, and the right one is the forgetful functor. By Dunn additivity theorem there is a canonical equivalence Alg m + 1 ( C ) ≃ Alg m ( Alg 1 ( C ) ) \operatorname{Alg}_{m+1}(\mathcal{C})\simeq\operatorname{Alg}_{m}(\operatorname{Alg}_{1}(\mathcal{C})) Alg m + 1 ( C ) ≃ Alg m ( Alg 1 ( C )) . Using this equivalence and θ n C \theta_{n\mathcal{C}} θ n C we have
θ n C , m : Alg m + 1 ( n C ) ≃ Alg m ( Alg 1 ( n C ) ) ⟶ Alg m ( ( n + 1 ) C ) . \theta_{n\mathcal{C},m}\colon\operatorname{Alg}_{m+1}(n\mathcal{C})\simeq\operatorname{Alg}_{m}(\operatorname{Alg}_{1}(n\mathcal{C}))\longrightarrow\operatorname{Alg}_{m}((n+1)\mathcal{C}). θ n C , m : Alg m + 1 ( n C ) ≃ Alg m ( Alg 1 ( n C )) ⟶ Alg m (( n + 1 ) C ) .
Our primary interest lies in the case n P r S t , n P r t + ∈ CAlg ( Cat ^ ∞ ) n\mathrm{Pr}_{\mathrm{St}},n\mathrm{Pr}_{\mathrm{t}+}\in\operatorname{CAlg}(\widehat{\operatorname{Cat}}_{\infty}) n Pr St , n Pr t + ∈ CAlg ( Cat ∞ ) :
Definition 3.21. Take C ⊗ \mathcal{C}^{\otimes} C ⊗ to be n P r S t n\mathrm{Pr}_{\mathrm{St}} n Pr St . We have symmetric monoidal functors
θ n P r S t : Alg 1 ( n P r S t ) ⟶ ( ( n + 1 ) P r S t ) n P r S t / ⟶ ( n + 1 ) P r S t \theta_{n\mathrm{Pr}_{\mathrm{St}}}\colon\operatorname{Alg}_{1}(n\mathrm{Pr}_{\mathrm{St}})\longrightarrow((n+1)\mathrm{Pr}_{\mathrm{St}})_{n\mathrm{Pr}_{\mathrm{St}}/}\longrightarrow(n+1)\mathrm{Pr}_{\mathrm{St}} θ n Pr St : Alg 1 ( n Pr St ) ⟶ (( n + 1 ) Pr St ) n Pr St / ⟶ ( n + 1 ) Pr St such that the left one is fully faithful, and the right one is the forgetful functor. This induces
θ n P r S t , m : Alg m + 1 ( n P r S t ) ⟶ Alg m ( ( n + 1 ) P r S t ) . \theta_{n\mathrm{Pr}_{\mathrm{St}},m}\colon\operatorname{Alg}_{m+1}(n\mathrm{Pr}_{\mathrm{St}})\longrightarrow\operatorname{Alg}_{m}((n+1)\mathrm{Pr}_{\mathrm{St}}). θ n Pr St , m : Alg m + 1 ( n Pr St ) ⟶ Alg m (( n + 1 ) Pr St ) .
Definition 3.22. Take C ⊗ \mathcal{C}^{\otimes} C ⊗ to be n P r t + n\mathrm{Pr}_{\mathrm{t}+} n Pr t + . We have symmetric monoidal functors
θ n P r t + : Alg 1 ( n P r t + ) ⟶ ( ( n + 1 ) P r t + ) n P r t + / ⟶ ( n + 1 ) P r t + \theta_{n\mathrm{Pr}_{\mathrm{t}+}}\colon\operatorname{Alg}_{1}(n\mathrm{Pr}_{\mathrm{t}+})\longrightarrow((n+1)\mathrm{Pr}_{\mathrm{t}+})_{n\mathrm{Pr}_{\mathrm{t}+}/}\longrightarrow(n+1)\mathrm{Pr}_{\mathrm{t}+} θ n Pr t + : Alg 1 ( n Pr t + ) ⟶ (( n + 1 ) Pr t + ) n Pr t + / ⟶ ( n + 1 ) Pr t + such that the left one is fully faithful, and the right one is the forgetful functor. This induces
θ n P r t + , m : Alg m + 1 ( n P r t + ) ⟶ Alg m ( ( n + 1 ) P r t + ) . \theta_{n\mathrm{Pr}_{\mathrm{t}+},m}\colon\operatorname{Alg}_{m+1}(n\mathrm{Pr}_{\mathrm{t}+})\longrightarrow\operatorname{Alg}_{m}((n+1)\mathrm{Pr}_{\mathrm{t}+}). θ n Pr t + , m : Alg m + 1 ( n Pr t + ) ⟶ Alg m (( n + 1 ) Pr t + ) .
Remark 3.23. We consider θ n P r S t , m : Alg m + 1 ( n P r S t ) ⟶ Alg m ( ( n + 1 ) P r S t ) \theta_{n\mathrm{Pr}_{\mathrm{St}},m}\colon\operatorname{Alg}_{m+1}(n\mathrm{Pr}_{\mathrm{St}})\longrightarrow\operatorname{Alg}_{m}((n+1)\mathrm{Pr}_{\mathrm{St}}) θ n Pr St , m : Alg m + 1 ( n Pr St ) ⟶ Alg m (( n + 1 ) Pr St ) . The symmetric monoidal functors λ n : n P r t + → n P r S t \lambda_{n}\colon n\mathrm{Pr}_{\mathrm{t}+}\to n\mathrm{Pr}_{\mathrm{St}} λ n : n Pr t + → n Pr St and λ n + 1 : ( n + 1 ) P r t + → ( n + 1 ) P r S t \lambda_{n+1}\colon(n+1)\mathrm{Pr}_{\mathrm{t}+}\to(n+1)\mathrm{Pr}_{\mathrm{St}} λ n + 1 : ( n + 1 ) Pr t + → ( n + 1 ) Pr St in Definition 3.19 induce the diagram
Alg m + 1 ( n P r t + ) → θ n P r t + , m Alg m ( ( n + 1 ) P r t + ) ↓ ↓ Alg m + 1 ( n P r S t ) → θ n P r S t , m Alg m ( ( n + 1 ) P r S t ) \begin{CD}
\operatorname{Alg}_{m+1}(n\mathrm{Pr}_{\mathrm{t}+}) @>{\theta_{n\mathrm{Pr}_{\mathrm{t}+},m}}>> \operatorname{Alg}_{m}((n+1)\mathrm{Pr}_{\mathrm{t}+}) \\
@VVV @VVV \\
\operatorname{Alg}_{m+1}(n\mathrm{Pr}_{\mathrm{St}}) @>{\theta_{n\mathrm{Pr}_{\mathrm{St}},m}}>> \operatorname{Alg}_{m}((n+1)\mathrm{Pr}_{\mathrm{St}})
\end{CD} Alg m + 1 ( n Pr t + ) ↓ ⏐ Alg m + 1 ( n Pr St ) θ n Pr t + , m θ n Pr St , m Alg m (( n + 1 ) Pr t + ) ↓ ⏐ Alg m (( n + 1 ) Pr St )
where the vertical functors are induced by λ n \lambda_n λ n and λ n + 1 \lambda_{n+1} λ n + 1 . There exists a canonical equivalence
LMod A ( n Prt + ) ⊗ n Prt + n PrSt ≃ LMod λ n ( A ) ( n PrSt ) \operatorname{LMod}_{A}(n\operatorname{Prt}^{+})\mathbin{\otimes}_{n\operatorname{Prt}^{+}}n\operatorname{PrSt}\simeq\operatorname{LMod}_{\lambda_n(A)}(n\operatorname{PrSt}) LMod A ( n Prt + ) ⊗ n Prt + n PrSt ≃ LMod λ n ( A ) ( n PrSt )
(cf. [24 4.8.4.6] ) so that the above diagram commutes up to canonical homotopy.
Definition 3.24. If we take C ⊗ \mathcal{C}^{\otimes} C ⊗ to be n Prt ± n\operatorname{Prt}^{\pm} n Prt ± , we then obtain a symmetric monoidal functor
θ n Prt ± : Alg 1 ( n Prt ± ) ⟶ ( ( n + 1 ) Prt ± ) n Prt ± / ⟶ ( n + 1 ) Prt ± \theta_{n\operatorname{Prt}^{\pm}}:\operatorname{Alg}_1(n\operatorname{Prt}^{\pm})\longrightarrow((n+1)\operatorname{Prt}^{\pm})_{n\operatorname{Prt}^{\pm}/}\longrightarrow(n+1)\operatorname{Prt}^{\pm} θ n Prt ± : Alg 1 ( n Prt ± ) ⟶ (( n + 1 ) Prt ± ) n Prt ± / ⟶ ( n + 1 ) Prt ± such that the left one is fully faithful, and the right one is the forgetful functor. We also have
θ n Prt ± , m : Alg m + 1 ( n Prt ± ) ⟶ Alg m ( ( n + 1 ) Prt ± ) . \theta_{n\operatorname{Prt}^{\pm},m}:\operatorname{Alg}_{m+1}(n\operatorname{Prt}^{\pm})\longrightarrow\operatorname{Alg}_{m}((n+1)\operatorname{Prt}^{\pm}). θ n Prt ± , m : Alg m + 1 ( n Prt ± ) ⟶ Alg m (( n + 1 ) Prt ± ) .
Construction 3.26. We will construct
θ Sp + : Alg 1 ( Sp ) ⟶ ( Prt + ) ( Sp , Sp c o n n ) / , \theta_{\operatorname{Sp}}^{+}:\operatorname{Alg}_1(\operatorname{Sp})\longrightarrow(\operatorname{Prt}^{+})_{(\operatorname{Sp},\operatorname{Sp}^{\mathrm{conn}})/}, θ Sp + : Alg 1 ( Sp ) ⟶ ( Prt + ) ( Sp , Sp conn ) / , which carries A A A to ( LMod A , LMod A ≤ 0 ) (\operatorname{LMod}_{A},\operatorname{LMod}_{A}^{\leq0}) ( LMod A , LMod A ≤ 0 ) , where LMod A ≤ 0 \operatorname{LMod}_{A}^{\leq0} LMod A ≤ 0 is the smallest full subcategory which contains A A A and is closed under small colimits and extensions. Recall that θ Sp : Alg 1 ( Sp ) → ( PrSt ) Sp / \theta_{\operatorname{Sp}}:\operatorname{Alg}_1(\operatorname{Sp})\to(\operatorname{PrSt})_{\operatorname{Sp}/} θ Sp : Alg 1 ( Sp ) → ( PrSt ) Sp / is fully faithful, we denote the essential image by T T T . Let S S S be the full subcategory of ( Prt + ) ( Sp , Sp c o n n ) / (\operatorname{Prt}^{+})_{(\operatorname{Sp},\operatorname{Sp}^{\mathrm{conn}})/} ( Prt + ) ( Sp , Sp conn ) / spanned by objects of the form ( LMod A , LMod A ≤ 0 ) (\operatorname{LMod}_{A},\operatorname{LMod}_{A}^{\leq0}) ( LMod A , LMod A ≤ 0 ) with A ∈ Alg 1 ( Sp ) A\in\operatorname{Alg}_1(\operatorname{Sp}) A ∈ Alg 1 ( Sp ) . The forgetful functor Prt + → PrSt \operatorname{Prt}^{+}\to\operatorname{PrSt} Prt + → PrSt induces f : S → T f:S\to T f : S → T . We claim that f f f is an equivalence. We define θ Sp + \theta_{\operatorname{Sp}}^{+} θ Sp + to be the composition g ∘ θ Sp g\circ\theta_{\operatorname{Sp}} g ∘ θ Sp where g g g is an inverse functor of f f f . Clearly, f f f is essentially surjective. Note that every functor F : LMod A → LMod B F:\operatorname{LMod}_{A}\to\operatorname{LMod}_{B} F : LMod A → LMod B in ( PrSt ) Sp / (\operatorname{PrSt})_{\operatorname{Sp}/} ( PrSt ) Sp / sends A A A to B B B . Since LMod A ≤ 0 \operatorname{LMod}_{A}^{\leq0} LMod A ≤ 0 (resp. LMod B ≤ 0 \operatorname{LMod}_{B}^{\leq0} LMod B ≤ 0 ) is the smallest full subcategory of LMod A \operatorname{LMod}_{A} LMod A (resp. LMod B \operatorname{LMod}_{B} LMod B ) which contains A A A (resp. B B B ) and is closed under small colimits and extensions, it follows that F F F is right t t t -exact. Thus, f f f is fully faithful. Since θ Sp \theta_{\operatorname{Sp}} θ Sp and Prt + → PrSt \operatorname{Prt}^{+}\to\operatorname{PrSt} Prt + → PrSt are promoted to symmetric monoidal functors, thus θ Sp + \theta_{\operatorname{Sp}}^{+} θ Sp + is also promoted to a symmetric monoidal functor.
Suppose that k k k is a commutative ring spectrum. If we replace Sp \operatorname{Sp} Sp with Mod k \operatorname{Mod}_{k} Mod k , then θ Sp + \theta_{\operatorname{Sp}}^{+} θ Sp + is generalized to
θ k + : Alg 1 ( Mod k ) ⟶ ( Prt + ) ( Mod k , Mod k ≤ 0 ) / \theta_{k}^{+}:\operatorname{Alg}_1(\operatorname{Mod}_{k})\longrightarrow(\operatorname{Prt}^{+})_{(\operatorname{Mod}_{k},\operatorname{Mod}_{k}^{\leq0})/} θ k + : Alg 1 ( Mod k ) ⟶ ( Prt + ) ( Mod k , Mod k ≤ 0 ) / in the same way.
If A A A is connective, then ( LMod A , LMod A ≤ 0 ) (\operatorname{LMod}_{A},\operatorname{LMod}_{A}^{\leq0}) ( LMod A , LMod A ≤ 0 ) is a both left and right complete t t t -structure (Example 3.2) so that ( LMod A , LMod A ≤ 0 ) (\operatorname{LMod}_{A},\operatorname{LMod}_{A}^{\leq0}) ( LMod A , LMod A ≤ 0 ) belongs to ( Prt ± ) ( Sp , Sp c o n n ) / (\operatorname{Prt}^{\pm})_{(\operatorname{Sp},\operatorname{Sp}^{\mathrm{conn}})/} ( Prt ± ) ( Sp , Sp conn ) / . We define
θ Sp ± : Alg 1 ( Sp c o n n ) ⟶ ( Prt ± ) ( Sp , Sp c o n n ) / \theta_{\operatorname{Sp}}^{\pm}:\operatorname{Alg}_1(\operatorname{Sp}^{\mathrm{conn}})\longrightarrow(\operatorname{Prt}^{\pm})_{(\operatorname{Sp},\operatorname{Sp}^{\mathrm{conn}})/} θ Sp ± : Alg 1 ( Sp conn ) ⟶ ( Prt ± ) ( Sp , Sp conn ) / to be the restriction of θ Sp + \theta_{\operatorname{Sp}}^{+} θ Sp + . For a connective commutative ring spectrum k k k , we also have
θ k ± : Alg 1 ( Mod k ≤ 0 ) ⟶ ( Prt ± ) ( Mod k , Mod k ≤ 0 ) / \theta_{k}^{\pm}:\operatorname{Alg}_1(\operatorname{Mod}_{k}^{\leq0})\longrightarrow(\operatorname{Prt}^{\pm})_{(\operatorname{Mod}_{k},\operatorname{Mod}_{k}^{\leq0})/} θ k ± : Alg 1 ( Mod k ≤ 0 ) ⟶ ( Prt ± ) ( Mod k , Mod k ≤ 0 ) / in the same way.
is the whole category. In particular, right completion does not change the image in Alg 1 ( Sp ) \operatorname{Alg}_{1}(\operatorname{Sp}) Alg 1 ( Sp ) under the right adjoint functor ( Pr t ) ( Sp , Sp c o n n ) / → Alg 1 ( Mod k ) (\operatorname{Pr}^{\mathrm{t}})_{(\operatorname{Sp},\operatorname{Sp}^{\mathrm{conn}})/} \to\operatorname{Alg}_{1}(\operatorname{Mod}_{k}) ( Pr t ) ( Sp , Sp conn ) / → Alg 1 ( Mod k ) .
Construction 3.28. For m 1 + m 2 = m + 1 m_{1}+m_{2}=m+1 m 1 + m 2 = m + 1 , 1 ≤ m 2 ≤ m + 1 1\leq m_{2}\leq m+1 1 ≤ m 2 ≤ m + 1 , using θ n Pr t + , m \theta_{n}^{\operatorname{Pr}^{\mathrm{t}+},m} θ n Pr t + , m we consider the sequence of symmetric monoidal fully faithful functors
Alg m + 1 ( Sp ) ⟶ Alg m ( ( Pr t + ) ( Sp , Sp c o n n ) / ) ≃ Alg m ( Pr t + ) ⟶ Alg m − 1 ( 2 Pr t + ) ⟶ Alg m − 2 ( 3 Pr t + ) ⋯ ⟶ Alg m 1 − 1 ( ( m 2 + 1 ) Pr t + ) . \begin{aligned}
\operatorname{Alg}_{m+1}(\operatorname{Sp}) &\longrightarrow\operatorname{Alg}_{m}\left((\operatorname{Pr}^{\mathrm{t}+})_{(\operatorname{Sp},\operatorname{Sp}^{\mathrm{conn}})/}\right) \simeq\operatorname{Alg}_{m}(\operatorname{Pr}^{\mathrm{t}+})\\
&\longrightarrow\operatorname{Alg}_{m-1}(2\operatorname{Pr}^{\mathrm{t}+})\\
&\longrightarrow\operatorname{Alg}_{m-2}(3\operatorname{Pr}^{\mathrm{t}+})\\
&\cdots\\
&\longrightarrow\operatorname{Alg}_{m_{1}-1}((m_{2}+1)\operatorname{Pr}^{\mathrm{t}+}).
\end{aligned} Alg m + 1 ( Sp ) ⟶ Alg m ( ( Pr t + ) ( Sp , Sp conn ) / ) ≃ Alg m ( Pr t + ) ⟶ Alg m − 1 ( 2 Pr t + ) ⟶ Alg m − 2 ( 3 Pr t + ) ⋯ ⟶ Alg m 1 − 1 (( m 2 + 1 ) Pr t + ) . where the first arrow is induced by θ Sp + \theta_{\operatorname{Sp}}^{+} θ Sp + , and the subsequent arrows are induced by θ r Pr t + \theta_{r}^{\operatorname{Pr}^{\mathrm{t}+}} θ r Pr t + . We denote the composite functor by
θ Sp , m + 1 → m 2 + : Alg m + 1 ( Sp ) → Alg m 1 − 1 ( ( m 2 + 1 ) Pr t + ) . \theta_{\operatorname{Sp},m+1\to m_{2}}^{+}\colon\operatorname{Alg}_{m+1}(\operatorname{Sp})\to\operatorname{Alg}_{m_{1}-1}((m_{2}+1)\operatorname{Pr}^{\mathrm{t}+}). θ Sp , m + 1 → m 2 + : Alg m + 1 ( Sp ) → Alg m 1 − 1 (( m 2 + 1 ) Pr t + ) . Using θ n Pr s t , m \theta_{n}^{\operatorname{Pr}^{\mathrm{st},m}} θ n Pr st , m instead of θ n Pr t + , m \theta_{n}^{\operatorname{Pr}^{\mathrm{t}+},m} θ n Pr t + , m , we also define
θ Sp , m + 1 → m 2 : Alg m + 1 ( Sp ) → Alg m 1 − 1 ( ( m 2 + 1 ) Pr s t ) . \theta_{\operatorname{Sp},m+1\to m_{2}}\colon\operatorname{Alg}_{m+1}(\operatorname{Sp})\to\operatorname{Alg}_{m_{1}-1}((m_{2}+1)\operatorname{Pr}^{\mathrm{st}}). θ Sp , m + 1 → m 2 : Alg m + 1 ( Sp ) → Alg m 1 − 1 (( m 2 + 1 ) Pr st ) . Note that θ k , m + 1 → m 2 + \theta_{k,m+1\to m_{2}}^{+} θ k , m + 1 → m 2 + and θ k , m + 1 → m 2 \theta_{k,m+1\to m_{2}} θ k , m + 1 → m 2 are fully faithful.
For A ∈ Alg m + 1 ( Sp ) A\in\operatorname{Alg}_{m+1}(\operatorname{Sp}) A ∈ Alg m + 1 ( Sp ) we write m 2 Pr A t + {}_{m_{2}}\operatorname{Pr}_{A}^{\mathrm{t}+} m 2 Pr A t + for θ Sp , m + 1 → m 2 + ( A ) \theta_{\operatorname{Sp},m+1\to m_{2}}^{+}(A) θ Sp , m + 1 → m 2 + ( A ) . For A ∈ Alg m + 1 ( Sp ) A\in\operatorname{Alg}_{m+1}(\operatorname{Sp}) A ∈ Alg m + 1 ( Sp ) we write m 2 Pr A {}_{m_{2}}\operatorname{Pr}_{A} m 2 Pr A for θ Sp , m + 1 → m 2 ( A ) \theta_{\operatorname{Sp},m+1\to m_{2}}(A) θ Sp , m + 1 → m 2 ( A ) . For ease of notation, we usually omit the monoidal structure from the notation.
Construction 3.30. We consider the sequence of symmetric monoidal fully faithful functors
Alg m + 1 ( Sp c o n n ) ⟶ Alg m ( ( Pr t ± ) ( Sp , Sp c o n n ) / ) ≃ Alg m ( Pr t ± ) ⟶ Alg m − 1 ( 2 Pr t ± ) ⟶ Alg m − 2 ( 3 Pr t ± ) ⋯ ⟶ Alg m 1 − 1 ( ( m 2 + 1 ) Pr t ± ) . \begin{aligned}
\operatorname{Alg}_{m+1}(\operatorname{Sp}^{\mathrm{conn}})&\longrightarrow\operatorname{Alg}_{m}\left((\operatorname{Pr}^{\mathrm{t}\pm})_{(\operatorname{Sp},\operatorname{Sp}^{\mathrm{conn}})/}\right)\simeq\operatorname{Alg}_{m}(\operatorname{Pr}^{\mathrm{t}\pm})\\
&\longrightarrow\operatorname{Alg}_{m-1}(2\operatorname{Pr}^{\mathrm{t}\pm})\\
&\longrightarrow\operatorname{Alg}_{m-2}(3\operatorname{Pr}^{\mathrm{t}\pm})\\
&\cdots\\
&\longrightarrow\operatorname{Alg}_{m_{1}-1}((m_{2}+1)\operatorname{Pr}^{\mathrm{t}\pm}).
\end{aligned} Alg m + 1 ( Sp conn ) ⟶ Alg m ( ( Pr t ± ) ( Sp , Sp conn ) / ) ≃ Alg m ( Pr t ± ) ⟶ Alg m − 1 ( 2 Pr t ± ) ⟶ Alg m − 2 ( 3 Pr t ± ) ⋯ ⟶ Alg m 1 − 1 (( m 2 + 1 ) Pr t ± ) . where the first arrow is induced by θ Sp ± \theta_{\operatorname{Sp}}^{\pm} θ Sp ± , and the subsequent arrows are induced by θ r Pr t ± \theta_{r}^{\operatorname{Pr}^{\mathrm{t}\pm}} θ r Pr t ± . We denote the composite fully faithful functor by
θ Sp , m + 1 → m 2 ± : Alg m + 1 ( Sp c o n n ) → Alg m 1 − 1 ( ( m 2 + 1 ) Pr t ± ) . \theta_{\operatorname{Sp},m+1\to m_{2}}^{\pm}\colon\operatorname{Alg}_{m+1}(\operatorname{Sp}^{\mathrm{conn}})\to\operatorname{Alg}_{m_{1}-1}((m_{2}+1)\operatorname{Pr}^{\mathrm{t}\pm}). θ Sp , m + 1 → m 2 ± : Alg m + 1 ( Sp conn ) → Alg m 1 − 1 (( m 2 + 1 ) Pr t ± ) . For A ∈ Alg m + 1 ( Sp c o n n ) A\in\operatorname{Alg}_{m+1}(\operatorname{Sp}^{\mathrm{conn}}) A ∈ Alg m + 1 ( Sp conn ) we write m 2 Pr t ± {}_{m_{2}}\operatorname{Pr}^{\mathrm{t}\pm} m 2 Pr t ± for θ Sp , m + 1 → m 2 ± ( A ) \theta_{\operatorname{Sp},m+1\to m_{2}}^{\pm}(A) θ Sp , m + 1 → m 2 ± ( A ) .
k k k -linear contexts
Let k k k be a connective commutative ring spectrum. We note that ( Mod k , Mod k ≤ 0 ) (\operatorname{Mod}_{k},\operatorname{Mod}_{k}^{\leq0}) ( Mod k , Mod k ≤ 0 ) belongs to CAlg ( Pr t ± ) \operatorname{CAlg}(\operatorname{Pr}^{\mathrm{t}\pm}) CAlg ( Pr t ± ) . We set
Pr k t + : = Mod ( Mod k , Mod k ≤ 0 ) ( Pr t + ) and Pr k t ± : = Mod ( Mod k , Mod k ≤ 0 ) ( Pr t ± ) . \operatorname{Pr}_{k}^{\mathrm{t}+}:=\operatorname{Mod}_{(\operatorname{Mod}_{k},\operatorname{Mod}_{k}^{\leq0})}(\operatorname{Pr}^{\mathrm{t}+})
\quad\text{and}\quad
\operatorname{Pr}_{k}^{\mathrm{t}\pm}:=\operatorname{Mod}_{(\operatorname{Mod}_{k},\operatorname{Mod}_{k}^{\leq0})}(\operatorname{Pr}^{\mathrm{t}\pm}). Pr k t + := Mod ( Mod k , Mod k ≤ 0 ) ( Pr t + ) and Pr k t ± := Mod ( Mod k , Mod k ≤ 0 ) ( Pr t ± ) .
We take C C C to be Pr k t + \operatorname{Pr}_{k}^{\mathrm{t}+} Pr k t + in Construction 3.20 instead of Pr t + \operatorname{Pr}^{\mathrm{t}+} Pr t + , then we get n Pr k t + ∈ CAlg ( Cat ^ † ) n\operatorname{Pr}_{k}^{\mathrm{t}+}\in\operatorname{CAlg}(\widehat{\operatorname{Cat}}^{\dagger}) n Pr k t + ∈ CAlg ( Cat † ) . Similarly, if C C C is Pr k t ± \operatorname{Pr}_{k}^{\mathrm{t}\pm} Pr k t ± , we have n Pr k t ± ∈ CAlg ( Cat ^ † ) n\operatorname{Pr}_{k}^{\mathrm{t}\pm}\in\operatorname{CAlg}(\widehat{\operatorname{Cat}}^{\dagger}) n Pr k t ± ∈ CAlg ( Cat † ) .
We have symmetric monoidal functors
θ n P r t k + : Alg 1 ( n P r t k + ) ⟶ ( ( n + 1 ) P r t k + ) n P r t k + / ⟶ ( n + 1 ) P r t k + \theta_{n\mathrm{Prt}^{+}_{k}}:\operatorname{Alg}_{1}(n\mathrm{Prt}^{+}_{k})\longrightarrow((n+1)\mathrm{Prt}^{+}_{k})_{n\mathrm{Prt}^{+}_{k}/}\longrightarrow(n+1)\mathrm{Prt}^{+}_{k} θ n Prt k + : Alg 1 ( n Prt k + ) ⟶ (( n + 1 ) Prt k + ) n Prt k + / ⟶ ( n + 1 ) Prt k +
which induces
θ n P r t k + , m : Alg m + 1 ( n P r t k + ) ⟶ Alg m ( ( n + 1 ) P r t k + ) \theta_{n\mathrm{Prt}^{+}_{k},m}:\operatorname{Alg}_{m+1}(n\mathrm{Prt}^{+}_{k})\longrightarrow\operatorname{Alg}_{m}((n+1)\mathrm{Prt}^{+}_{k}) θ n Prt k + , m : Alg m + 1 ( n Prt k + ) ⟶ Alg m (( n + 1 ) Prt k + )
for m ≥ 0 m \ge0 m ≥ 0 .
We define θ n P r t k ± : Alg 1 ( n P r t k ± ) ⟶ ( ( n + 1 ) P r t k ± ) n P r t k ± / \theta_{n\mathrm{Prt}^{\pm}_{k}}:\operatorname{Alg}_{1}(n\mathrm{Prt}^{\pm}_{k})\longrightarrow((n+1)\mathrm{Prt}^{\pm}_{k})_{n\mathrm{Prt}^{\pm}_{k}/} θ n Prt k ± : Alg 1 ( n Prt k ± ) ⟶ (( n + 1 ) Prt k ± ) n Prt k ± / and θ n P r t k ± , m : Alg m + 1 ( n P r t k ± ) ⟶ Alg m ( ( n + 1 ) P r t k ± ) \theta_{n\mathrm{Prt}^{\pm}_{k},m}:\operatorname{Alg}_{m+1}(n\mathrm{Prt}^{\pm}_{k})\longrightarrow\operatorname{Alg}_{m}((n+1)\mathrm{Prt}^{\pm}_{k}) θ n Prt k ± , m : Alg m + 1 ( n Prt k ± ) ⟶ Alg m (( n + 1 ) Prt k ± ) in the same way. Using them, we also define
θ k , m + 1 → m 2 + : Alg m + 1 ( Mod k ) ⟶ Alg m 1 − 1 ( ( m 2 + 1 ) P r t k + ) . \theta^{+}_{k,m+1\to m_{2}}:\operatorname{Alg}_{m+1}(\operatorname{Mod}_{k})\longrightarrow\operatorname{Alg}_{m_{1}-1}((m_{2}+1)\mathrm{Prt}^{+}_{k}). θ k , m + 1 → m 2 + : Alg m + 1 ( Mod k ) ⟶ Alg m 1 − 1 (( m 2 + 1 ) Prt k + ) .
and
θ k , m + 1 → m 2 ± : Alg m + 1 ( Mod k ≤ 0 ) ⟶ Alg m 1 − 1 ( ( m 2 + 1 ) P r t k ± ) . \theta^{\pm}_{k,m+1\to m_{2}}:\operatorname{Alg}_{m+1}(\operatorname{Mod}^{\leq0}_{k})\longrightarrow\operatorname{Alg}_{m_{1}-1}((m_{2}+1)\mathrm{Prt}^{\pm}_{k}). θ k , m + 1 → m 2 ± : Alg m + 1 ( Mod k ≤ 0 ) ⟶ Alg m 1 − 1 (( m 2 + 1 ) Prt k ± ) .
in a similar way to θ S p , m + 1 → m 2 + \theta^{+}_{\mathrm{Sp},m+1\to m_{2}} θ Sp , m + 1 → m 2 + and θ S p , m + 1 → m 2 ± \theta^{\pm}_{\mathrm{Sp},m+1\to m_{2}} θ Sp , m + 1 → m 2 ± . Similarly, in the setting of P r k \mathrm{Pr}_{k} Pr k we have
θ k , m + 1 → m 2 : Alg m + 1 ( Mod k ) ⟶ Alg m 1 − 1 ( ( m 2 + 1 ) P r k ) . \theta_{k,m+1\to m_{2}}:\operatorname{Alg}_{m+1}(\operatorname{Mod}_{k})\longrightarrow\operatorname{Alg}_{m_{1}-1}((m_{2}+1)\mathrm{Pr}_{k}). θ k , m + 1 → m 2 : Alg m + 1 ( Mod k ) ⟶ Alg m 1 − 1 (( m 2 + 1 ) Pr k ) .
The functors θ k , m + 1 → m 2 + \theta^{+}_{k,m+1\to m_{2}} θ k , m + 1 → m 2 + , θ k , m + 1 → m 2 ± \theta^{\pm}_{k,m+1\to m_{2}} θ k , m + 1 → m 2 ± and θ k , m + 1 → m 2 \theta_{k,m+1\to m_{2}} θ k , m + 1 → m 2 are (promoted to) symmetric monoidal functors and are fully faithful.
In the k k k -linear setting, we continue to use the notation m 2 P r t A + {}_{m_{2}}\mathrm{Prt}^{+}_{A} m 2 Prt A + , m 2 P r t A ± {}_{m_{2}}\mathrm{Prt}^{\pm}_{A} m 2 Prt A ± and m 2 P r A {}_{m_{2}}\mathrm{Pr}_{A} m 2 Pr A for θ k , m + 1 → m 2 ( A ) \theta_{k,m+1\to m_{2}}(A) θ k , m + 1 → m 2 ( A ) , θ k , m + 1 → m 2 ± ( A ) \theta^{\pm}_{k,m+1\to m_{2}}(A) θ k , m + 1 → m 2 ± ( A ) and θ k , m + 1 → m 2 ( A ) \theta_{k,m+1\to m_{2}}(A) θ k , m + 1 → m 2 ( A ) , respectively. For instance, 2 P r t A + {}_{2}\mathrm{Prt}^{+}_{A} 2 Prt A + is defined to be
LMod LMod LMod A ( P r t k + ) ( 2 P r t k + ) . \operatorname{LMod}_{\operatorname{LMod}_{\operatorname{LMod}_{A}}(\mathrm{Prt}^{+}_{k})}(2\mathrm{Prt}^{+}_{k}). LMod LMod LMod A ( Prt k + ) ( 2 Prt k + ) .
This notation overlaps with those in Construction 3.28 and Construction 3.30. However, this is harmless. Indeed, using definitions and [24 7.1.3.1] , one sees that m 2 P r t A + {}_{m_{2}}\mathrm{Prt}^{+}_{A} m 2 Prt A + , m 2 P r t A ± {}_{m_{2}}\mathrm{Prt}^{\pm}_{A} m 2 Prt A ± and m 2 P r A {}_{m_{2}}\mathrm{Pr}_{A} m 2 Pr A in this k k k -linear setting are equivalent to those in Construction 3.28 and Construction 3.30. By default, in the subsequent sections, we adopt m 2 P r t A + {}_{m_{2}}\mathrm{Prt}^{+}_{A} m 2 Prt A + , m 2 P r t A ± {}_{m_{2}}\mathrm{Prt}^{\pm}_{A} m 2 Prt A ± and m 2 P r A {}_{m_{2}}\mathrm{Pr}_{A} m 2 Pr A in the k k k -linear setting.
Higher categorified Koszul duality
Let k k k be a base field.
Algebraization
We introduce the notion of algebraization and discuss a relation with Koszul duals of E n \mathrm{E}_{n} E n -algebras.
Definition 4.1. Let n ≥ 0 n \ge0 n ≥ 0 be a nonnegative integer. Let P be any of P r k \mathrm{Pr}_{k} Pr k or P r t k + \mathrm{Prt}^{+}_{k} Prt k + or P r t k ± \mathrm{Prt}^{\pm}_{k} Prt k ± . Let ( L , ϕ : n P → L ) (\mathcal{L},\phi:n\mathrm{P}\to\mathcal{L}) ( L , ϕ : n P → L ) be an object in ( ( n + 1 ) P ) n P / ((n+1)\mathrm{P})_{n\mathrm{P}/} (( n + 1 ) P ) n P / . Here L \mathcal{L} L is the image of ( n − 1 ) P ∈ n P (n-1)\mathrm{P}\in n\mathrm{P} ( n − 1 ) P ∈ n P under ϕ \phi ϕ . Conversely, this L \mathcal{L} L determines ϕ \phi ϕ so that ( L , ϕ ) (\mathcal{L},\phi) ( L , ϕ ) amounts to the pair ( L , L ) (\mathcal{L},\mathcal{L}) ( L , L ) . We consider the functor j ( L , L ) : Alg n + 1 ( Mod k ) o p → S ^ j_{(\mathcal{L},\mathcal{L})}:\operatorname{Alg}_{n+1}(\operatorname{Mod}_{k})^{\mathrm{op}}\to\widehat{\mathcal{S}} j ( L , L ) : Alg n + 1 ( Mod k ) op → S given informally by A ↦ Map ( ( n + 1 ) P ) n P / ( ( n P A , ( n − 1 ) P A ) , ( L , L ) ) A\mapsto\operatorname{Map}_{((n+1)\mathrm{P})_{n\mathrm{P}/}}((n\mathrm{P}_{A},(n-1)\mathrm{P}_{A}),(\mathcal{L},\mathcal{L})) A ↦ Map (( n + 1 ) P ) n P / (( n P A , ( n − 1 ) P A ) , ( L , L )) . We regard ( L , L ) (\mathcal{L},\mathcal{L}) ( L , L ) as an object of ( ( n + 1 ) P ) n P / ((n+1)\mathrm{P})_{n\mathrm{P}/} (( n + 1 ) P ) n P / . Here P A \mathrm{P}_{A} P A is P r A \mathrm{Pr}_{A} Pr A or P r t A + \mathrm{Prt}^{+}_{A} Prt A + or P r t A ± \mathrm{Prt}^{\pm}_{A} Prt A ± , depending on the choice of P. If j ( L , L ) j_{(\mathcal{L},\mathcal{L})} j ( L , L ) is representable by some C ∈ Alg n + 1 ( Mod k ) C\in\operatorname{Alg}_{n+1}(\operatorname{Mod}_{k}) C ∈ Alg n + 1 ( Mod k ) , we say that ( L , L ) (\mathcal{L},\mathcal{L}) ( L , L ) has an algebraization with respect to P. If the context is obvious, we say that ( L , L ) (\mathcal{L},\mathcal{L}) ( L , L ) has an algebraization. Let ( n P C , ( n − 1 ) P C ) → ( L , L ) (n\mathrm{P}_{C},(n-1)\mathrm{P}_{C})\to(\mathcal{L},\mathcal{L}) ( n P C , ( n − 1 ) P C ) → ( L , L ) be the morphism that corresponds to the identity C → C C\to C C → C through the equivalence Map Alg n + 1 ( Mod k ) ( C , C ) ≃ Map ( ( n + 1 ) P ) n P / ( n P C , ( L , L ) ) \operatorname{Map}_{\operatorname{Alg}_{n+1}(\operatorname{Mod}_{k})}(C,C)\simeq\operatorname{Map}_{((n+1)\mathrm{P})_{n\mathrm{P}/}}(n\mathrm{P}_{C},(\mathcal{L},\mathcal{L})) Map Alg n + 1 ( Mod k ) ( C , C ) ≃ Map (( n + 1 ) P ) n P / ( n P C , ( L , L )) . We refer to it as the algebraization functor (with respect to P).
Notation 4.2. Let P be any of P r k \mathrm{Pr}_{k} Pr k or P r t k + \mathrm{Prt}^{+}_{k} Prt k + or P r t k ± \mathrm{Prt}^{\pm}_{k} Prt k ± . Let ( L , ϕ : n P → L ) (\mathcal{L},\phi:n\mathrm{P}\to\mathcal{L}) ( L , ϕ : n P → L ) be an object in ( ( n + 1 ) P ) n P / ((n+1)\mathrm{P})_{n\mathrm{P}/} (( n + 1 ) P ) n P / . As explained in Definition 4.1, we usually use a pair ( L , L ) (\mathcal{L},\mathcal{L}) ( L , L ) to indicate an object of ( L , ϕ : n P → L ) (\mathcal{L},\phi:n\mathrm{P}\to\mathcal{L}) ( L , ϕ : n P → L ) .
Definition 4.3. Let n ≥ 0 n \ge0 n ≥ 0 be a nonnegative integer. Let P be any of P r k \mathrm{Pr}_{k} Pr k or P r t k + \mathrm{Prt}^{+}_{k} Prt k + or P r t k ± \mathrm{Prt}^{\pm}_{k} Prt k ± . Let M ⊗ \mathcal{M}^{\otimes} M ⊗ be an E 1 \mathrm{E}_{1} E 1 -algebra object in ( n + 1 ) P (n+1)\mathrm{P} ( n + 1 ) P , that is, an object of Alg 1 ( ( n + 1 ) P ) \operatorname{Alg}_{1}((n+1)\mathrm{P}) Alg 1 (( n + 1 ) P ) . (For later use, more generally M ⊗ \mathcal{M}^{\otimes} M ⊗ is allowed to be an E 1 \mathrm{E}_{1} E 1 -algebra object in Mod P r ( C a t ^ ) \operatorname{Mod}_{\mathrm{Pr}(\widehat{\mathcal{C}at})} Mod Pr ( C a t ) .) We consider the functor j M ⊗ : Alg n + 2 ( Mod k ) o p → S ^ j_{\mathcal{M}^{\otimes}}:\operatorname{Alg}_{n+2}(\operatorname{Mod}_{k})^{\mathrm{op}}\to\widehat{\mathcal{S}} j M ⊗ : Alg n + 2 ( Mod k ) op → S given informally by A ↦ Map Alg 1 ( ( n + 1 ) P ) ( n P A , M ⊗ ) A\mapsto\operatorname{Map}_{\operatorname{Alg}_{1}((n+1)\mathrm{P})}(n\mathrm{P}_{A},\mathcal{M}^{\otimes}) A ↦ Map Alg 1 (( n + 1 ) P ) ( n P A , M ⊗ ) . Here P A \mathrm{P}_{A} P A is P r A \mathrm{Pr}_{A} Pr A or P r t A + \mathrm{Prt}^{+}_{A} Prt A + or P r t A ± \mathrm{Prt}^{\pm}_{A} Prt A ± . If j M ⊗ j_{\mathcal{M}^{\otimes}} j M ⊗ is representable by some C ∈ Alg n + 2 ( Mod k ) C\in\operatorname{Alg}_{n+2}(\operatorname{Mod}_{k}) C ∈ Alg n + 2 ( Mod k ) , we say that M ⊗ \mathcal{M}^{\otimes} M ⊗ has an algebraization. Let n P C → M ⊗ n\mathrm{P}_{C}\to\mathcal{M}^{\otimes} n P C → M ⊗ be the morphism that corresponds to the identity C → C C\to C C → C through the equivalence Map Alg n + 2 ( Mod k ) ( C , C ) ≃ Map Alg 1 ( ( n + 1 ) P ) ( n P C , M ⊗ ) \operatorname{Map}_{\operatorname{Alg}_{n+2}(\operatorname{Mod}_{k})}(C,C)\simeq\operatorname{Map}_{\operatorname{Alg}_{1}((n+1)\mathrm{P})}(n\mathrm{P}_{C},\mathcal{M}^{\otimes}) Map Alg n + 2 ( Mod k ) ( C , C ) ≃ Map Alg 1 (( n + 1 ) P ) ( n P C , M ⊗ ) . We refer to it as the algebraization functor (with respect to P).
Remark 4.4. In the setting of Definition 4.3, we suppose that M ⊗ M^{\otimes} M ⊗ be an E 1 \mathrm{E}_1 E 1 -algebra object in ( n + 1 ) P r k ± (n+1)\mathrm{Pr}^{\pm}_k ( n + 1 ) Pr k ± . If we regard M ⊗ M^{\otimes} M ⊗ as an object of P r k + \mathrm{Pr}^{+}_k Pr k + , we have
Map Alg 1 ( ( n + 1 ) P r k + ) ( n P r k , A + , M ⊗ ) ≃ Map Alg 1 ( ( n + 1 ) P r k ± ) ( n P r k , A ± , M ⊗ ) . \operatorname{Map}_{\operatorname{Alg}_1((n+1)\mathrm{Pr}^{+}_k)}(n\mathrm{Pr}^{+}_{k,A},M^{\otimes}) \simeq\operatorname{Map}_{\operatorname{Alg}_1((n+1)\mathrm{Pr}^{\pm}_k)}(n\mathrm{Pr}^{\pm}_{k,A},M^{\otimes}). Map Alg 1 (( n + 1 ) Pr k + ) ( n Pr k , A + , M ⊗ ) ≃ Map Alg 1 (( n + 1 ) Pr k ± ) ( n Pr k , A ± , M ⊗ ) .
Note that n P r k , A ± = n P r k , A + ⊗ n P r k + n P r k ± n\mathrm{Pr}^{\pm}_{k,A}=n\mathrm{Pr}^{+}_{k,A}\otimes_{n\mathrm{Pr}^{+}_k}n\mathrm{Pr}^{\pm}_k n Pr k , A ± = n Pr k , A + ⊗ n Pr k + n Pr k ± . If M ⊗ M^{\otimes} M ⊗ has an algebraization with respect to P r k + \mathrm{Pr}^{+}_k Pr k + , then M ⊗ M^{\otimes} M ⊗ has an algebraization with respect to P r k ± \mathrm{Pr}^{\pm}_k Pr k ± .
We will relate the algebraizations and Koszul duals of algebras. Let A A A be an augmented E n + 2 \mathrm{E}_{n+2} E n + 2 -algebra in Mod k \operatorname{Mod}_k Mod k . By the base change n P r k , A + → n P r k + n\mathrm{Pr}^{+}_{k,A}\to n\mathrm{Pr}^{+}_k n Pr k , A + → n Pr k + induced by the augmentation A → k A\to k A → k , we regard n P r k + n\mathrm{Pr}^{+}_k n Pr k + as a left n P r k , A + n\mathrm{Pr}^{+}_{k,A} n Pr k , A + -module. That is, n P r k + n\mathrm{Pr}^{+}_k n Pr k + is an object of LMod n P r k , A + ( ( n + 1 ) P r k + ) ≃ ( n + 1 ) P r k , A + \operatorname{LMod}_{n\mathrm{Pr}^{+}_{k,A}}((n+1)\mathrm{Pr}^{+}_k)\simeq(n+1)\mathrm{Pr}^{+}_{k,A} LMod n Pr k , A + (( n + 1 ) Pr k + ) ≃ ( n + 1 ) Pr k , A + (see Proposition 4.9 (1) for the equivalence). We consider the endomorphism algebra End n P r k , A + l ( n P r k + ) \operatorname{End}^{l}_{n\mathrm{Pr}^{+}_{k,A}}(n\mathrm{Pr}^{+}_k) End n Pr k , A + l ( n Pr k + ) of n P r k + ∈ ( n + 1 ) P r k , A + n\mathrm{Pr}^{+}_k\in(n+1)\mathrm{Pr}^{+}_{k,A} n Pr k + ∈ ( n + 1 ) Pr k , A + , defined in Alg 1 ( ( n + 1 ) P r k + ) \operatorname{Alg}_1((n+1)\mathrm{Pr}^{+}_k) Alg 1 (( n + 1 ) Pr k + ) (Lemma 4.10 for the existence). Here the superscript l l l and the subscript n P r k , A + n\mathrm{Pr}^{+}_{k,A} n Pr k , A + in End n P r k , A + l \operatorname{End}^{l}_{n\mathrm{Pr}^{+}_{k,A}} End n Pr k , A + l indicate “left n P r k , A + n\mathrm{Pr}^{+}_{k,A} n Pr k , A + -linear functors”.
Proposition 4.5. The following holds.
Let B B B be an augmented E n + 1 \mathrm{E}_{n+1} E n + 1 -algebra in Mod k \operatorname{Mod}_k Mod k . By the base change ( n − 1 ) P r k , B + → ( n − 1 ) P r k + (n-1)\mathrm{Pr}^{+}_{k,B}\to(n-1)\mathrm{Pr}^{+}_k ( n − 1 ) Pr k , B + → ( n − 1 ) Pr k + induced by the augmentation B → k B\to k B → k , we regard ( n − 1 ) P r k + (n-1)\mathrm{Pr}^{+}_k ( n − 1 ) Pr k + as a left ( n − 1 ) P r k , B + (n-1)\mathrm{Pr}^{+}_{k,B} ( n − 1 ) Pr k , B + -module. Then ( n P r k , B + , ( n − 1 ) P r k + ) (n\mathrm{Pr}^{+}_{k,B},(n-1)\mathrm{Pr}^{+}_k) ( n Pr k , B + , ( n − 1 ) Pr k + ) has an algebraization C ∈ Alg n + 1 ( Mod k ) C\in\operatorname{Alg}_{n+1}(\operatorname{Mod}_k) C ∈ Alg n + 1 ( Mod k ) . Moreover, C C C is equivalent to the E n + 1 \mathrm{E}_{n+1} E n + 1 -Koszul dual D n + 1 ( B ) D_{n+1}(B) D n + 1 ( B ) .
Let A A A be an augmented E n + 2 \mathrm{E}_{n+2} E n + 2 -algebra in Mod k \operatorname{Mod}_k Mod k . Then there exists an endomorphism algebra End n P r k , A + l ( n P r k + ) \operatorname{End}^{l}_{n\mathrm{Pr}^{+}_{k,A}}(n\mathrm{Pr}^{+}_k) End n Pr k , A + l ( n Pr k + ) , and End n P r k , A + l ( n P r k + ) \operatorname{End}^{l}_{n\mathrm{Pr}^{+}_{k,A}}(n\mathrm{Pr}^{+}_k) End n Pr k , A + l ( n Pr k + ) has an algebraization C ∈ Alg n + 2 ( Mod k ) C\in\operatorname{Alg}_{n+2}(\operatorname{Mod}_k) C ∈ Alg n + 2 ( Mod k ) . Moreover, C C C is equivalent to the E n + 2 \mathrm{E}_{n+2} E n + 2 -Koszul dual D n + 2 ( A ) D_{n+2}(A) D n + 2 ( A ) .
Remark 4.6. The E n + 2 \mathrm{E}_{n+2} E n + 2 -algebra C C C has a canonically defined augmentation p C : C → k p_C:C\to k p C : C → k , and ( C , p C ) (C,p_C) ( C , p C ) is equivalent to D n + 2 ( A ) D_{n+2}(A) D n + 2 ( A ) as an augmented E n + 2 \mathrm{E}_{n+2} E n + 2 -algebra. See the proof of Proposition 4.5.
The proof of Proposition 4.5 will require some preliminaries.
Recall that n P r k + n\mathrm{Pr}^{+}_k n Pr k + belongs to CAlg ( Cat ^ † ) \operatorname{CAlg}(\widehat{\operatorname{Cat}}_{\dagger}) CAlg ( Cat † ) . We have symmetric monoidal functors
θ n P r k + : Alg 1 ( n P r k + ) ⟶ ( ( n + 1 ) P r k + ) n P r k + / ⟶ ( n + 1 ) P r k + \theta_{n\mathrm{Pr}^{+}_k}:\operatorname{Alg}_1(n\mathrm{Pr}^{+}_k)\longrightarrow((n+1)\mathrm{Pr}^{+}_k)_{n\mathrm{Pr}^{+}_k/}\longrightarrow(n+1)\mathrm{Pr}^{+}_k θ n Pr k + : Alg 1 ( n Pr k + ) ⟶ (( n + 1 ) Pr k + ) n Pr k + / ⟶ ( n + 1 ) Pr k +
such that the left one is fully faithful, and the right one is the forgetful functor (cf. Construction 3.20). This induces a fully faithful functor
θ n P r k + , m : Alg m + 1 ( n P r k + ) ⟶ Alg m ( ( n + 1 ) P r k + ) . \theta_{n\mathrm{Pr}^{+}_k,m}:\operatorname{Alg}_{m+1}(n\mathrm{Pr}^{+}_k)\longrightarrow\operatorname{Alg}_m((n+1)\mathrm{Pr}^{+}_k). θ n Pr k + , m : Alg m + 1 ( n Pr k + ) ⟶ Alg m (( n + 1 ) Pr k + ) .
We consider whether or not θ n P r k + , m \theta_{n\mathrm{Pr}^{+}_k,m} θ n Pr k + , m has a right adjoint. From the construction θ n P r k + , m \theta_{n\mathrm{Pr}^{+}_k,m} θ n Pr k + , m is extended to
Alg m + 1 ( n P r k + ) ⟶ Alg m ( Mod n P r k + ( Cat ^ † ) ) . \operatorname{Alg}_{m+1}(n\mathrm{Pr}^{+}_k)\longrightarrow\operatorname{Alg}_m(\operatorname{Mod}_{n\mathrm{Pr}^{+}_k}(\widehat{\operatorname{Cat}}_{\dagger})). Alg m + 1 ( n Pr k + ) ⟶ Alg m ( Mod n Pr k + ( Cat † )) .
For D ∈ ( n + 1 ) P r k + \mathcal{D}\in(n+1)\mathrm{Pr}^{+}_k D ∈ ( n + 1 ) Pr k + and for D ∈ D D\in\mathcal{D} D ∈ D , its Hom (or endomorphism algebra) object Hom D ( D , D ) \operatorname{Hom}_{\mathcal{D}}(D,D) Hom D ( D , D ) exists in Mod ( n − 1 ) P r k + ( Cat ^ † ) \operatorname{Mod}_{(n-1)\mathrm{Pr}^{+}_k}(\widehat{\operatorname{Cat}}_{\dagger}) Mod ( n − 1 ) Pr k + ( Cat † ) . We can ask if Hom D ( D , D ′ ) \operatorname{Hom}_{\mathcal{D}}(D,D') Hom D ( D , D ′ ) exists in n P r k + n\mathrm{Pr}^{+}_k n Pr k + . Motivated by this question, we consider the following condition:
Definition 4.7. Let n ≥ 0 n\geq0 n ≥ 0 be a nonnegative integer. Let P \mathrm{P} P be any of P r k \mathrm{Pr}_k Pr k or P r k + \mathrm{Pr}^{+}_k Pr k + or P r k ± \mathrm{Pr}^{\pm}_k Pr k ± .
Let ( L , L ) (\mathcal{L},L) ( L , L ) be an object in ( ( n + 1 ) P ) n P / ((n+1)\mathrm{P})_{n\mathrm{P}/} (( n + 1 ) P ) n P / . If there exists an endomorphism algebra object of L L L in Alg 1 ( n P ) \operatorname{Alg}_1(n\mathrm{P}) Alg 1 ( n P ) , we say that ( L , L ) (\mathcal{L},L) ( L , L ) has an endomorphism algebra. If exists, we write E ( ( L , L ) ) \mathrm{E}((\mathcal{L},L)) E (( L , L )) for an endomorphism algebra.
Let 1 \mathbf{1} 1 denote a unit object in E ( ( L , L ) ) \mathrm{E}((\mathcal{L},L)) E (( L , L )) . Suppose that ( E ( ( L , L ) ) , 1 ) (\mathrm{E}((\mathcal{L},L)),\mathbf{1}) ( E (( L , L )) , 1 ) also has an endomorphism algebra which we denote by E 2 ( ( L , L ) ) ∈ Alg 2 ( ( n − 1 ) P ) , … , E n − 1 ( ( L , L ) ) \mathrm{E}^2((\mathcal{L},L))\in\operatorname{Alg}_2((n-1)\mathrm{P}),\ldots,\mathrm{E}^{n-1}((\mathcal{L},L)) E 2 (( L , L )) ∈ Alg 2 (( n − 1 ) P ) , … , E n − 1 (( L , L )) with a unit also has an endomorphism algebra which we denote by E n ( ( L , L ) ) ∈ Alg n ( P ) \mathrm{E}^n((\mathcal{L},L))\in\operatorname{Alg}_n(\mathrm{P}) E n (( L , L )) ∈ Alg n ( P ) . In this case, we say that ( L , L ) (\mathcal{L},L) ( L , L ) is a good endomorphism algebra.
Let M ⊗ \mathcal{M}^{\otimes} M ⊗ be an E 1 \mathrm{E}_1 E 1 -algebra object in ( n + 1 ) P (n+1)\mathrm{P} ( n + 1 ) P , that is, an object of Alg 1 ( ( n + 1 ) P ) \operatorname{Alg}_1((n+1)\mathrm{P}) Alg 1 (( n + 1 ) P ) . Let 1 M \mathbf{1}_{\mathcal{M}} 1 M denote a unit object of M ⊗ \mathcal{M}^{\otimes} M ⊗ . If ( M , 1 M ) (\mathcal{M},\mathbf{1}_{\mathcal{M}}) ( M , 1 M ) has an endomorphism algebra, we say that M ⊗ \mathcal{M}^{\otimes} M ⊗ has an endomorphism algebra of a unit. If exists, we write E ( M ⊗ ) \mathrm{E}(\mathcal{M}^{\otimes}) E ( M ⊗ ) for an endomorphism algebra object of the unit. Suppose that E ( M ⊗ ) \mathrm{E}(\mathcal{M}^{\otimes}) E ( M ⊗ ) also has an endomorphism algebra of a unit which we denote by E 2 ( M ⊗ ) ∈ Alg 3 ( ( n − 1 ) P ) , … , E n − 1 ( M ⊗ ) \mathrm{E}^2(\mathcal{M}^{\otimes})\in\operatorname{Alg}_3((n-1)\mathrm{P}),\ldots,\mathrm{E}^{n-1}(\mathcal{M}^{\otimes}) E 2 ( M ⊗ ) ∈ Alg 3 (( n − 1 ) P ) , … , E n − 1 ( M ⊗ ) also has an endomorphism algebra of
a unit which we denote by E n ( M ⊗ ) ∈ Alg n + 1 ( P ) E^{n}(\mathcal{M}^{\otimes}) \in\operatorname{Alg}_{n+1}(\mathcal{P}) E n ( M ⊗ ) ∈ Alg n + 1 ( P ) . In this case, we say that M ⊗ \mathcal{M}^{\otimes} M ⊗ has a good endomorphism algebra of a unit.
Lemma 4.8. Let n ≥ 0 n \ge0 n ≥ 0 be a nonnegative integer. Let P \mathcal{P} P be either P k t + \mathcal{P}^{\mathrm{t}+}_{k} P k t + or P k \mathcal{P}_{k} P k . The following holds.
Let ( L , L ) (\mathcal{L},L) ( L , L ) be an object in ( ( n + 1 ) P ) n P / ((n+1)\mathcal{P})_{n\mathcal{P}/} (( n + 1 ) P ) n P / . Suppose that ( L , L ) (\mathcal{L},L) ( L , L ) has a good endomorphism algebra. Then ( L , L ) (\mathcal{L},L) ( L , L ) has an algebraization.
Let M ⊗ \mathcal{M}^{\otimes} M ⊗ be an E 1 \mathbb{E}_{1} E 1 -algebra object in ( n + 1 ) P (n+1)\mathcal{P} ( n + 1 ) P . Suppose that M ⊗ \mathcal{M}^{\otimes} M ⊗ has a good endomorphism algebra of a unit. Then M ⊗ \mathcal{M}^{\otimes} M ⊗ has an algebraization.
Proof. We treat the case of P = P k t + \mathcal{P}=\mathcal{P}^{\mathrm{t}+}_{k} P = P k t + ; the case of P = P k \mathcal{P}=\mathcal{P}_{k} P = P k is similar. We first prove (1). We treat the case of n = 0 n=0 n = 0 . In this situation, ( L , L ) ∈ ( P k t + ) Mod k / (\mathcal{L},L)\in(\mathcal{P}^{\mathrm{t}+}_{k})_{\operatorname{Mod}_{k}/} ( L , L ) ∈ ( P k t + ) Mod k / . For any L , L ′ ∈ L L,L'\in\mathcal{L} L , L ′ ∈ L there exists a Hom object Hom L ( L , L ′ ) \operatorname{Hom}_{\mathcal{L}}(L,L') Hom L ( L , L ′ ) in Mod k \operatorname{Mod}_{k} Mod k since Mod k \operatorname{Mod}_{k} Mod k is presentable. Recall that θ 0 P k , 0 : Alg 1 ( Mod k ) → ( P k ) Mod k / \theta^{\mathcal{P}_{k},0}_{0}:\operatorname{Alg}_{1}(\operatorname{Mod}_{k})\to(\mathcal{P}_{k})_{\operatorname{Mod}_{k}/} θ 0 P k , 0 : Alg 1 ( Mod k ) → ( P k ) Mod k / is extended to θ 0 P k t + , 0 : Alg 1 ( Mod k ) → ( P k t + ) Mod k / \theta^{\mathcal{P}^{\mathrm{t}+}_{k},0}_{0}:\operatorname{Alg}_{1}(\operatorname{Mod}_{k})\to(\mathcal{P}^{\mathrm{t}+}_{k})_{\operatorname{Mod}_{k}/} θ 0 P k t + , 0 : Alg 1 ( Mod k ) → ( P k t + ) Mod k / (see Construction 3.26). The functor θ 0 P k , 0 \theta^{\mathcal{P}_{k},0}_{0} θ 0 P k , 0 has a right adjoint which carries ( A , A ) (A,A) ( A , A ) to the opposite of the endomorphism algebra End ( A ) op \operatorname{End}(A)^{\operatorname{op}} End ( A ) op of the object A A A . By the definition of the connective part of the standard t t t -structure (cf. Remark 3.3), θ 0 P k t + , 0 \theta^{\mathcal{P}^{\mathrm{t}+}_{k},0}_{0} θ 0 P k t + , 0 also has a right adjoint E + : ( P k t + ) Mod k / → Alg 1 ( Mod k ) E^{+}:(\mathcal{P}^{\mathrm{t}+}_{k})_{\operatorname{Mod}_{k}/}\to\operatorname{Alg}_{1}(\operatorname{Mod}_{k}) E + : ( P k t + ) Mod k / → Alg 1 ( Mod k ) given informally by ( A , A ) ↦ End ( A ) op (A,A)\mapsto\operatorname{End}(A)^{\operatorname{op}} ( A , A ) ↦ End ( A ) op . Note that θ 0 P k t + , 0 \theta^{\mathcal{P}^{\mathrm{t}+}_{k},0}_{0} θ 0 P k t + , 0 is fully faithful. Thus for any A ∈ Alg 1 ( Mod k ) A\in\operatorname{Alg}_{1}(\operatorname{Mod}_{k}) A ∈ Alg 1 ( Mod k ) we have
Map ( P k t + ) Mod k / ( ( LMod A , A ) , ( L , L ) ) ≃ Map Alg 1 ( Mod k ) ( A , End ( L ) op ) . \operatorname{Map}_{(\mathcal{P}^{\mathrm{t}+}_{k})_{\operatorname{Mod}_{k}/}}\bigl((\operatorname{LMod}_{A},A),(\mathcal{L},L)\bigr)\simeq\operatorname{Map}_{\operatorname{Alg}_{1}(\operatorname{Mod}_{k})}\bigl(A,\operatorname{End}(L)^{\operatorname{op}}\bigr). Map ( P k t + ) Mod k / ( ( LMod A , A ) , ( L , L ) ) ≃ Map Alg 1 ( Mod k ) ( A , End ( L ) op ) .
Thus, the claim holds in the case of n = 0 n=0 n = 0 .
We will show the case of n ≥ 1 n\ge1 n ≥ 1 . Let ( ( n + 1 ) P k t ± ) n P k t + / ⋆ ((n+1)\mathcal{P}^{\mathrm{t}\pm}_{k})^{\star}_{n\mathcal{P}^{\mathrm{t}+}_{k}/} (( n + 1 ) P k t ± ) n P k t + / ⋆ be the full subcategory of ( ( n + 1 ) P k t + ) n P k t + / ((n+1)\mathcal{P}^{\mathrm{t}+}_{k})_{n\mathcal{P}^{\mathrm{t}+}_{k}/} (( n + 1 ) P k t + ) n P k t + / spanned by f : n P k t + → B f:n\mathcal{P}^{\mathrm{t}+}_{k}\to\mathcal{B} f : n P k t + → B such that a Hom object Hom ( B , B ) \operatorname{Hom}(B,B) Hom ( B , B ) exists in n P k t + n\mathcal{P}^{\mathrm{t}+}_{k} n P k t + where B : = f ( ( n − 1 ) P t + ) B:=f((n-1)\mathcal{P}^{\mathrm{t}+}) B := f (( n − 1 ) P t + ) . Note that the essential image of the fully faithful functor θ n P k t + , 0 : Alg 1 ( n P k t + ) → ( ( n + 1 ) P k t + ) n P k t + / \theta^{\mathcal{P}^{\mathrm{t}+}_{k},0}_{n}:\operatorname{Alg}_{1}(n\mathcal{P}^{\mathrm{t}+}_{k})\to((n+1)\mathcal{P}^{\mathrm{t}+}_{k})_{n\mathcal{P}^{\mathrm{t}+}_{k}/} θ n P k t + , 0 : Alg 1 ( n P k t + ) → (( n + 1 ) P k t + ) n P k t + / is contained in ( ( n + 1 ) P k t + ) n P k t + / ⋆ ((n+1)\mathcal{P}^{\mathrm{t}+}_{k})^{\star}_{n\mathcal{P}^{\mathrm{t}+}_{k}/} (( n + 1 ) P k t + ) n P k t + / ⋆ since for any C ∈ Alg 1 ( n P k t + ) C\in\operatorname{Alg}_{1}(n\mathcal{P}^{\mathrm{t}+}_{k}) C ∈ Alg 1 ( n P k t + ) the free module C C C in LMod C ( n P t + ) \operatorname{LMod}_{C}(n\mathcal{P}^{\mathrm{t}+}) LMod C ( n P t + ) has the Hom object C = Hom ( C , C ) C=\operatorname{Hom}(C,C) C = Hom ( C , C ) . Observe that the restricted functor θ n P k t + , 0 ⋆ : Alg 1 ( n P k t + ) → ( ( n + 1 ) P k t + ) n P k t + / ⋆ \theta^{\star}_{n\mathcal{P}^{\mathrm{t}+}_{k},0}:\operatorname{Alg}_{1}(n\mathcal{P}^{\mathrm{t}+}_{k})\to((n+1)\mathcal{P}^{\mathrm{t}+}_{k})^{\star}_{n\mathcal{P}^{\mathrm{t}+}_{k}/} θ n P k t + , 0 ⋆ : Alg 1 ( n P k t + ) → (( n + 1 ) P k t + ) n P k t + / ⋆ has a right adjoint E n P k t + ⋆ E^{\star}_{n\mathcal{P}^{\mathrm{t}+}_{k}} E n P k t + ⋆ . For this purpose, it will suffice to show that for any [ f : n P k t + → B ] ∈ ( ( n + 1 ) P k t + ) n P k t + / ⋆ [f:n\mathcal{P}^{\mathrm{t}+}_{k}\to\mathcal{B}]\in((n+1)\mathcal{P}^{\mathrm{t}+}_{k})^{\star}_{n\mathcal{P}^{\mathrm{t}+}_{k}/} [ f : n P k t + → B ] ∈ (( n + 1 ) P k t + ) n P k t + / ⋆ , the object B : = f ( ( n − 1 ) P t + ) B:=f((n-1)\mathcal{P}^{\mathrm{t}+}) B := f (( n − 1 ) P t + ) has an endomorphism algebra defined as an object of Alg 1 ( n P k t + ) \operatorname{Alg}_{1}(n\mathcal{P}^{\mathrm{t}+}_{k}) Alg 1 ( n P k t + ) . According to [24 4.7.1.40, 4.7.1.41] , this is equivalent to the condition that there exists a Hom object Hom ( B , B ) \operatorname{Hom}(B,B) Hom ( B , B ) in n P k t + n\mathcal{P}^{\mathrm{t}+}_{k} n P k t + . However, this is our assumption.
Since ( L , L ) (\mathcal{L},L) ( L , L ) has a good endomorphism algebra, the image E n P k t + ⋆ ( L , L ) E^{\star}_{n\mathcal{P}^{\mathrm{t}+}_{k}}(\mathcal{L},L) E n P k t + ⋆ ( L , L ) lies in Alg 1 ( n P k t + ) ⋆ \operatorname{Alg}_{1}(n\mathcal{P}^{\mathrm{t}+}_{k})^{\star} Alg 1 ( n P k t + ) ⋆ . Its image under the functor E ( n − 1 ) P k t + ⋆ : Alg 1 ( n P k t + ) ⋆ → Alg 2 ( ( n − 1 ) P k t + ) E^{\star}_{(n-1)\mathcal{P}^{\mathrm{t}+}_{k}}:\operatorname{Alg}_{1}(n\mathcal{P}^{\mathrm{t}+}_{k})^{\star}\to\operatorname{Alg}_{2}((n-1)\mathcal{P}^{\mathrm{t}+}_{k}) E ( n − 1 ) P k t + ⋆ : Alg 1 ( n P k t + ) ⋆ → Alg 2 (( n − 1 ) P k t + ) lies in Alg 2 ( ( n − 1 ) P k t + ) ⋆ \operatorname{Alg}_{2}((n-1)\mathcal{P}^{\mathrm{t}+}_{k})^{\star} Alg 2 (( n − 1 ) P k t + ) ⋆ . Iterating this procedure, we get the image D 1 D_{1} D 1 in Alg n ( P k t + ) \operatorname{Alg}_{n}(\mathcal{P}^{\mathrm{t}+}_{k}) Alg n ( P k t + ) . Namely, if we denote by ( ( n + 1 ) P k t + ) n P k t + / ⋆ ⋆ ((n+1)\mathcal{P}^{\mathrm{t}+}_{k})^{\star\star}_{n\mathcal{P}^{\mathrm{t}+}_{k}/} (( n + 1 ) P k t + ) n P k t + / ⋆⋆ the full subcategory spanned by those objects ( L , L ) (\mathcal{L},L) ( L , L ) such that ( L , L ) (\mathcal{L},L) ( L , L ) has a good endomorphism algebra, then the composition yields the adjoint pair
U : Alg n ( P k t + ) ⇄ ( ( n + 1 ) P k t + ) n P k t + / ⋆ ⋆ : V . U:\operatorname{Alg}_{n}(\mathcal{P}^{\mathrm{t}+}_{k})\rightleftarrows((n+1)\mathcal{P}^{\mathrm{t}+}_{k})^{\star\star}_{n\mathcal{P}^{\mathrm{t}+}_{k}/}:V. U : Alg n ( P k t + ) ⇄ (( n + 1 ) P k t + ) n P k t + / ⋆⋆ : V .
Consider the adjoint pair
Alg n ( θ P k t + , 0 ) : Alg n + 1 ( Mod k ) ⇄ Alg n ( P k t + ) : Alg n ( E + ) \operatorname{Alg}_{n}(\theta^{\mathcal{P}^{\mathrm{t}+}_{k},0}):\operatorname{Alg}_{n+1}(\operatorname{Mod}_{k})\rightleftarrows\operatorname{Alg}_{n}(\mathcal{P}^{\mathrm{t}+}_{k}):\operatorname{Alg}_{n}(E^{+}) Alg n ( θ P k t + , 0 ) : Alg n + 1 ( Mod k ) ⇄ Alg n ( P k t + ) : Alg n ( E + )
induced by ( θ P k t + , 0 , E + ) (\theta^{\mathcal{P}^{\mathrm{t}+}_{k},0},E^{+}) ( θ P k t + , 0 , E + ) . Applying Alg n ( E + ) : Alg n ( P k t + ) → Alg n + 1 ( Mod k ) \operatorname{Alg}_{n}(E^{+}):\operatorname{Alg}_{n}(\mathcal{P}^{\mathrm{t}+}_{k})\to\operatorname{Alg}_{n+1}(\operatorname{Mod}_{k}) Alg n ( E + ) : Alg n ( P k t + ) → Alg n + 1 ( Mod k ) to D 1 D_{1} D 1 , we have D 0 : = Alg n ( E + ) ( D 1 ) ∈ Alg n + 1 ( Mod k ) D_{0}:=\operatorname{Alg}_{n}(E^{+})(D_{1})\in\operatorname{Alg}_{n+1}(\operatorname{Mod}_{k}) D 0 := Alg n ( E + ) ( D 1 ) ∈ Alg n + 1 ( Mod k ) . Using the sequence of adjoint pairs we have the canonical equivalence
Map Alg n + 1 ( Mod k ) ( C , D 0 ) ≃ Map Alg n ( P k t + ) ( LMod C , D 1 ) ≃ Map ( ( n + 1 ) P k t + ) n P k t + / ( ( ( n P k t + , ( n − 1 ) P k t + ) , ( L , L ) ) ) \operatorname{Map}_{\operatorname{Alg}_{n+1}(\operatorname{Mod}_{k})}(C,D_{0})\simeq\operatorname{Map}_{\operatorname{Alg}_{n}(\mathcal{P}^{\mathrm{t}+}_{k})}(\operatorname{LMod}_{C},D_{1})\simeq\operatorname{Map}_{((n+1)\mathcal{P}^{\mathrm{t}+}_{k})_{n\mathcal{P}^{\mathrm{t}+}_{k}/}}\bigl(((n\mathcal{P}^{\mathrm{t}+}_{k},(n-1)\mathcal{P}^{\mathrm{t}+}_{k}),(\mathcal{L},L))\bigr) Map Alg n + 1 ( Mod k ) ( C , D 0 ) ≃ Map Alg n ( P k t + ) ( LMod C , D 1 ) ≃ Map (( n + 1 ) P k t + ) n P k t + / ( (( n P k t + , ( n − 1 ) P k t + ) , ( L , L )) )
(notice that U ( LMod C ) ≃ n P k t + U(\operatorname{LMod}_{C})\simeq n\mathcal{P}^{\mathrm{t}+}_{k} U ( LMod C ) ≃ n P k t + ). Thus ( L , L ) (\mathcal{L},L) ( L , L ) has an algebraization.
We next prove (2). The proof proceeds by passing to algebra objects of the constructions in (1). In the case of n = 0 n=0 n = 0 , applying Alg 1 \operatorname{Alg}_{1} Alg 1 to the adjoint pair ( θ P k t + , 0 , E + ) (\theta^{\mathcal{P}^{\mathrm{t}+}_{k},0},E^{+}) ( θ P k t + , 0 , E + ) yields the adjoint pair
θ P k t + , 1 : Alg 2 ( Mod k ) ⇄ Alg 1 ( P k t + ) : E + , \theta^{\mathcal{P}^{\mathrm{t}+}_{k},1}:\operatorname{Alg}_{2}(\operatorname{Mod}_{k})\rightleftarrows\operatorname{Alg}_{1}(\mathcal{P}^{\mathrm{t}+}_{k}):E^{+}, θ P k t + , 1 : Alg 2 ( Mod k ) ⇄ Alg 1 ( P k t + ) : E + ,
where the right adjoint E + E^{+} E + is informally given by A ⊗ ↦ End ( 1 A ) op \mathcal{A}^{\otimes}\mapsto\operatorname{End}(\mathbb{1}_{\mathcal{A}})^{\operatorname{op}} A ⊗ ↦ End ( 1 A ) op . Since θ P k t + , 1 \theta^{\mathcal{P}^{\mathrm{t}+}_{k},1} θ P k t + , 1 is fully faithful, the claim holds as in (1). For n ≥ 1 n \ge1 n ≥ 1 , according to [24 4.7.1.40, 4.7.1.41] , the existence of an endomorphism algebra as an object of Alg 1 ( n Prt k + ) \operatorname{Alg}_{1}(n\operatorname{Prt}_{k}^{+}) Alg 1 ( n Prt k + ) is equivalent to the existence of a Hom object in n Prt k + n\operatorname{Prt}_{k}^{+} n Prt k + , which is exactly the defining condition of ( ( n + 1 ) Prt k + ) n Prt k + / ⋆ \left((n+1)\operatorname{Prt}_{k}^{+}\right)^{\star}_{n\operatorname{Prt}_{k}^{+}/} ( ( n + 1 ) Prt k + ) n Prt k + / ⋆ . Consequently, we have a canonical equivalence of full subcategories Alg 1 ( ( n + 1 ) Prt k + ) ⋆ ≃ Alg 1 ( ( ( n + 1 ) Prt k + ) n Prt k + / ⋆ ) \operatorname{Alg}_{1}\left((n+1)\operatorname{Prt}_{k}^{+}\right)^{\star} \simeq\operatorname{Alg}_{1}\left(\left((n+1)\operatorname{Prt}_{k}^{+}\right)^{\star}_{n\operatorname{Prt}_{k}^{+}/}\right) Alg 1 ( ( n + 1 ) Prt k + ) ⋆ ≃ Alg 1 ( ( ( n + 1 ) Prt k + ) n Prt k + / ⋆ ) , where the left hand side is spanned by those objects having a good endomorphism algebra of a unit. Taking algebra objects of the adjoint pair ( θ n Prt k + / ⋆ , E n Prt k + ⋆ ) \left(\theta^{\star}_{n\operatorname{Prt}_{k}^{+}/},E^{\star}_{n\operatorname{Prt}_{k}^{+}}\right) ( θ n Prt k + / ⋆ , E n Prt k + ⋆ ) from (1) gives the adjoint pair
Alg 1 ( θ n Prt k + / ⋆ ) : Alg 2 ( n Prt k + ) ⇄ Alg 1 ( ( n + 1 ) Prt k + ) ⋆ : Alg 1 ( E n Prt k + ⋆ ) . \operatorname{Alg}_{1}\left(\theta^{\star}_{n\operatorname{Prt}_{k}^{+}/}\right):\operatorname{Alg}_{2}\left(n\operatorname{Prt}_{k}^{+}\right)\rightleftarrows\operatorname{Alg}_{1}\left((n+1)\operatorname{Prt}_{k}^{+}\right)^{\star}:\operatorname{Alg}_{1}\left(E^{\star}_{n\operatorname{Prt}_{k}^{+}}\right). Alg 1 ( θ n Prt k + / ⋆ ) : Alg 2 ( n Prt k + ) ⇄ Alg 1 ( ( n + 1 ) Prt k + ) ⋆ : Alg 1 ( E n Prt k + ⋆ ) .
Iterating this procedure
Alg n + 2 ( Mod k ) ⇄ θ Prt k + θ Prt k + ⋆ Alg n + 1 ( Prt k + ) ⇄ U ′ V ′ Alg 1 ( ( n + 1 ) Prt k + ) ⋆ ⋆ \operatorname{Alg}_{n+2}(\operatorname{Mod}_{k}) \underset{\theta^{\star}_{\operatorname{Prt}_{k}^{+}}}{\overset{\theta_{\operatorname{Prt}_{k}^{+}}}{\rightleftarrows}} \operatorname{Alg}_{n+1}(\operatorname{Prt}_{k}^{+}) \underset{V'}{\overset{U'}{\rightleftarrows}} \operatorname{Alg}_{1}\left((n+1)\operatorname{Prt}_{k}^{+}\right)^{\star\star} Alg n + 2 ( Mod k ) θ Prt k + ⋆ ⇄ θ Prt k + Alg n + 1 ( Prt k + ) V ′ ⇄ U ′ Alg 1 ( ( n + 1 ) Prt k + ) ⋆⋆
and evaluating the resulting sequence of right adjoints on M ⊗ \mathcal{M}^{\otimes} M ⊗ to obtain D 1 ∈ Alg n + 1 ( Prt k + ) D_{1}\in\operatorname{Alg}_{n+1}(\operatorname{Prt}_{k}^{+}) D 1 ∈ Alg n + 1 ( Prt k + ) and D 0 ∈ Alg n + 2 ( Mod k ) D_{0}\in\operatorname{Alg}_{n+2}(\operatorname{Mod}_{k}) D 0 ∈ Alg n + 2 ( Mod k ) , we deduce the canonical equivalence
Map Alg n + 2 ( Mod k ) ( C , D 0 ) ≃ Map Alg n + 1 ( Prt k + ) ( LMod C , D 1 ) ≃ Map Alg 1 ( ( n + 1 ) Prt k + ) ( n Prt k + , M ⊗ ) . \operatorname{Map}_{\operatorname{Alg}_{n+2}(\operatorname{Mod}_{k})}(C,D_{0})\simeq\operatorname{Map}_{\operatorname{Alg}_{n+1}(\operatorname{Prt}_{k}^{+})}(\operatorname{LMod}_{C},D_{1})\simeq\operatorname{Map}_{\operatorname{Alg}_{1}((n+1)\operatorname{Prt}_{k}^{+})}(n\operatorname{Prt}_{k}^{+},\mathcal{M}^{\otimes}). Map Alg n + 2 ( Mod k ) ( C , D 0 ) ≃ Map Alg n + 1 ( Prt k + ) ( LMod C , D 1 ) ≃ Map Alg 1 (( n + 1 ) Prt k + ) ( n Prt k + , M ⊗ ) .
exactly as in (1). Thus M ⊗ \mathcal{M}^{\otimes} M ⊗ has an algebraization. □
Proposition 4.9. (1) Let A A A be an E n + 2 \mathbb{E}_{n+2} E n + 2 -algebra in Mod k \operatorname{Mod}_{k} Mod k . Then there exists a canonical equivalence of monoidal ∞ \infty ∞ -categories
n Prt A + ≃ LMod ( n − 1 ) Prt A + ( n Prt k + ) . n\operatorname{Prt}_{A}^{+}\simeq\operatorname{LMod}_{(n-1)\operatorname{Prt}_{A}^{+}}\left(n\operatorname{Prt}_{k}^{+}\right). n Prt A + ≃ LMod ( n − 1 ) Prt A + ( n Prt k + ) . Similarly, n Prt A ± ≃ LMod ( n − 1 ) Prt A ± ( n Prt k ± ) n\operatorname{Prt}_{A}^{\pm}\simeq\operatorname{LMod}_{(n-1)\operatorname{Prt}_{A}^{\pm}}\left(n\operatorname{Prt}_{k}^{\pm}\right) n Prt A ± ≃ LMod ( n − 1 ) Prt A ± ( n Prt k ± ) .
(2) Let A → B A\to B A → B be a morphism of E n + 2 \mathbb{E}_{n+2} E n + 2 -algebras in Mod k \operatorname{Mod}_{k} Mod k . The induced morphism n Prt A + → n Prt B + n\operatorname{Prt}_{A}^{+}\to n\operatorname{Prt}_{B}^{+} n Prt A + → n Prt B + in Alg 1 ( ( n + 1 ) Prt k + ) \operatorname{Alg}_{1}\left((n+1)\operatorname{Prt}_{k}^{+}\right) Alg 1 ( ( n + 1 ) Prt k + ) defines the restriction functor LMod n Prt A + ( Cat ^ † ) → LMod n Prt A + ( Cat ^ † ) \operatorname{LMod}_{n\operatorname{Prt}_{A}^{+}}(\widehat{\operatorname{Cat}}_{\dagger})\to\operatorname{LMod}_{n\operatorname{Prt}_{A}^{+}}(\widehat{\operatorname{Cat}}_{\dagger}) LMod n Prt A + ( Cat † ) → LMod n Prt A + ( Cat † ) . The image of n Prt B + n\operatorname{Prt}_{B}^{+} n Prt B + lies in ( n + 1 ) Prt A + (n+1)\operatorname{Prt}_{A}^{+} ( n + 1 ) Prt A + . Namely, n Prt B + n\operatorname{Prt}_{B}^{+} n Prt B + is κ 0 \kappa_{0} κ 0 -compact in LMod n Prt A + ( Cat ^ † ) \operatorname{LMod}_{n\operatorname{Prt}_{A}^{+}}(\widehat{\operatorname{Cat}}_{\dagger}) LMod n Prt A + ( Cat † ) .
Similarly, n Prt B ± n\operatorname{Prt}_{B}^{\pm} n Prt B ± is κ 0 \kappa_{0} κ 0 -compact in LMod n Prt A ± ( Cat ^ † ) \operatorname{LMod}_{n\operatorname{Prt}_{A}^{\pm}}(\widehat{\operatorname{Cat}}_{\dagger}) LMod n Prt A ± ( Cat † ) .
(3) In the situation of (2), the induced morphism ( n − 1 ) Prt A + → ( n − 1 ) Prt B + (n-1)\operatorname{Prt}_{A}^{+}\to(n-1)\operatorname{Prt}_{B}^{+} ( n − 1 ) Prt A + → ( n − 1 ) Prt B + in Alg 2 ( n Prt k + ) \operatorname{Alg}_{2}(n\operatorname{Prt}_{k}^{+}) Alg 2 ( n Prt k + ) exhibits ( n − 1 ) Prt B + (n-1)\operatorname{Prt}_{B}^{+} ( n − 1 ) Prt B + as an algebra object in LMod ( n − 1 ) Prt A + ( n Prt k + ) \operatorname{LMod}_{(n-1)\operatorname{Prt}_{A}^{+}}(n\operatorname{Prt}_{k}^{+}) LMod ( n − 1 ) Prt A + ( n Prt k + ) . There exist canonical equivalences
n Prt B + ≃ LMod ( n − 1 ) Prt B + ( n Prt k + ) ≃ LMod ( n − 1 ) Prt A + ( LMod ( n − 1 ) Prt A + ( n Prt k + ) ) ≃ LMod ( n − 1 ) Prt B + ( n Prt A + ) . n\operatorname{Prt}_{B}^{+}\simeq\operatorname{LMod}_{(n-1)\operatorname{Prt}_{B}^{+}}(n\operatorname{Prt}_{k}^{+})\simeq\operatorname{LMod}_{(n-1)\operatorname{Prt}_{A}^{+}}\left(\operatorname{LMod}_{(n-1)\operatorname{Prt}_{A}^{+}}(n\operatorname{Prt}_{k}^{+})\right)\simeq\operatorname{LMod}_{(n-1)\operatorname{Prt}_{B}^{+}}(n\operatorname{Prt}_{A}^{+}). n Prt B + ≃ LMod ( n − 1 ) Prt B + ( n Prt k + ) ≃ LMod ( n − 1 ) Prt A + ( LMod ( n − 1 ) Prt A + ( n Prt k + ) ) ≃ LMod ( n − 1 ) Prt B + ( n Prt A + ) . Similarly, n Prt B ± ≃ LMod ( n − 1 ) Prt B ± ( n Prt A ± ) n\operatorname{Prt}_{B}^{\pm}\simeq\operatorname{LMod}_{(n-1)\operatorname{Prt}_{B}^{\pm}}(n\operatorname{Prt}_{A}^{\pm}) n Prt B ± ≃ LMod ( n − 1 ) Prt B ± ( n Prt A ± ) .
Proof. We treat the case of Prt + \operatorname{Prt}^{+} Prt + . The case of Prt ± \operatorname{Prt}^{\pm} Prt ± is the same. We prove (1). We first note that ( n − 1 ) Prt A + (n-1)\operatorname{Prt}_{A}^{+} ( n − 1 ) Prt A + belongs to n Prt k + n\operatorname{Prt}_{k}^{+} n Prt k + since ( n − 1 ) Prt A + = LMod ( n − 2 ) Prt A + ( ( n − 1 ) Prt k + ) (n-1)\operatorname{Prt}_{A}^{+}=\operatorname{LMod}_{(n-2)\operatorname{Prt}_{A}^{+}}((n-1)\operatorname{Prt}_{k}^{+}) ( n − 1 ) Prt A + = LMod ( n − 2 ) Prt A + (( n − 1 ) Prt k + ) is dualizable and κ 0 \kappa_{0} κ 0 -compact in Mod ( n − 1 ) Prt k + ( Cat ^ † ) \operatorname{Mod}_{(n-1)\operatorname{Prt}_{k}^{+}}(\widehat{\operatorname{Cat}}_{\dagger}) Mod ( n − 1 ) Prt k + ( Cat † ) (see [30 Proposition 5.1.10] ). We observe an equivalence
Mod ( n − 1 ) Prt A + ( Cat ^ † ) κ 0 ≃ Mod ( n − 1 ) Prt A + ( n Prt k + ) . \operatorname{Mod}_{(n-1)\operatorname{Prt}_{A}^{+}}(\widehat{\operatorname{Cat}}_{\dagger})^{\kappa_{0}}\simeq\operatorname{Mod}_{(n-1)\operatorname{Prt}_{A}^{+}}(n\operatorname{Prt}_{k}^{+}). Mod ( n − 1 ) Prt A + ( Cat † ) κ 0 ≃ Mod ( n − 1 ) Prt A + ( n Prt k + ) .
This right-hand side is Mod ( n − 1 ) Prt A + ( Mod ( n − 1 ) Prt k + ( Cat ^ † ) κ 0 ) \operatorname{Mod}_{(n-1)\operatorname{Prt}_{A}^{+}}\left(\operatorname{Mod}_{(n-1)\operatorname{Prt}_{k}^{+}}(\widehat{\operatorname{Cat}}_{\dagger})^{\kappa_{0}}\right) Mod ( n − 1 ) Prt A + ( Mod ( n − 1 ) Prt k + ( Cat † ) κ 0 ) , and
Mod ( n − 1 ) Prt A + ( Mod ( n − 1 ) Prt k + ( Cat ^ † ) ) ⊂ Mod ( n − 1 ) Prt A + ( Mod ( n − 1 ) Prt k + ( Cat ^ † ) ) ≃ Mod ( n − 1 ) Prt A + ( Cat ^ † ) \operatorname{Mod}_{(n-1)\operatorname{Prt}_{A}^{+}}\left(\operatorname{Mod}_{(n-1)\operatorname{Prt}_{k}^{+}}(\widehat{\operatorname{Cat}}_{\dagger})\right)\subset\operatorname{Mod}_{(n-1)\operatorname{Prt}_{A}^{+}}\left(\operatorname{Mod}_{(n-1)\operatorname{Prt}_{k}^{+}}(\widehat{\operatorname{Cat}}_{\dagger})\right)\simeq\operatorname{Mod}_{(n-1)\operatorname{Prt}_{A}^{+}}(\widehat{\operatorname{Cat}}_{\dagger}) Mod ( n − 1 ) Prt A + ( Mod ( n − 1 ) Prt k + ( Cat † ) ) ⊂ Mod ( n − 1 ) Prt A + ( Mod ( n − 1 ) Prt k + ( Cat † ) ) ≃ Mod ( n − 1 ) Prt A + ( Cat † )
where ⊂ \subset ⊂ means the inclusion as a full subcategory (see [24 7.1.3.1, 7.1.3.3] for the latter monoidal equivalence). We note that Mod ( n − 1 ) Prt A + ( Mod ( n − 1 ) Prt k + ( Cat ^ † ) κ 0 ) \operatorname{Mod}_{(n-1)\operatorname{Prt}_{A}^{+}}\left(\operatorname{Mod}_{(n-1)\operatorname{Prt}_{k}^{+}}(\widehat{\operatorname{Cat}}_{\dagger})^{\kappa_{0}}\right) Mod ( n − 1 ) Prt A + ( Mod ( n − 1 ) Prt k + ( Cat † ) κ 0 ) is the smallest full subcategory of Mod ( n − 1 ) Prt A + ( Cat ^ † ) \operatorname{Mod}_{(n-1)\operatorname{Prt}_{A}^{+}}(\widehat{\operatorname{Cat}}_{\dagger}) Mod ( n − 1 ) Prt A + ( Cat † ) which contains the collection of objects { ( n − 1 ) Prt A + ⊗ ( n − 1 ) Prt k + ( n − 1 ) Prt k + ⊗ P } P ∈ Pr \{(n-1)\operatorname{Prt}_{A}^{+}\otimes_{(n-1)\operatorname{Prt}_{k}^{+}}(n-1)\operatorname{Prt}_{k}^{+}\otimes\mathcal{P}\}_{\mathcal{P}\in\operatorname{Pr}} {( n − 1 ) Prt A + ⊗ ( n − 1 ) Prt k + ( n − 1 ) Prt k + ⊗ P } P ∈ Pr and is closed under small colimits (see [30 Proposition 5.1.7] ). Again by [30 Proposition 5.1.7] , Mod ( n − 1 ) Prt A + ( Cat ^ † ) κ 0 \operatorname{Mod}_{(n-1)\operatorname{Prt}_{A}^{+}}(\widehat{\operatorname{Cat}}_{\dagger})^{\kappa_{0}} Mod ( n − 1 ) Prt A + ( Cat † ) κ 0 is the smallest full subcategory of Mod ( n − 1 ) Prt A + ( Cat ^ † ) \operatorname{Mod}_{(n-1)\operatorname{Prt}_{A}^{+}}(\widehat{\operatorname{Cat}}_{\dagger}) Mod ( n − 1 ) Prt A + ( Cat † ) which contains the collection of objects { ( n − 1 ) Prt A + ⊗ P } P ∈ Pr \{(n-1)\operatorname{Prt}_{A}^{+}\otimes\mathcal{P}\}_{\mathcal{P}\in\operatorname{Pr}} {( n − 1 ) Prt A + ⊗ P } P ∈ Pr and is closed under small colimits. Thus, we see the equivalence. We prove (2) and (3). By (1), ϕ : RMod ( n − 1 ) P r B t + ( n P r A t + ) ≃ LMod ( n − 1 ) P r B t + ( n P r A t + ) ≃ n P r A t + \phi:\operatorname{RMod}_{(n-1)\mathrm{Pr}^{t+}_{B}}(n\mathrm{Pr}^{t+}_{A}) \simeq\operatorname{LMod}_{(n-1)\mathrm{Pr}^{t+}_{B}}(n\mathrm{Pr}^{t+}_{A}) \simeq n\mathrm{Pr}^{t+}_{A} ϕ : RMod ( n − 1 ) Pr B t + ( n Pr A t + ) ≃ LMod ( n − 1 ) Pr B t + ( n Pr A t + ) ≃ n Pr A t + (the first equivalence follows from the fact that ( n − 1 ) P r A t + (n-1)\mathrm{Pr}^{t+}_{A} ( n − 1 ) Pr A t + is an E 2 \mathrm{E}_{2} E 2 -algebra). Thus we have
RMod ( n − 1 ) P r B t + ( n P r A t + ) ≃ LMod ( n − 1 ) P r B t + ( n P r A t + ) ≃ RMod ( n − 1 ) P r B t + ( RMod ( n − 1 ) P r A t + ( n P r k t + ) ) ≃ RMod ( n − 1 ) P r B t + ( n P r k t + ) ≃ LMod ( n − 1 ) P r B t + ( n P r k t + ) = n P r B t + . \begin{aligned}
\operatorname{RMod}_{(n-1)\mathrm{Pr}^{t+}_{B}}(n\mathrm{Pr}^{t+}_{A}) &\simeq\operatorname{LMod}_{(n-1)\mathrm{Pr}^{t+}_{B}}(n\mathrm{Pr}^{t+}_{A})\\
&\simeq\operatorname{RMod}_{(n-1)\mathrm{Pr}^{t+}_{B}}\left(\operatorname{RMod}_{(n-1)\mathrm{Pr}^{t+}_{A}}(n\mathrm{Pr}^{t+}_{k})\right)\\
&\simeq\operatorname{RMod}_{(n-1)\mathrm{Pr}^{t+}_{B}}(n\mathrm{Pr}^{t+}_{k})\\
&\simeq\operatorname{LMod}_{(n-1)\mathrm{Pr}^{t+}_{B}}(n\mathrm{Pr}^{t+}_{k})=n\mathrm{Pr}^{t+}_{B}.
\end{aligned} RMod ( n − 1 ) Pr B t + ( n Pr A t + ) ≃ LMod ( n − 1 ) Pr B t + ( n Pr A t + ) ≃ RMod ( n − 1 ) Pr B t + ( RMod ( n − 1 ) Pr A t + ( n Pr k t + ) ) ≃ RMod ( n − 1 ) Pr B t + ( n Pr k t + ) ≃ LMod ( n − 1 ) Pr B t + ( n Pr k t + ) = n Pr B t + .
where the first and 4-th equivalences follow from the fact that ( n − 1 ) P r B t + (n-1)\mathrm{Pr}^{t+}_{B} ( n − 1 ) Pr B t + is a E 2 \mathrm{E}_{2} E 2 -algebra, ϕ \phi ϕ induces the second equivalence, and the third equivalence is induced by the forgetful functor (see [24 7.1.3.1] ). By [30 Proposition 5.1.10] , RMod ( n − 1 ) P r B t + ( n P r A t + ) \operatorname{RMod}_{(n-1)\mathrm{Pr}^{t+}_{B}}(n\mathrm{Pr}^{t+}_{A}) RMod ( n − 1 ) Pr B t + ( n Pr A t + ) is κ 0 \kappa_{0} κ 0 -compact in LMod n P r A t + ( Cat ^ ) \operatorname{LMod}_{n\mathrm{Pr}^{t+}_{A}}(\widehat{\operatorname{Cat}}) LMod n Pr A t + ( Cat ) . Thus, n P r B t + ∈ ( n + 1 ) P r A t + n\mathrm{Pr}^{t+}_{B}\in(n+1)\mathrm{Pr}^{t+}_{A} n Pr B t + ∈ ( n + 1 ) Pr A t + . □ \square □
Lemma 4.10. Let A A A be an augmented E n + 2 \mathrm{E}_{n+2} E n + 2 -algebra. There exists an endomorphism algebra object End n P r A t + l ( n P r k t + ) ∈ Alg 1 ( ( n + 1 ) P r k t + ) \operatorname{End}^{l}_{n\mathrm{Pr}^{t+}_{A}}(n\mathrm{Pr}^{t+}_{k})\in\operatorname{Alg}_{1}((n+1)\mathrm{Pr}^{t+}_{k}) End n Pr A t + l ( n Pr k t + ) ∈ Alg 1 (( n + 1 ) Pr k t + ) of ( ( n + 1 ) P r A t + , n P r k t + ∈ ( n + 1 ) P r A t + ) ((n+1)\mathrm{Pr}^{t+}_{A},n\mathrm{Pr}^{t+}_{k}\in(n+1)\mathrm{Pr}^{t+}_{A}) (( n + 1 ) Pr A t + , n Pr k t + ∈ ( n + 1 ) Pr A t + ) . Similarly, there exists an endomorphism algebra object End n P r A t ± l ( n P r k t ± ) ∈ Alg 1 ( ( n + 1 ) P r k t ± ) \operatorname{End}^{l}_{n\mathrm{Pr}^{t\pm}_{A}}(n\mathrm{Pr}^{t\pm}_{k})\in\operatorname{Alg}_{1}((n+1)\mathrm{Pr}^{t\pm}_{k}) End n Pr A t ± l ( n Pr k t ± ) ∈ Alg 1 (( n + 1 ) Pr k t ± ) of ( ( n + 1 ) P r A t + , n P r k t ± ∈ ( n + 1 ) P r A t ± ) ((n+1)\mathrm{Pr}^{t+}_{A},n\mathrm{Pr}^{t\pm}_{k}\in(n+1)\mathrm{Pr}^{t\pm}_{A}) (( n + 1 ) Pr A t + , n Pr k t ± ∈ ( n + 1 ) Pr A t ± ) .
Proof. The case of End n P r A t ± l ( n P r k t ± ) \operatorname{End}^{l}_{n\mathrm{Pr}^{t\pm}_{A}}(n\mathrm{Pr}^{t\pm}_{k}) End n Pr A t ± l ( n Pr k t ± ) is the same with End n P r A t + l ( n P r k t + ) \operatorname{End}^{l}_{n\mathrm{Pr}^{t+}_{A}}(n\mathrm{Pr}^{t+}_{k}) End n Pr A t + l ( n Pr k t + ) . We prove the case of End n P r A t + l ( n P r k t + ) \operatorname{End}^{l}_{n\mathrm{Pr}^{t+}_{A}}(n\mathrm{Pr}^{t+}_{k}) End n Pr A t + l ( n Pr k t + ) . By [24 4.7.1.40, 4.7.1.41] (or [24 3.2.2.4] ), it is enough to prove that there exists a Hom object E = Hom n P r A t + l ( n P r k t + , n P r k t + ) E=\operatorname{Hom}^{l}_{n\mathrm{Pr}^{t+}_{A}}(n\mathrm{Pr}^{t+}_{k},n\mathrm{Pr}^{t+}_{k}) E = Hom n Pr A t + l ( n Pr k t + , n Pr k t + ) in ( n + 1 ) P r k t + (n+1)\mathrm{Pr}^{t+}_{k} ( n + 1 ) Pr k t + . Since Mod n P r k t + ( Cat ^ ) \operatorname{Mod}_{n\mathrm{Pr}^{t+}_{k}}(\widehat{\operatorname{Cat}}) Mod n Pr k t + ( Cat ) is a V \mathrm{V} V -presentable ∞ \infty ∞ -category, E E E exists in Mod n P r k t + ( Cat ^ ) \operatorname{Mod}_{n\mathrm{Pr}^{t+}_{k}}(\widehat{\operatorname{Cat}}) Mod n Pr k t + ( Cat ) . There exist equivalences in Mod n P r k t + ( Cat ^ ) \operatorname{Mod}_{n\mathrm{Pr}^{t+}_{k}}(\widehat{\operatorname{Cat}}) Mod n Pr k t + ( Cat ) :
E ≃ Hom n P r k t + l ( n P r k t + ⊗ n P r A t + n P r k t + , n P r k t + ) ≃ Hom n P r k t + l ( n P r k t + ⊗ A k , n P r k t + ) ≃ Hom n P r k t + l ( LMod ( n − 1 ) P r k t + ⊗ A k ( n P r k t + ) , n P r k t + ) ≃ Hom n P r k t + l ( RMod ( n − 1 ) P r k t + ⊗ A k ( n P r k t + ) , n P r k t + ) ≃ LMod ( n − 1 ) P r k t + ⊗ A k ( n P r k t + ) \begin{aligned}
E &\simeq\operatorname{Hom}^{l}_{n\mathrm{Pr}^{t+}_{k}}(n\mathrm{Pr}^{t+}_{k}\otimes_{n\mathrm{Pr}^{t+}_{A}}n\mathrm{Pr}^{t+}_{k},n\mathrm{Pr}^{t+}_{k})\\
&\simeq\operatorname{Hom}^{l}_{n\mathrm{Pr}^{t+}_{k}}(n\mathrm{Pr}^{t+}_{k}\otimes_{A}k,n\mathrm{Pr}^{t+}_{k})\\
&\simeq\operatorname{Hom}^{l}_{n\mathrm{Pr}^{t+}_{k}}(\operatorname{LMod}_{(n-1)\mathrm{Pr}^{t+}_{k}\otimes_{A}k}(n\mathrm{Pr}^{t+}_{k}),n\mathrm{Pr}^{t+}_{k})\\
&\simeq\operatorname{Hom}^{l}_{n\mathrm{Pr}^{t+}_{k}}(\operatorname{RMod}_{(n-1)\mathrm{Pr}^{t+}_{k}\otimes_{A}k}(n\mathrm{Pr}^{t+}_{k}),n\mathrm{Pr}^{t+}_{k})\\
&\simeq\operatorname{LMod}_{(n-1)\mathrm{Pr}^{t+}_{k}\otimes_{A}k}(n\mathrm{Pr}^{t+}_{k})
\end{aligned} E ≃ Hom n Pr k t + l ( n Pr k t + ⊗ n Pr A t + n Pr k t + , n Pr k t + ) ≃ Hom n Pr k t + l ( n Pr k t + ⊗ A k , n Pr k t + ) ≃ Hom n Pr k t + l ( LMod ( n − 1 ) Pr k t + ⊗ A k ( n Pr k t + ) , n Pr k t + ) ≃ Hom n Pr k t + l ( RMod ( n − 1 ) Pr k t + ⊗ A k ( n Pr k t + ) , n Pr k t + ) ≃ LMod ( n − 1 ) Pr k t + ⊗ A k ( n Pr k t + )
where the first one is induced by base change along n P r A t + → n P r k t + n\mathrm{Pr}^{t+}_{A}\to n\mathrm{Pr}^{t+}_{k} n Pr A t + → n Pr k t + , the second one comes from Lemma 4.16, the third equivalence arises from the symmetry of bar construction, and the final equivalence is induced by the P r k t + \mathrm{Pr}^{t+}_{k} Pr k t + -linear functor LMod ( n − 1 ) P r k t + ⊗ A k ( n P r k t + ) ⊗ P r k t + RMod ( n − 1 ) P r k t + ⊗ A k ( n P r k t + ) → P r k t + \operatorname{LMod}_{(n-1)\mathrm{Pr}^{t+}_{k}\otimes_{A}k}(n\mathrm{Pr}^{t+}_{k})\otimes_{\mathrm{Pr}^{t+}_{k}}\operatorname{RMod}_{(n-1)\mathrm{Pr}^{t+}_{k}\otimes_{A}k}(n\mathrm{Pr}^{t+}_{k})\to\mathrm{Pr}^{t+}_{k} LMod ( n − 1 ) Pr k t + ⊗ A k ( n Pr k t + ) ⊗ Pr k t + RMod ( n − 1 ) Pr k t + ⊗ A k ( n Pr k t + ) → Pr k t + induced by the integral kernel of the diagonal bimodule ( n − 1 ) P r k ⊗ A k t + (n-1)\mathrm{Pr}^{t+}_{k\otimes_{A}k} ( n − 1 ) Pr k ⊗ A k t + (cf. [24 4.8.4.1, 4.8.4.8] ). By [30 Proposition 5.1.10] , LMod ( n − 1 ) P r k ⊗ A k t + ( n P r k t + ) \operatorname{LMod}_{(n-1)\mathrm{Pr}^{t+}_{k\otimes_{A}k}}(n\mathrm{Pr}^{t+}_{k}) LMod ( n − 1 ) Pr k ⊗ A k t + ( n Pr k t + ) belongs to ( n + 1 ) P r k t + (n+1)\mathrm{Pr}^{t+}_{k} ( n + 1 ) Pr k t + . □ \square □
Lemma 4.11. Suppose n ≥ 1 n\ge1 n ≥ 1 . The following holds.
Let B B B be an augmented E n + 1 \mathrm{E}_{n+1} E n + 1 -algebra. The pair ( n P r B t + , ( n − 1 ) P r k t + ) (n\mathrm{Pr}^{t+}_{B},(n-1)\mathrm{Pr}^{t+}_{k}) ( n Pr B t + , ( n − 1 ) Pr k t + ) has a good endomorphism algebra.
Let A A A be an augmented E n + 2 \mathrm{E}_{n+2} E n + 2 -algebra. The endomorphism algebra End n P r A t + l ( n P r k t + ) ∈ Alg 1 ( ( n + 1 ) P r k t ± ) \operatorname{End}^{l}_{n\mathrm{Pr}^{t+}_{A}}(n\mathrm{Pr}^{t+}_{k})\in\operatorname{Alg}_{1}((n+1)\mathrm{Pr}^{t\pm}_{k}) End n Pr A t + l ( n Pr k t + ) ∈ Alg 1 (( n + 1 ) Pr k t ± ) has a good endomorphism algebra of a unit.
Proof. We prove (2). The proof of (1) is similar to that of (2): the argument works by replacing k ⊗ A k k\otimes_{A}k k ⊗ A k with B B B . The proof of (2) is a generalization of the proof of Lemma 4.10. From the proof of Lemma 4.10, there exists an equivalence Hom n P r A t + l ( n P r k t + , n P r k t + ) ≃ n P r k ⊗ A k t + \operatorname{Hom}^{l}_{n\mathrm{Pr}^{t+}_{A}}(n\mathrm{Pr}^{t+}_{k},n\mathrm{Pr}^{t+}_{k})\simeq n\mathrm{Pr}^{t+}_{k\otimes_{A}k} Hom n Pr A t + l ( n Pr k t + , n Pr k t + ) ≃ n Pr k ⊗ A k t + in ( n + 1 ) P r k t + (n+1)\mathrm{Pr}^{t+}_{k} ( n + 1 ) Pr k t + . This equivalence sends the identity functor of n P r k t + n\mathrm{Pr}^{t+}_{k} n Pr k t + to the left ( n − 1 ) P r k ⊗ A k t + (n-1)\mathrm{Pr}^{t+}_{k\otimes_{A}k} ( n − 1 ) Pr k ⊗ A k t + -module ( n − 1 ) P r k t + (n-1)\mathrm{Pr}^{t+}_{k} ( n − 1 ) Pr k t + , which is the integral kernel of the identity functor. We write Bar ( A ) \operatorname{Bar}(A) Bar ( A ) for k ⊗ A k k\otimes_{A}k k ⊗ A k . We regard it as an E n + 1 \mathrm{E}_{n+1} E n + 1 -algebra in Mod k \operatorname{Mod}_{k} Mod k , that is defined by the bar construction (more generally, it is an E n + 1 \mathrm{E}_{n+1} E n + 1 -algebra in Alg 1 ( Mod k ) \operatorname{Alg}_{1}(\operatorname{Mod}_{k}) Alg 1 ( Mod k ) ). Next, we prove that there exists an endomorphism algebra End ( n − 1 ) P r Bar ( A ) t + l ( ( n − 1 ) P r k t + ) \operatorname{End}^{l}_{(n-1)\mathrm{Pr}^{t+}_{\operatorname{Bar}(A)}}((n-1)\mathrm{Pr}^{t+}_{k}) End ( n − 1 ) Pr Bar ( A ) t + l (( n − 1 ) Pr k t + ) in Alg 1 ( n P r k t + ) \operatorname{Alg}_{1}(n\mathrm{Pr}^{t+}_{k}) Alg 1 ( n Pr k t + ) . As in the proof of Lemma 4.10, it is enough to show that there exists a Hom object Hom ( n − 1 ) Prt Bar ( A ) + l ( ( n − 1 ) Prt k + , ( n − 1 ) Prt k + ) \operatorname{Hom}^{l}_{(n-1)\operatorname{Prt}^{+}_{\operatorname{Bar}(A)}}((n-1)\operatorname{Prt}^{+}_{k},(n-1)\operatorname{Prt}^{+}_{k}) Hom ( n − 1 ) Prt Bar ( A ) + l (( n − 1 ) Prt k + , ( n − 1 ) Prt k + ) in n Prt k + n\operatorname{Prt}^{+}_{k} n Prt k + . By the same argument, we have an equivalence
Hom ( n − 1 ) Prt Bar ( A ) + l ( ( n − 1 ) Prt k + , ( n − 1 ) Prt k + ) ≃ ( n − 1 ) Prt Bar 2 ( A ) + \operatorname{Hom}^{l}_{(n-1)\operatorname{Prt}^{+}_{\operatorname{Bar}(A)}}((n-1)\operatorname{Prt}^{+}_{k},(n-1)\operatorname{Prt}^{+}_{k}) \simeq(n-1)\operatorname{Prt}^{+}_{\operatorname{Bar}^{2}(A)} Hom ( n − 1 ) Prt Bar ( A ) + l (( n − 1 ) Prt k + , ( n − 1 ) Prt k + ) ≃ ( n − 1 ) Prt Bar 2 ( A ) +
in n Prt k + n\operatorname{Prt}^{+}_{k} n Prt k + , where Bar 2 ( A ) = Bar ( Bar ( A ) ) = k ⊗ Bar ( A ) k \operatorname{Bar}^{2}(A)=\operatorname{Bar}(\operatorname{Bar}(A))=k\otimes_{\operatorname{Bar}(A)}k Bar 2 ( A ) = Bar ( Bar ( A )) = k ⊗ Bar ( A ) k is an E n \mathbb{E}_{n} E n -algebra obtained by the bar construction. This equivalence maps the identity functor of ( n − 1 ) Prt k + (n-1)\operatorname{Prt}^{+}_{k} ( n − 1 ) Prt k + to the left ( n − 2 ) Prt Bar 2 ( A ) + (n-2)\operatorname{Prt}^{+}_{\operatorname{Bar}^{2}(A)} ( n − 2 ) Prt Bar 2 ( A ) + -module ( n − 2 ) Prt k + (n-2)\operatorname{Prt}^{+}_{k} ( n − 2 ) Prt k + . Thus, E ( End n Prt k + l ( n Prt k + ) ) = End ( n − 1 ) Prt Bar ( A ) + l ( ( n − 1 ) Prt k + ) \mathbb{E}(\operatorname{End}^{l}_{n\operatorname{Prt}^{+}_{k}}(n\operatorname{Prt}^{+}_{k}))=\operatorname{End}^{l}_{(n-1)\operatorname{Prt}^{+}_{\operatorname{Bar}(A)}}((n-1)\operatorname{Prt}^{+}_{k}) E ( End n Prt k + l ( n Prt k + )) = End ( n − 1 ) Prt Bar ( A ) + l (( n − 1 ) Prt k + ) exists in Alg 1 ( n Prt k + ) \operatorname{Alg}_{1}(n\operatorname{Prt}^{+}_{k}) Alg 1 ( n Prt k + ) . Repeating this argument, we see that for 1 ≤ r ≤ n − 1 1\leq r\leq n-1 1 ≤ r ≤ n − 1 there exists E r ( End n Prt k + l ( n Prt k + ) ) \mathbb{E}^{r}(\operatorname{End}^{l}_{n\operatorname{Prt}^{+}_{k}}(n\operatorname{Prt}^{+}_{k})) E r ( End n Prt k + l ( n Prt k + )) in Alg 1 ( ( n + 1 − r ) Prt k + ) \operatorname{Alg}_{1}((n+1-r)\operatorname{Prt}^{+}_{k}) Alg 1 (( n + 1 − r ) Prt k + ) , whose underlying ( n − r ) Prt k + (n-r)\operatorname{Prt}^{+}_{k} ( n − r ) Prt k + -module is ( n − r ) Prt Bar r + 1 ( A ) + (n-r)\operatorname{Prt}^{+}_{\operatorname{Bar}^{r+1}(A)} ( n − r ) Prt Bar r + 1 ( A ) + . In particular, E n − 1 ( End n Prt k + l ( n Prt k + ) ) ≃ Prt Bar n ( A ) + = LMod LMod Bar n ( A ) ( Prt k + ) \mathbb{E}^{n-1}(\operatorname{End}^{l}_{n\operatorname{Prt}^{+}_{k}}(n\operatorname{Prt}^{+}_{k}))\simeq\operatorname{Prt}^{+}_{\operatorname{Bar}^{n}(A)}=\operatorname{LMod}_{\operatorname{LMod}_{\operatorname{Bar}^{n}(A)}(\operatorname{Prt}^{+}_{k})} E n − 1 ( End n Prt k + l ( n Prt k + )) ≃ Prt Bar n ( A ) + = LMod LMod Bar n ( A ) ( Prt k + ) in 2 Prt k + 2\operatorname{Prt}^{+}_{k} 2 Prt k + . Here LMod Bar n ( A ) \operatorname{LMod}_{\operatorname{Bar}^{n}(A)} LMod Bar n ( A ) is regarded as an object of Alg 1 ( Prt k + ) \operatorname{Alg}_{1}(\operatorname{Prt}^{+}_{k}) Alg 1 ( Prt k + ) . Thus, it is enough to show that there exists an endomorphism algebra of Mod k ∈ Prt Bar n ( A ) + \operatorname{Mod}_{k}\in\operatorname{Prt}^{+}_{\operatorname{Bar}^{n}(A)} Mod k ∈ Prt Bar n ( A ) + in Alg 1 ( Prt k + ) \operatorname{Alg}_{1}(\operatorname{Prt}^{+}_{k}) Alg 1 ( Prt k + ) . The existence follows from Lemma 4.12. Consequently, our assertion follows.
Lemma 4.12. Let B B B be an augmented E 2 \mathbb{E}_{2} E 2 -algebra, Prt B + = LMod LMod B ( Prt k + ) \operatorname{Prt}^{+}_{B}=\operatorname{LMod}_{\operatorname{LMod}_{B}(\operatorname{Prt}^{+}_{k})} Prt B + = LMod LMod B ( Prt k + ) in 2 Prt k + 2\operatorname{Prt}^{+}_{k} 2 Prt k + . There exists an endomorphism algebra of Mod k ∈ Prt B + \operatorname{Mod}_{k}\in\operatorname{Prt}^{+}_{B} Mod k ∈ Prt B + in Alg 1 ( Prt k + ) \operatorname{Alg}_{1}(\operatorname{Prt}^{+}_{k}) Alg 1 ( Prt k + ) .
Proof. There exists a Hom object from Mod k \operatorname{Mod}_{k} Mod k to itself in Prt k t \operatorname{Prt}^{t}_{k} Prt k t , defined by
H = ( Fun LMod B ( Mod k , Mod k ) , Fun LMod B ≤ 0 ( Mod k ≤ 0 , Mod k ≤ 0 ) ) . H=\left(\operatorname{Fun}_{\operatorname{LMod}_{B}}(\operatorname{Mod}_{k},\operatorname{Mod}_{k}),\operatorname{Fun}_{\operatorname{LMod}^{\leq0}_{B}}(\operatorname{Mod}^{\leq0}_{k},\operatorname{Mod}^{\leq0}_{k})\right). H = ( Fun LMod B ( Mod k , Mod k ) , Fun LMod B ≤ 0 ( Mod k ≤ 0 , Mod k ≤ 0 ) ) .
Note that it is generally not right complete (cf. [21 ] ). Passing to R ^ : Prt k t → Prt k t + \widehat{R}:\operatorname{Prt}^{t}_{k}\to\operatorname{Prt}^{t+}_{k} R : Prt k t → Prt k t + which is a right adjoint of the symmetric monoidal inclusion Prt k t + ⊂ Prt k t \operatorname{Prt}^{t+}_{k}\subset\operatorname{Prt}^{t}_{k} Prt k t + ⊂ Prt k t , we obtain R ^ ( H ) \widehat{R}(H) R ( H ) , that is a Hom object defined in Prt k t + \operatorname{Prt}^{t+}_{k} Prt k t + . Finally, by [24 4.7.1.40, 4.7.1.41] , the Hom object R ^ ( H ) \widehat{R}(H) R ( H ) is (uniquely) promoted to an endomorphism algebra defined in Alg 1 ( Prt k t + ) \operatorname{Alg}_{1}(\operatorname{Prt}^{t+}_{k}) Alg 1 ( Prt k t + ) .
Lemma 4.13. Let A A A and B B B be E n + 1 \mathbb{E}_{n+1} E n + 1 -algebras over k k k . Then the functor θ k , n + 1 → n + : Alg n + 1 ( Mod k ) → ( ( n + 1 ) Prt k t + ) n Prt k t + \theta^{+}_{k,n+1\to n}:\operatorname{Alg}_{n+1}(\operatorname{Mod}_{k})\to((n+1)\operatorname{Prt}^{t+}_{k})_{n\operatorname{Prt}^{t+}_{k}} θ k , n + 1 → n + : Alg n + 1 ( Mod k ) → (( n + 1 ) Prt k t + ) n Prt k t + given by R ↦ ( n Prt R t + , ( n − 1 ) Prt R t + ) R\mapsto(n\operatorname{Prt}^{t+}_{R},(n-1)\operatorname{Prt}^{t+}_{R}) R ↦ ( n Prt R t + , ( n − 1 ) Prt R t + ) induces an equivalence
Map Alg n + 1 ( Mod k ) ( A , B ) → Map ( ( n + 1 ) Prt k t + ) n Prt k t + ( ( n Prt A t ± , ( n − 1 ) Prt A t + ) , ( n Prt B t + , ( n − 1 ) Prt B t + ) ) . \operatorname{Map}_{\operatorname{Alg}_{n+1}(\operatorname{Mod}_{k})}(A,B)\to\operatorname{Map}_{((n+1)\operatorname{Prt}^{t+}_{k})_{n\operatorname{Prt}^{t+}_{k}}}((n\operatorname{Prt}^{t\pm}_{A},(n-1)\operatorname{Prt}^{t+}_{A}),(n\operatorname{Prt}^{t+}_{B},(n-1)\operatorname{Prt}^{t+}_{B})). Map Alg n + 1 ( Mod k ) ( A , B ) → Map (( n + 1 ) Prt k t + ) n Prt k t + (( n Prt A t ± , ( n − 1 ) Prt A t + ) , ( n Prt B t + , ( n − 1 ) Prt B t + )) . Suppose that A A A and B B B are promoted to E n + 2 \mathbb{E}_{n+2} E n + 2 -algebras. Then the functor Alg 1 ( θ k , n + 1 → n + ) : Alg n + 2 ( Mod k ) → Alg 1 ( ( n + 1 ) Prt k t + ) \operatorname{Alg}_{1}(\theta_{k,n+1\to n}^{+}):\operatorname{Alg}_{n+2}(\operatorname{Mod}_{k})\to\operatorname{Alg}_{1}((n+1)\operatorname{Prt}^{t+}_{k}) Alg 1 ( θ k , n + 1 → n + ) : Alg n + 2 ( Mod k ) → Alg 1 (( n + 1 ) Prt k t + ) given by R ↦ n Prt R t + R\mapsto n\operatorname{Prt}^{t+}_{R} R ↦ n Prt R t + induces an equivalence
Map Alg n + 2 ( Mod k ) ( A , B ) → Map Alg 1 ( ( n + 1 ) Prt k t + ) ( n Prt A t + , n Prt B t + ) . \operatorname{Map}_{\operatorname{Alg}_{n+2}(\operatorname{Mod}_{k})}(A,B)\to\operatorname{Map}_{\operatorname{Alg}_{1}((n+1)\operatorname{Prt}^{t+}_{k})}(n\operatorname{Prt}^{t+}_{A},n\operatorname{Prt}^{t+}_{B}). Map Alg n + 2 ( Mod k ) ( A , B ) → Map Alg 1 (( n + 1 ) Prt k t + ) ( n Prt A t + , n Prt B t + ) . Suppose furthermore that B B B is connective. Then the functor Alg 1 ( θ k , n + 1 → n ± ) : Alg n + 2 ( Mod k ) → Alg 1 ( ( n + 1 ) Prt k t ± ) \operatorname{Alg}_{1}(\theta_{k,n+1\to n}^{\pm}):\operatorname{Alg}_{n+2}(\operatorname{Mod}_{k})\to\operatorname{Alg}_{1}((n+1)\operatorname{Prt}^{t\pm}_{k}) Alg 1 ( θ k , n + 1 → n ± ) : Alg n + 2 ( Mod k ) → Alg 1 (( n + 1 ) Prt k t ± ) given by R ↦ n Prt R t ± R\mapsto n\operatorname{Prt}^{t\pm}_{R} R ↦ n Prt R t ± induces an equivalence
Map Alg n + 2 ( Mod k ) ( A , B ) → Map Alg 1 ( ( n + 1 ) Prt k t ± ) ( n Prt A t ± , n Prt B t ± ) . \operatorname{Map}_{\operatorname{Alg}_{n+2}(\operatorname{Mod}_{k})}(A,B)\to\operatorname{Map}_{\operatorname{Alg}_{1}((n+1)\operatorname{Prt}^{t\pm}_{k})}(n\operatorname{Prt}^{t\pm}_{A},n\operatorname{Prt}^{t\pm}_{B}). Map Alg n + 2 ( Mod k ) ( A , B ) → Map Alg 1 (( n + 1 ) Prt k t ± ) ( n Prt A t ± , n Prt B t ± ) .
Proof. Recall the sequence of adjoint pairs from the proof of Lemma 4.8:
Alg n + 1 ( Mod k ) ⇄ ? θ 0 Prt k t + , 0 + Alg n ( Prt k t + ) ⇄ ? U ( ( n + 1 ) Prt k t ± ) n Prt k t + ∗ ∗ . \operatorname{Alg}_{n+1}(\operatorname{Mod}_{k})\mathop{\rightleftarrows}^{\theta^{+}_{0\operatorname{Prt}^{t+}_{k},0}}_{\operatorname{?}}\operatorname{Alg}_{n}(\operatorname{Prt}^{t+}_{k})\mathop{\rightleftarrows}^{U}_{\operatorname{?}}((n+1)\operatorname{Prt}^{t\pm}_{k})^{\ast\ast}_{n\operatorname{Prt}^{t+}_{k}}. Alg n + 1 ( Mod k ) ⇄ ? θ 0 Prt k t + , 0 + Alg n ( Prt k t + ) ⇄ ? U (( n + 1 ) Prt k t ± ) n Prt k t + ∗∗ .
Here by abuse of notation we write θ 0 Prt k t + , 0 \theta_{0\operatorname{Prt}^{t+}_{k},0} θ 0 Prt k t + , 0 for Alg n ( θ 0 Prt k t + , 0 ) \operatorname{Alg}_{n}(\theta_{0\operatorname{Prt}^{t+}_{k},0}) Alg n ( θ 0 Prt k t + , 0 ) . The functor in the statement is the composite U ∘ θ 0 Prt k t + , 0 U\circ\theta_{0\operatorname{Prt}^{t+}_{k},0} U ∘ θ 0 Prt k t + , 0 of left adjoint functors. The first assertion follows from the fact that θ 0 Prt k t + , 0 \theta_{0\operatorname{Prt}^{t+}_{k},0} θ 0 Prt k t + , 0 and U U U are fully faithful. Applying Alg 1 ( − ) \operatorname{Alg}_{1}(-) Alg 1 ( − ) to the sequence of the above adjoint pairs, we also see the second assertion.
It remains to prove the final assertion: We assume that B B B is connective. Consider a variant of the above sequence of adjoint pairs:
Alg n + 2 ( Mod k ) ⇄ ? θ 0 Prt k t + , 1 + Alg n + 1 ( Prt k t + ) ⇄ ? L ^ Alg n + 1 ( Prt k t ± ) ⇄ ? U ± Alg 1 ( ( n + 1 ) Prt k t ± ) ∗ ∗ . \operatorname{Alg}_{n+2}(\operatorname{Mod}_{k})\mathop{\rightleftarrows}^{\theta^{+}_{0\operatorname{Prt}^{t+}_{k},1}}_{\operatorname{?}}\operatorname{Alg}_{n+1}(\operatorname{Prt}^{t+}_{k})\mathop{\rightleftarrows}^{\widehat{L}}_{\operatorname{?}}\operatorname{Alg}_{n+1}(\operatorname{Prt}^{t\pm}_{k})\mathop{\rightleftarrows}^{U^{\pm}}_{\operatorname{?}}\operatorname{Alg}_{1}((n+1)\operatorname{Prt}^{t\pm}_{k})^{\ast\ast}. Alg n + 2 ( Mod k ) ⇄ ? θ 0 Prt k t + , 1 + Alg n + 1 ( Prt k t + ) ⇄ ? L Alg n + 1 ( Prt k t ± ) ⇄ ? U ± Alg 1 (( n + 1 ) Prt k t ± ) ∗∗ .
The left adjoint U ± U^{\pm} U ± is a Prt t ± \operatorname{Prt}^{t\pm} Prt t ± -version of U ′ U^{\prime} U ′ , that is defined in a similar way. Here by abuse of notation we write θ 0 Prt k t + , 1 \theta_{0\operatorname{Prt}^{t+}_{k},1} θ 0 Prt k t + , 1 and L ^ \widehat{L} L for Alg 1 ( θ 0 Prt k t + , 1 ) \operatorname{Alg}_{1}(\theta_{0\operatorname{Prt}^{t+}_{k},1}) Alg 1 ( θ 0 Prt k t + , 1 ) and Alg 1 ( L ^ ) \operatorname{Alg}_{1}(\widehat{L}) Alg 1 ( L ) , respectively. The map of mapping spaces is induced by the composite of left adjoint functors. Note that the functors θ Prt k , + t + 1 \theta_{\operatorname{Prt}_{k,+}^{\operatorname{t}+1}} θ Prt k , + t + 1 and U ± U^{\pm} U ± are fully faithful. Thus, it is enough to show that L ^ \widehat{L} L induces an equivalence
Map Alg n + 1 ( Prt k t + ) ( LMod A , LMod B ) ⟶ Map Alg n + 1 ( Prt k t ± ) ( L ^ ( LMod A ) , L ^ ( LMod B ) ) . \operatorname{Map}_{\operatorname{Alg}_{n+1}(\operatorname{Prt}_{k}^{\operatorname{t}+})}(\operatorname{LMod}_{A},\operatorname{LMod}_{B}) \longrightarrow\operatorname{Map}_{\operatorname{Alg}_{n+1}(\operatorname{Prt}_{k}^{\operatorname{t}\pm})}(\widehat{L}(\operatorname{LMod}_{A}),\widehat{L}(\operatorname{LMod}_{B})). Map Alg n + 1 ( Prt k t + ) ( LMod A , LMod B ) ⟶ Map Alg n + 1 ( Prt k t ± ) ( L ( LMod A ) , L ( LMod B )) .
Indeed, the induced map is an equivalence since the middle adjoint pair is a localization (whose right adjoint is fully faithful), and by the connectivity of B B B , the category LMod B \operatorname{LMod}_{B} LMod B lies in Alg n + 1 ( Prt k t ± ) \operatorname{Alg}_{n+1}(\operatorname{Prt}_{k}^{\operatorname{t}\pm}) Alg n + 1 ( Prt k t ± ) . □ \square □
Proof of Proposition 4.5. The existence of algebraizations of (1) and (2) follows from Lemma 4.8 and Lemma 4.11 .
In the case (1), there exist C ∈ Alg n + 1 ( Mod k ) C \in\operatorname{Alg}_{n+1}(\operatorname{Mod}_{k}) C ∈ Alg n + 1 ( Mod k ) and an algebraization functor α : ( n Prt C t + , ( n − 1 ) Prt C t + ) → ( n Prt B t + , ( n − 1 ) Prt k t + ) \alpha\colon(n\operatorname{Prt}_{C}^{\operatorname{t}+},(n-1)\operatorname{Prt}_{C}^{\operatorname{t}+}) \to(n\operatorname{Prt}_{B}^{\operatorname{t}+},(n-1)\operatorname{Prt}_{k}^{\operatorname{t}+}) α : ( n Prt C t + , ( n − 1 ) Prt C t + ) → ( n Prt B t + , ( n − 1 ) Prt k t + ) . We will prove that C C C is equivalent to D n + 1 ( B ) \mathbb{D}_{n+1}(B) D n + 1 ( B ) . Consider the composition P C : n Prt C t + → α n Prt B t + → n Prt k t + P_{C} \colon n\operatorname{Prt}_{C}^{\operatorname{t}+} \xrightarrow{\alpha} n\operatorname{Prt}_{B}^{\operatorname{t}+} \to n\operatorname{Prt}_{k}^{\operatorname{t}+} P C : n Prt C t + α n Prt B t + → n Prt k t + with the forgetful functor P B : n Prt B t + → n Prt k t + P_{B} \colon n\operatorname{Prt}_{B}^{\operatorname{t}+} \to n\operatorname{Prt}_{k}^{\operatorname{t}+} P B : n Prt B t + → n Prt k t + . Note that P C P_{C} P C carries ( n − 1 ) Prt C t + (n-1)\operatorname{Prt}_{C}^{\operatorname{t}+} ( n − 1 ) Prt C t + to ( n − 1 ) Prt k t + (n-1)\operatorname{Prt}_{k}^{\operatorname{t}+} ( n − 1 ) Prt k t + . From Lemma 4.13 we deduce that there exists an essentially unique augmentation p C : C → k p_{C} \colon C \to k p C : C → k that induces/corresponds to P C P_{C} P C . By Lemma 4.13 , for any D ∈ Alg n + 1 ( Mod k ) D \in\operatorname{Alg}_{n+1}(\operatorname{Mod}_{k}) D ∈ Alg n + 1 ( Mod k ) there exists an equivalence in S ^ \widehat{\mathcal{S}} S :
Map Alg n + 1 ( Mod k ) ( B ⊗ k D , k ) ≃ Map Alg 1 ( ( n + 1 ) Prt k t + ) ( n Prt B t + ⊗ n Prt D t + , n Prt k t + ) . \operatorname{Map}_{\operatorname{Alg}_{n+1}(\operatorname{Mod}_{k})}(B \otimes_{k} D,k) \simeq\operatorname{Map}_{\operatorname{Alg}_{1}((n+1)\operatorname{Prt}_{k}^{\operatorname{t}+})} (n\operatorname{Prt}_{B}^{\operatorname{t}+} \otimes n\operatorname{Prt}_{D}^{\operatorname{t}+},n\operatorname{Prt}_{k}^{\operatorname{t}+}). Map Alg n + 1 ( Mod k ) ( B ⊗ k D , k ) ≃ Map Alg 1 (( n + 1 ) Prt k t + ) ( n Prt B t + ⊗ n Prt D t + , n Prt k t + ) .
By Lemma 4.13 again, we have equivalences Map Alg n + 1 ( Mod k ) ( B , k ) ≃ Map Alg 1 ( ( n + 1 ) Prt k t + ) ( n Prt B t + , n Prt k t + ) \operatorname{Map}_{\operatorname{Alg}_{n+1}(\operatorname{Mod}_{k})}(B,k) \simeq\operatorname{Map}_{\operatorname{Alg}_{1}((n+1)\operatorname{Prt}_{k}^{\operatorname{t}+})}(n\operatorname{Prt}_{B}^{\operatorname{t}+},n\operatorname{Prt}_{k}^{\operatorname{t}+}) Map Alg n + 1 ( Mod k ) ( B , k ) ≃ Map Alg 1 (( n + 1 ) Prt k t + ) ( n Prt B t + , n Prt k t + ) and Map Alg n + 1 ( Mod k ) ( D , k ) ≃ Map Alg 1 ( ( n + 1 ) Prt k t + ) ( n Prt D t + , n Prt k t + ) \operatorname{Map}_{\operatorname{Alg}_{n+1}(\operatorname{Mod}_{k})}(D,k) \simeq\operatorname{Map}_{\operatorname{Alg}_{1}((n+1)\operatorname{Prt}_{k}^{\operatorname{t}+})}(n\operatorname{Prt}_{D}^{\operatorname{t}+},n\operatorname{Prt}_{k}^{\operatorname{t}+}) Map Alg n + 1 ( Mod k ) ( D , k ) ≃ Map Alg 1 (( n + 1 ) Prt k t + ) ( n Prt D t + , n Prt k t + ) . Let p D : D → k p_{D} \colon D \to k p D : D → k be an augmentation. By these equivalences,
Pair ( p B , p D ) = Map Alg n + 1 ( Mod k ) ( B ⊗ k D , k ) × Map ( B , k ) × Map ( D , k ) { ( p A , p D ) } \operatorname{Pair}(p_{B},p_{D})=\operatorname{Map}_{\operatorname{Alg}_{n+1}(\operatorname{Mod}_{k})}(B\otimes_{k}D,k)\mathop{\times}_{\operatorname{Map}(B,k)\times\operatorname{Map}(D,k)}\{(p_{A},p_{D})\} Pair ( p B , p D ) = Map Alg n + 1 ( Mod k ) ( B ⊗ k D , k ) × Map ( B , k ) × Map ( D , k ) {( p A , p D )}
is equivalent to
M = Map Alg 1 ( n Prt k t + ) ( ( n − 1 ) Prt B t + ⊗ ( n − 1 ) Prt D t + , ( n − 1 ) Prt k t + ) × Map ( ( n − 1 ) Prt k t + , ( n − 1 ) Prt k t + ) × Map ( ( n − 1 ) Prt k t + , ( n − 1 ) Prt k t + ) { ( P B , P D ) } . \mathcal{M}=\operatorname{Map}_{\operatorname{Alg}_{1}(n\operatorname{Prt}_{k}^{\operatorname{t}+})}((n-1)\operatorname{Prt}_{B}^{\operatorname{t}+}\otimes(n-1)\operatorname{Prt}_{D}^{\operatorname{t}+},(n-1)\operatorname{Prt}_{k}^{\operatorname{t}+})\mathop{\times}_{\operatorname{Map}_{((n-1)\operatorname{Prt}_{k}^{\operatorname{t}+},(n-1)\operatorname{Prt}_{k}^{\operatorname{t}+})}\times\operatorname{Map}_{((n-1)\operatorname{Prt}_{k}^{\operatorname{t}+},(n-1)\operatorname{Prt}_{k}^{\operatorname{t}+})}\{(P_{B},P_{D})\}.} M = Map Alg 1 ( n Prt k t + ) (( n − 1 ) Prt B t + ⊗ ( n − 1 ) Prt D t + , ( n − 1 ) Prt k t + ) × Map (( n − 1 ) Prt k t + , ( n − 1 ) Prt k t + ) × Map (( n − 1 ) Prt k t + , ( n − 1 ) Prt k t + ) {( P B , P D )} .
where P D : ( n − 1 ) Prt D t + → ( n − 1 ) Prt k t + P_{D} \colon(n-1)\operatorname{Prt}_{D}^{\operatorname{t}+} \to(n-1)\operatorname{Prt}_{k}^{\operatorname{t}+} P D : ( n − 1 ) Prt D t + → ( n − 1 ) Prt k t + is induced by p D p_{D} p D (we omit some subscripts in Map \operatorname{Map} Map and also omit specified objects from the notation). Since End ( n − 1 ) Prt B t + l ( ( n − 1 ) Prt k t + ) ∈ Alg 1 ( n Prt k t + ) \operatorname{End}^{l}_{(n-1)\operatorname{Prt}_{B}^{\operatorname{t}+}}((n-1)\operatorname{Prt}_{k}^{\operatorname{t}+}) \in\operatorname{Alg}_{1}(n\operatorname{Prt}_{k}^{\operatorname{t}+}) End ( n − 1 ) Prt B t + l (( n − 1 ) Prt k t + ) ∈ Alg 1 ( n Prt k t + ) is an E 1 \mathbb{E}_{1} E 1 -Koszul dual of ( n − 1 ) Prt B t + (n-1)\operatorname{Prt}_{B}^{\operatorname{t}+} ( n − 1 ) Prt B t + in Alg 1 ( n Prt k t + ) \operatorname{Alg}_{1}(n\operatorname{Prt}_{k}^{\operatorname{t}+}) Alg 1 ( n Prt k t + ) for n ≥ 2 n \ge2 n ≥ 2 , the latter space M \mathcal{M} M is equivalent to Map Alg 1 ( n Prt k t + ) ( ( n − 1 ) Prt D t + , End ( n − 1 ) Prt B t + l ( ( n − 1 ) Prt k t + ) ) \operatorname{Map}_{\operatorname{Alg}_{1}(n\operatorname{Prt}_{k}^{\operatorname{t}+})}((n-1)\operatorname{Prt}_{D}^{\operatorname{t}+},\operatorname{End}^{l}_{(n-1)\operatorname{Prt}_{B}^{\operatorname{t}+}}((n-1)\operatorname{Prt}_{k}^{\operatorname{t}+})) Map Alg 1 ( n Prt k t + ) (( n − 1 ) Prt D t + , End ( n − 1 ) Prt B t + l (( n − 1 ) Prt k t + )) . Since End ( n − 1 ) Prt B t + l ( ( n − 1 ) Prt k t + ) \operatorname{End}^{l}_{(n-1)\operatorname{Prt}_{B}^{\operatorname{t}+}}((n-1)\operatorname{Prt}_{k}^{\operatorname{t}+}) End ( n − 1 ) Prt B t + l (( n − 1 ) Prt k t + ) has an algebraization C C C , we have
Map Alg 1 ( n Prt k t + ) ( ( n − 1 ) Prt D t + , End ( n − 1 ) Prt B t + l ( ( n − 1 ) Prt k t + ) ) ≃ Map Alg n + 1 ( Mod k ) ( D , C ) . \operatorname{Map}_{\operatorname{Alg}_{1}(n\operatorname{Prt}_{k}^{\operatorname{t}+})}((n-1)\operatorname{Prt}_{D}^{\operatorname{t}+},\operatorname{End}^{l}_{(n-1)\operatorname{Prt}_{B}^{\operatorname{t}+}}((n-1)\operatorname{Prt}_{k}^{\operatorname{t}+}))\simeq\operatorname{Map}_{\operatorname{Alg}_{n+1}(\operatorname{Mod}_{k})}(D,C). Map Alg 1 ( n Prt k t + ) (( n − 1 ) Prt D t + , End ( n − 1 ) Prt B t + l (( n − 1 ) Prt k t + )) ≃ Map Alg n + 1 ( Mod k ) ( D , C ) .
When n = 1 n=1 n = 1 , we have
Map Alg 1 ( Prt k t + ) ( Mod D , End Mod B l ( Mod k ) ) ≃ Map Alg 1 ( Prt k t + ) ( Mod D , R ^ ( End Mod B l ( Mod k ) ) ) ≃ Map Alg 2 ( Mod k ) ( D , C ) . \operatorname{Map}_{\operatorname{Alg}_{1}(\operatorname{Prt}_{k}^{\operatorname{t}+})}(\operatorname{Mod}_{D},\operatorname{End}^{l}_{\operatorname{Mod}_{B}}(\operatorname{Mod}_{k}))\simeq\operatorname{Map}_{\operatorname{Alg}_{1}(\operatorname{Prt}_{k}^{\operatorname{t}+})}(\operatorname{Mod}_{D},\widehat{R}(\operatorname{End}^{l}_{\operatorname{Mod}_{B}}(\operatorname{Mod}_{k})))\simeq\operatorname{Map}_{\operatorname{Alg}_{2}(\operatorname{Mod}_{k})}(D,C). Map Alg 1 ( Prt k t + ) ( Mod D , End Mod B l ( Mod k )) ≃ Map Alg 1 ( Prt k t + ) ( Mod D , R ( End Mod B l ( Mod k ))) ≃ Map Alg 2 ( Mod k ) ( D , C ) .
Combined with Map Alg n + 1 ( Mod k ) ( D , k ) ≃ Map Alg 1 ( n Prt k t + ) ( ( n − 1 ) Prt D t + , ( n − 1 ) Prt k t + ) \operatorname{Map}_{\operatorname{Alg}_{n+1}(\operatorname{Mod}_{k})}(D,k)\simeq\operatorname{Map}_{\operatorname{Alg}_{1}(n\operatorname{Prt}_{k}^{\operatorname{t}+})}((n-1)\operatorname{Prt}_{D}^{\operatorname{t}+},(n-1)\operatorname{Prt}_{k}^{\operatorname{t}+}) Map Alg n + 1 ( Mod k ) ( D , k ) ≃ Map Alg 1 ( n Prt k t + ) (( n − 1 ) Prt D t + , ( n − 1 ) Prt k t + ) , we have
Pair ( p B , p D ) ≃ Map Alg 1 + ( n Prt k t + ) ( ( n − 1 ) Prt D t + , End ( n − 1 ) Prt B t + l ( ( n − 1 ) Prt k t + ) ) ≃ Map Alg n + 1 + ( Mod k ) ( ( D , p D ) , ( C , p C ) ) . \operatorname{Pair}(p_{B},p_{D})\simeq\operatorname{Map}_{\operatorname{Alg}_{1}^{+}(n\operatorname{Prt}_{k}^{\operatorname{t}+})}((n-1)\operatorname{Prt}_{D}^{\operatorname{t}+},\operatorname{End}^{l}_{(n-1)\operatorname{Prt}_{B}^{\operatorname{t}+}}((n-1)\operatorname{Prt}_{k}^{\operatorname{t}+}))\simeq\operatorname{Map}_{\operatorname{Alg}_{n+1}^{+}(\operatorname{Mod}_{k})}((D,p_{D}),(C,p_{C})). Pair ( p B , p D ) ≃ Map Alg 1 + ( n Prt k t + ) (( n − 1 ) Prt D t + , End ( n − 1 ) Prt B t + l (( n − 1 ) Prt k t + )) ≃ Map Alg n + 1 + ( Mod k ) (( D , p D ) , ( C , p C )) .
It follows that ( C , p C ) ∈ Alg n + 1 + ( Mod k ) (C,p_{C})\in\operatorname{Alg}_{n+1}^{+}(\operatorname{Mod}_{k}) ( C , p C ) ∈ Alg n + 1 + ( Mod k ) is an E n + 1 \mathbb{E}_{n+1} E n + 1 -Koszul dual D n + 1 ( B ) \mathbb{D}_{n+1}(B) D n + 1 ( B ) of B B B .
We prove (2). First, by Lemma 4.10 , there exists an endomorphism algebra End n Prt A t + l ( n Prt A t + ) \operatorname{End}^{l}_{n\operatorname{Prt}_{A}^{\operatorname{t}+}}(n\operatorname{Prt}_{A}^{\operatorname{t}+}) End n Prt A t + l ( n Prt A t + ) . The following argument is similar to (1). In the case (2), there exist C ∈ Alg n + 2 ( Mod k ) C\in\operatorname{Alg}_{n+2}(\operatorname{Mod}_{k}) C ∈ Alg n + 2 ( Mod k ) and an algebraization functor n Prt C t + → End n Prt A t + l ( n Prt k t + ) n\operatorname{Prt}_{C}^{\operatorname{t}+}\to\operatorname{End}^{l}_{n\operatorname{Prt}_{A}^{\operatorname{t}+}}(n\operatorname{Prt}_{k}^{\operatorname{t}+}) n Prt C t + → End n Prt A t + l ( n Prt k t + ) . It remains to prove that C C C is equivalent to D n + 2 ( A ) \mathbb{D}_{n+2}(A) D n + 2 ( A ) . Consider the composite P C : n Prt C t + → End n Prt A t + l ( n Prt k t + ) → End n Prt A t + l ( n Prt k t + ) = n Prt k t + P_{C}\colon n\operatorname{Prt}_{C}^{\operatorname{t}+}\to\operatorname{End}^{l}_{n\operatorname{Prt}_{A}^{\operatorname{t}+}}(n\operatorname{Prt}_{k}^{\operatorname{t}+})\to\operatorname{End}^{l}_{n\operatorname{Prt}_{A}^{\operatorname{t}+}}(n\operatorname{Prt}_{k}^{\operatorname{t}+})=n\operatorname{Prt}_{k}^{\operatorname{t}+} P C : n Prt C t + → End n Prt A t + l ( n Prt k t + ) → End n Prt A t + l ( n Prt k t + ) = n Prt k t + induced by the restriction along the augmentation P A : n Prt A t + → n Prt k t + P_{A}\colon n\operatorname{Prt}_{A}^{\operatorname{t}+}\to n\operatorname{Prt}_{k}^{\operatorname{t}+} P A : n Prt A t + → n Prt k t + . From Lemma 4.13 , we note that there exists an essentially unique augmentation p C : C → k p_{C}\colon C\to k p C : C → k that induces/corresponds to P C P_{C} P C . By Lemma 4.13 , there exists an equivalence in S ^ \widehat{\mathcal{S}} S . Let D ∈ Alg n + 2 ( Mod k ) D\in\operatorname{Alg}_{n+2}(\operatorname{Mod}_{k}) D ∈ Alg n + 2 ( Mod k ) .
Map Alg n + 2 ( Mod k ) ( A ⊗ k D , k ) ≃ Map Alg 1 ( ( n + 1 ) Prt k t + ) ( n Prt A t + ⊗ n Prt k t + n Prt D t + , n Prt k t + ) \operatorname{Map}_{\operatorname{Alg}_{n+2}(\operatorname{Mod}_{k})}(A\otimes_{k}D,k)\simeq\operatorname{Map}_{\operatorname{Alg}_{1}((n+1)\operatorname{Prt}_{k}^{\operatorname{t}+})}(n\operatorname{Prt}_{A}^{\operatorname{t}+}\otimes n\operatorname{Prt}_{k}^{\operatorname{t}+}n\operatorname{Prt}_{D}^{\operatorname{t}+},n\operatorname{Prt}_{k}^{\operatorname{t}+}) Map Alg n + 2 ( Mod k ) ( A ⊗ k D , k ) ≃ Map Alg 1 (( n + 1 ) Prt k t + ) ( n Prt A t + ⊗ n Prt k t + n Prt D t + , n Prt k t + )
By Lemma 4.13 again, we have equivalences Map Alg n + 2 ( Mod k ) ( A , k ) ≃ Map Alg 1 ( ( n + 1 ) Prt k t + ) ( n Prt A t + , n Prt k t + ) \operatorname{Map}_{\operatorname{Alg}_{n+2}(\operatorname{Mod}_{k})}(A,k)\simeq\operatorname{Map}_{\operatorname{Alg}_{1}((n+1)\operatorname{Prt}_{k}^{\operatorname{t}+})}(n\operatorname{Prt}_{A}^{\operatorname{t}+},n\operatorname{Prt}_{k}^{\operatorname{t}+}) Map Alg n + 2 ( Mod k ) ( A , k ) ≃ Map Alg 1 (( n + 1 ) Prt k t + ) ( n Prt A t + , n Prt k t + ) and Map Alg n + 2 ( Mod k ) ( D , k ) ≃ Map Alg 1 ( ( n + 1 ) Prt k t + ) ( n Prt D t + , n Prt k t + ) \operatorname{Map}_{\operatorname{Alg}_{n+2}(\operatorname{Mod}_{k})}(D,k)\simeq\operatorname{Map}_{\operatorname{Alg}_{1}((n+1)\operatorname{Prt}_{k}^{\operatorname{t}+})}(n\operatorname{Prt}_{D}^{\operatorname{t}+},n\operatorname{Prt}_{k}^{\operatorname{t}+}) Map Alg n + 2 ( Mod k ) ( D , k ) ≃ Map Alg 1 (( n + 1 ) Prt k t + ) ( n Prt D t + , n Prt k t + ) . Let p D : D → k p_{D}\colon D\to k p D : D → k be an augmentation. By these equivalences,
Pair ( p A , p D ) = Map Alg n + 2 ( Mod k ) ( A ⊗ k D , k ) × Map ( A , k ) × Map ( C , k ) { ( p A , p D ) } \operatorname{Pair}(p_A,p_D)=\operatorname{Map}_{\operatorname{Alg}_{n+2}(\operatorname{Mod}_k)}(A\otimes_k D,k)\times_{\operatorname{Map}(A,k)\times\operatorname{Map}(C,k)}\{(p_A,p_D)\} Pair ( p A , p D ) = Map Alg n + 2 ( Mod k ) ( A ⊗ k D , k ) × Map ( A , k ) × Map ( C , k ) {( p A , p D )}
is equivalent to
M = Map Alg 1 ( ( n + 1 ) Pr k t + ) ( n Pr A t + ⊗ n Pr D t + , n Pr k t + ) × Map ( n Pr A t + , n Pr k t + ) × Map ( n Pr D t + , n Pr k t + ) { ( P A , P D ) } \mathrm{M}=\operatorname{Map}_{\operatorname{Alg}_1((n+1)\operatorname{Pr}_k^{t+})}(n\operatorname{Pr}_{A}^{t+}\otimes n\operatorname{Pr}_{D}^{t+},n\operatorname{Pr}_{k}^{t+})\times_{\operatorname{Map}(n\operatorname{Pr}_{A}^{t+},n\operatorname{Pr}_{k}^{t+})\times\operatorname{Map}(n\operatorname{Pr}_{D}^{t+},n\operatorname{Pr}_{k}^{t+})}\{(P_A,P_D)\} M = Map Alg 1 (( n + 1 ) Pr k t + ) ( n Pr A t + ⊗ n Pr D t + , n Pr k t + ) × Map ( n Pr A t + , n Pr k t + ) × Map ( n Pr D t + , n Pr k t + ) {( P A , P D )}
where P D : n Pr D t + → n Pr k t + P_D:n\operatorname{Pr}_{D}^{t+}\to n\operatorname{Pr}_{k}^{t+} P D : n Pr D t + → n Pr k t + is induced by p D p_D p D (we omit some subscripts in Map). Since End n Pr A t + l ( n Pr k t + ) ∈ Alg 1 ( ( n + 1 ) Pr k t + ) \operatorname{End}^{l}_{n\operatorname{Pr}_{A}^{t+}}(n\operatorname{Pr}_{k}^{t+})\in\operatorname{Alg}_1((n+1)\operatorname{Pr}_{k}^{t+}) End n Pr A t + l ( n Pr k t + ) ∈ Alg 1 (( n + 1 ) Pr k t + ) is an E 1 \mathbf{E}_1 E 1 -Koszul dual of n Pr A t + n\operatorname{Pr}_{A}^{t+} n Pr A t + in Alg 1 ( ( n + 1 ) Pr k t + ) \operatorname{Alg}_1((n+1)\operatorname{Pr}_{k}^{t+}) Alg 1 (( n + 1 ) Pr k t + ) , then M \mathrm{M} M is equivalent to Map Alg 1 + ( ( n + 1 ) Pr k t + ) ( n Pr D t + , End n Pr A t + l ( n Pr k t + ) ) \operatorname{Map}_{\operatorname{Alg}_1^{+}((n+1)\operatorname{Pr}_{k}^{t+})}(n\operatorname{Pr}_{D}^{t+},\operatorname{End}^{l}_{n\operatorname{Pr}_{A}^{t+}}(n\operatorname{Pr}_{k}^{t+})) Map Alg 1 + (( n + 1 ) Pr k t + ) ( n Pr D t + , End n Pr A t + l ( n Pr k t + )) . Since End n Pr A t + l ( n Pr k t + ) \operatorname{End}^{l}_{n\operatorname{Pr}_{A}^{t+}}(n\operatorname{Pr}_{k}^{t+}) End n Pr A t + l ( n Pr k t + ) has an algebraization C C C , we have
Map Alg 1 ( ( n + 1 ) Pr k t + ) ( n Pr D t + , End n Pr A t + l ( n Pr k t + ) ) ≃ Map Alg n + 2 ( Mod k ) ( D , C ) . \operatorname{Map}_{\operatorname{Alg}_1((n+1)\operatorname{Pr}_{k}^{t+})}(n\operatorname{Pr}_{D}^{t+},\operatorname{End}^{l}_{n\operatorname{Pr}_{A}^{t+}}(n\operatorname{Pr}_{k}^{t+}))\simeq\operatorname{Map}_{\operatorname{Alg}_{n+2}(\operatorname{Mod}_k)}(D,C). Map Alg 1 (( n + 1 ) Pr k t + ) ( n Pr D t + , End n Pr A t + l ( n Pr k t + )) ≃ Map Alg n + 2 ( Mod k ) ( D , C ) .
Combined with Map Alg n + 2 ( Mod k ) ( D , k ) ≃ Map Alg 1 ( ( n + 1 ) Pr k t + ) ( n Pr D t + , n Pr k t + ) \operatorname{Map}_{\operatorname{Alg}_{n+2}(\operatorname{Mod}_k)}(D,k)\simeq\operatorname{Map}_{\operatorname{Alg}_1((n+1)\operatorname{Pr}_{k}^{t+})}(n\operatorname{Pr}_{D}^{t+},n\operatorname{Pr}_{k}^{t+}) Map Alg n + 2 ( Mod k ) ( D , k ) ≃ Map Alg 1 (( n + 1 ) Pr k t + ) ( n Pr D t + , n Pr k t + ) , we have
Pair ( p A , p D ) ≃ Map Alg 1 + ( ( n + 1 ) Pr k t + ) ( n Pr D t + , End n Pr A t + l ( n Pr k t + ) ) ≃ Map Alg n + 2 + ( Mod k ) ( ( D , p D ) , ( C , p C ) ) . \operatorname{Pair}(p_A,p_D)\simeq\operatorname{Map}_{\operatorname{Alg}_1^{+}((n+1)\operatorname{Pr}_{k}^{t+})}(n\operatorname{Pr}_{D}^{t+},\operatorname{End}^{l}_{n\operatorname{Pr}_{A}^{t+}}(n\operatorname{Pr}_{k}^{t+}))\simeq\operatorname{Map}_{\operatorname{Alg}_{n+2}^{+}(\operatorname{Mod}_k)}((D,p_D),(C,p_C)). Pair ( p A , p D ) ≃ Map Alg 1 + (( n + 1 ) Pr k t + ) ( n Pr D t + , End n Pr A t + l ( n Pr k t + )) ≃ Map Alg n + 2 + ( Mod k ) (( D , p D ) , ( C , p C )) .
It follows that ( C , p C ) ∈ Alg n + 2 + ( Mod k ) (C,p_C)\in\operatorname{Alg}_{n+2}^{+}(\operatorname{Mod}_k) ( C , p C ) ∈ Alg n + 2 + ( Mod k ) is an E n + 2 \mathbf{E}_{n+2} E n + 2 -Koszul dual D n + 2 ( A ) \mathbb{D}_{n+2}(A) D n + 2 ( A ) of A A A . □
Remark 4.14. From the final portion of the above proof, the algebraization functor n Pr D n + 2 ( A ) t + → End n Pr A t + l ( n Pr k t + ) n\operatorname{Pr}_{\mathbb{D}_{n+2}(A)}^{t+}\to\operatorname{End}^{l}_{n\operatorname{Pr}_{A}^{t+}}(n\operatorname{Pr}_{k}^{t+}) n Pr D n + 2 ( A ) t + → End n Pr A t + l ( n Pr k t + ) corresponds to the pairing n Pr A t + ⊗ n Pr k t + n Pr D n + 2 ( A ) t + → n Pr k t + n\operatorname{Pr}_{A}^{t+}\otimes_{n\operatorname{Pr}_{k}^{t+}}n\operatorname{Pr}_{\mathbb{D}_{n+2}(A)}^{t+}\to n\operatorname{Pr}_{k}^{t+} n Pr A t + ⊗ n Pr k t + n Pr D n + 2 ( A ) t + → n Pr k t + induced by the universal E n + 2 \mathbf{E}_{n+2} E n + 2 -Koszul dual pairing A ⊗ k D n + 2 ( A ) → k A\otimes_k\mathbb{D}_{n+2}(A)\to k A ⊗ k D n + 2 ( A ) → k .
As a byproduct of the proof we reach a categorical understanding of the following well-known fact:
Corollary 4.15. Let A A A be an augmented E n + 2 \mathbf{E}_{n+2} E n + 2 -algebra. Let D n + 2 ( A ) \mathbb{D}_{n+2}(A) D n + 2 ( A ) be an E n + 2 \mathbf{E}_{n+2} E n + 2 -Koszul dual of A A A . Then there exists an equivalence D n + 2 ( A ) ≃ D n + 1 ( Bar ( A ) ) \mathbb{D}_{n+2}(A)\simeq\mathbb{D}_{n+1}(\operatorname{Bar}(A)) D n + 2 ( A ) ≃ D n + 1 ( Bar ( A )) in Alg n + 1 ( Mod k ) \operatorname{Alg}_{n+1}(\operatorname{Mod}_k) Alg n + 1 ( Mod k ) .
Proof. Recall from Proposition 4.5 and its proof that D n + 2 ( A ) \mathbb{D}_{n+2}(A) D n + 2 ( A ) is the image of End n Pr A t + l ( n Pr k t + ) ∈ Alg 1 ( ( n + 1 ) Pr k t + ) \operatorname{End}^{l}_{n\operatorname{Pr}_{A}^{t+}}(n\operatorname{Pr}_{k}^{t+})\in\operatorname{Alg}_1((n+1)\operatorname{Pr}_{k}^{t+}) End n Pr A t + l ( n Pr k t + ) ∈ Alg 1 (( n + 1 ) Pr k t + ) under the right adjoint functors in the sequence of adjoint pairs
Alg n + 2 ( Mod k ) ⇄ Pr k t + θ Pr k t + , 0 Alg n + 1 ( Pr k t + ) ⇄ U Alg 1 ( ( ( n + 1 ) Pr k t ± ) ∗ ∗ ) . \operatorname{Alg}_{n+2}(\operatorname{Mod}_k)\underset{\theta_{\operatorname{Pr}_{k}^{t+},0}}{\stackrel{\operatorname{Pr}_{k}^{t+}}{\rightleftarrows}}\operatorname{Alg}_{n+1}(\operatorname{Pr}_{k}^{t+})\underset{U}{\stackrel{}{ \rightleftarrows}}\operatorname{Alg}_1(((n+1)\operatorname{Pr}_{k}^{t\pm})^{**}). Alg n + 2 ( Mod k ) θ Pr k t + , 0 ⇄ Pr k t + Alg n + 1 ( Pr k t + ) U ⇄ Alg 1 ((( n + 1 ) Pr k t ± ) ∗∗ ) .
See the proof of Lemma 4.8 for this sequence. The equivalence End n Pr A t + l ( n Pr k t + ) ≃ n Pr k t + ⊗ n Pr A t + n Pr k t + ≃ n Pr k ⊗ A k t + \operatorname{End}^{l}_{n\operatorname{Pr}_{A}^{t+}}(n\operatorname{Pr}_{k}^{t+})\simeq n\operatorname{Pr}_{k}^{t+}\otimes_{n\operatorname{Pr}_{A}^{t+}}n\operatorname{Pr}_{k}^{t+}\simeq n\operatorname{Pr}_{k\otimes_A k}^{t+} End n Pr A t + l ( n Pr k t + ) ≃ n Pr k t + ⊗ n Pr A t + n Pr k t + ≃ n Pr k ⊗ A k t + carries the identity functor to ( n − 1 ) Pr k t + ∈ n Pr k ⊗ A k t + (n-1)\operatorname{Pr}_{k}^{t+}\in n\operatorname{Pr}_{k\otimes_A k}^{t+} ( n − 1 ) Pr k t + ∈ n Pr k ⊗ A k t + (see Lemma 4.16 for n Pr k t + ⊗ n Pr A t + n Pr k t + ≃ n Pr k ⊗ A k t + n\operatorname{Pr}_{k}^{t+}\otimes_{n\operatorname{Pr}_{A}^{t+}}n\operatorname{Pr}_{k}^{t+}\simeq n\operatorname{Pr}_{k\otimes_A k}^{t+} n Pr k t + ⊗ n Pr A t + n Pr k t + ≃ n Pr k ⊗ A k t + ). We identify ( n Pr k ⊗ A k t + , ( n − 1 ) Pr k t + ) (n\operatorname{Pr}_{k\otimes_A k}^{t+},(n-1)\operatorname{Pr}_{k}^{t+}) ( n Pr k ⊗ A k t + , ( n − 1 ) Pr k t + ) with the image of End n Pr A t + l ( n Pr k t + ) \operatorname{End}^{l}_{n\operatorname{Pr}_{A}^{t+}}(n\operatorname{Pr}_{k}^{t+}) End n Pr A t + l ( n Pr k t + ) in ( ( n + 1 ) Pr k t ± ) n Pr k t + ∗ ∗ ((n+1)\operatorname{Pr}_{k}^{t\pm})^{**}_{n\operatorname{Pr}_{k}^{t+}} (( n + 1 ) Pr k t ± ) n Pr k t + ∗∗ .
The above sequence of adjoint pairs is obtained from the following sequence of adjoint functors
Alg n + 1 ( Mod k ) ⇄ Alg 1 ( n Pr k t + ) ∗ ∗ ⇄ ( ( n + 1 ) Pr k t ± ) n Pr k t + ∗ ∗ \operatorname{Alg}_{n+1}(\operatorname{Mod}_k)\rightleftarrows\operatorname{Alg}_1(n\operatorname{Pr}_{k}^{t+})^{**}\rightleftarrows((n+1)\operatorname{Pr}_{k}^{t\pm})^{**}_{n\operatorname{Pr}_{k}^{t+}} Alg n + 1 ( Mod k ) ⇄ Alg 1 ( n Pr k t + ) ∗∗ ⇄ (( n + 1 ) Pr k t ± ) n Pr k t + ∗∗
by applying Alg 1 ( − ) \operatorname{Alg}_1(-) Alg 1 ( − ) . In particular, the right adjoint functor carries ( n Pr k ⊗ A k t + , ( n − 1 ) Pr k t + ) (n\operatorname{Pr}_{k\otimes_A k}^{t+},(n-1)\operatorname{Pr}_{k}^{t+}) ( n Pr k ⊗ A k t + , ( n − 1 ) Pr k t + ) to D n + 2 ( A ) \mathbb{D}_{n+2}(A) D n + 2 ( A ) in Alg n + 1 ( Mod k ) \operatorname{Alg}_{n+1}(\operatorname{Mod}_k) Alg n + 1 ( Mod k ) . The image of ( n Pr k ⊗ A k t + , ( n − 1 ) Pr k t + ) (n\operatorname{Pr}_{k\otimes_A k}^{t+},(n-1)\operatorname{Pr}_{k}^{t+}) ( n Pr k ⊗ A k t + , ( n − 1 ) Pr k t + ) in Alg 1 ( n Pr k t + ) \operatorname{Alg}_1(n\operatorname{Pr}_{k}^{t+}) Alg 1 ( n Pr k t + ) under the right adjoint functor is End ( n − 1 ) Pr k ⊗ A k t + l ( ( n − 1 ) Pr k t + ) \operatorname{End}^{l}_{(n-1)\operatorname{Pr}_{k\otimes_A k}^{t+}}((n-1)\operatorname{Pr}_{k}^{t+}) End ( n − 1 ) Pr k ⊗ A k t + l (( n − 1 ) Pr k t + ) . By Proposition 4.5 and its proof again, the image of End ( n − 1 ) Pr k ⊗ A k t + l ( ( n − 1 ) Pr k t + ) \operatorname{End}^{l}_{(n-1)\operatorname{Pr}_{k\otimes_A k}^{t+}}((n-1)\operatorname{Pr}_{k}^{t+}) End ( n − 1 ) Pr k ⊗ A k t + l (( n − 1 ) Pr k t + ) under the right adjoint functor is D n + 1 ( k ⊗ A k ) \mathbb{D}_{n+1}(k\otimes_A k) D n + 1 ( k ⊗ A k ) . Thus, D n + 2 ( A ) ≃ D n + 1 ( k ⊗ A k ) \mathbb{D}_{n+2}(A)\simeq\mathbb{D}_{n+1}(k\otimes_A k) D n + 2 ( A ) ≃ D n + 1 ( k ⊗ A k ) . □
Lemma 4.16. Let A A A be an E n + 2 \mathbf{E}_{n+2} E n + 2 -algebra. There exists a canonical equivalence n Pr k t + ⊗ n Pr A t + n Pr k t + ≃ n Pr k ⊗ A k t + n\operatorname{Pr}_{k}^{t+}\otimes_{n\operatorname{Pr}_{A}^{t+}}n\operatorname{Pr}_{k}^{t+}\simeq n\operatorname{Pr}_{k\otimes_A k}^{t+} n Pr k t + ⊗ n Pr A t + n Pr k t + ≃ n Pr k ⊗ A k t + .
Proof. The proof proceeds by induction on n n n . When n = 0 n=0 n = 0 , the functor c 0 : Mod k ⊗ Mod d Mod k → LMod k ⊗ A k c_0:\operatorname{Mod}_k\otimes_{\operatorname{Mod}_d}\operatorname{Mod}_k\to\operatorname{LMod}_{k\otimes_A k} c 0 : Mod k ⊗ Mod d Mod k → LMod k ⊗ A k induced by the augmented simplicial diagram A ⊗ ∙ → k → k ⊗ A k A^{\otimes\bullet}\to k\to k\otimes_A k A ⊗∙ → k → k ⊗ A k of the bar construction of k ⊗ A k k\otimes_A k k ⊗ A k is an equivalence (see [24 4.8.5.13] ). Next, we assume that for an E n + 1 \mathbf{E}_{n+1} E n + 1 -algebra A A A the canonical map c n − 1 : ( n − 1 ) Pr k t + ⊗ ( n − 1 ) Pr A t + ( n − 1 ) Pr k t + → ( n − 1 ) Pr k ⊗ A k t + c_{n-1}:(n-1)\operatorname{Pr}_{k}^{t+}\otimes_{(n-1)\operatorname{Pr}_{A}^{t+}}(n-1)\operatorname{Pr}_{k}^{t+}\to(n-1)\operatorname{Pr}_{k\otimes_A k}^{t+} c n − 1 : ( n − 1 ) Pr k t + ⊗ ( n − 1 ) Pr A t + ( n − 1 ) Pr k t + → ( n − 1 ) Pr k ⊗ A k t + induced by the bar construction is an equivalence. We will prove that c n : n Pr k t + ⊗ n Pr A t + n Pr k t + → n Pr k ⊗ A k t + c_n:n\operatorname{Pr}_{k}^{t+}\otimes_{n\operatorname{Pr}_{A}^{t+}}n\operatorname{Pr}_{k}^{t+}\to n\operatorname{Pr}_{k\otimes_A k}^{t+} c n : n Pr k t + ⊗ n Pr A t + n Pr k t + → n Pr k ⊗ A k t + is an equivalence. We note that θ n Pr k t + : Alg 1 ( n Pr k t + ) → ( ( n + 1 ) Pr k t + ) n Pr k t + \theta_{n\operatorname{Pr}_{k}^{t+}}:\operatorname{Alg}_1(n\operatorname{Pr}_{k}^{t+})\to((n+1)\operatorname{Pr}_{k}^{t+})_{n\operatorname{Pr}_{k}^{t+}} θ n Pr k t + : Alg 1 ( n Pr k t + ) → (( n + 1 ) Pr k t + ) n Pr k t + is fully faithful (cf. Construction 3.20), and the composition Alg 1 ( n Pr k t + ) → ( ( n + 1 ) Pr k t + ) n Pr k t + → ( ( n + 1 ) Pr k t + ) \operatorname{Alg}_1(n\operatorname{Pr}_{k}^{t+})\to((n+1)\operatorname{Pr}_{k}^{t+})_{n\operatorname{Pr}_{k}^{t+}}\to((n+1)\operatorname{Pr}_{k}^{t+}) Alg 1 ( n Pr k t + ) → (( n + 1 ) Pr k t + ) n Pr k t + → (( n + 1 ) Pr k t + ) with the forgetful functor preserves geometric realizations into the essential images of the composite. Thus, under the assumption, we are reduced to showing that there exists an equivalence n P r k t + ⊗ n P r A t + n P r k t + ≃ LMod D ( n P r k t + ) \mathrm{nPr}_{k}^{t+}\otimes_{\mathrm{nPr}_{A}^{t+}}\mathrm{nPr}_{k}^{t+}\simeq\operatorname{LMod}_{D}(\mathrm{nPr}_{k}^{t+}) nPr k t + ⊗ nPr A t + nPr k t + ≃ LMod D ( nPr k t + ) for some D ∈ Alg 1 ( n P r k t + ) D\in\operatorname{Alg}_{1}(\mathrm{nPr}_{k}^{t+}) D ∈ Alg 1 ( nPr k t + ) . By definition and [24 7.1.3.1] , n P r k t + ⊗ n P r A t + n P r k t + \mathrm{nPr}_{k}^{t+}\otimes_{\mathrm{nPr}_{A}^{t+}}\mathrm{nPr}_{k}^{t+} nPr k t + ⊗ nPr A t + nPr k t + is equivalent to
LMod ( n − 1 ) P r A t + ( LMod ( n − 1 ) P r A t + ( n P r k t + ) ) ⊗ LMod ( n − 1 ) P r A t + ( n P r k t + ) RMod ( n − 1 ) P r k t + ( LMod ( n − 1 ) P r A t + ( n P r k t + ) ) . \operatorname{LMod}_{(n-1)\mathrm{Pr}_{A}^{t+}}\left(\operatorname{LMod}_{(n-1)\mathrm{Pr}_{A}^{t+}}(\mathrm{nPr}_{k}^{t+})\right)\otimes_{\operatorname{LMod}_{(n-1)\mathrm{Pr}_{A}^{t+}}(\mathrm{nPr}_{k}^{t+})}\operatorname{RMod}_{(n-1)\mathrm{Pr}_{k}^{t+}}\left(\operatorname{LMod}_{(n-1)\mathrm{Pr}_{A}^{t+}}(\mathrm{nPr}_{k}^{t+})\right). LMod ( n − 1 ) Pr A t + ( LMod ( n − 1 ) Pr A t + ( nPr k t + ) ) ⊗ LMod ( n − 1 ) Pr A t + ( nPr k t + ) RMod ( n − 1 ) Pr k t + ( LMod ( n − 1 ) Pr A t + ( nPr k t + ) ) .
where we regard ( n − 1 ) P r k t + (n-1)\mathrm{Pr}_{k}^{t+} ( n − 1 ) Pr k t + as an E 1 \mathrm{E}_{1} E 1 -algebra in LMod ( n − 1 ) P r A t + ( n P r k t + ) \operatorname{LMod}_{(n-1)\mathrm{Pr}_{A}^{t+}(\mathrm{nPr}_{k}^{t+})} LMod ( n − 1 ) Pr A t + ( nPr k t + ) . This is equivalent to
B n P r k t + = RMod ( n − 1 ) P r k t + ( LMod ( n − 1 ) P r k t + ( LMod ( n − 1 ) P r A t + ( n P r k t + ) ) ) . \mathcal{B}_{\mathrm{nPr}_{k}^{t+}}=\operatorname{RMod}_{(n-1)\mathrm{Pr}_{k}^{t+}}\left(\operatorname{LMod}_{(n-1)\mathrm{Pr}_{k}^{t+}}\left(\operatorname{LMod}_{(n-1)\mathrm{Pr}_{A}^{t+}}(\mathrm{nPr}_{k}^{t+})\right)\right). B nPr k t + = RMod ( n − 1 ) Pr k t + ( LMod ( n − 1 ) Pr k t + ( LMod ( n − 1 ) Pr A t + ( nPr k t + ) ) ) .
(cf. [24 4.8.4.6] ). For ease of notation, we write RMod B ( LMod C ( LMod A ′ ( P ) ) ) \operatorname{RMod}_{B}(\operatorname{LMod}_{C}(\operatorname{LMod}_{A'}(P))) RMod B ( LMod C ( LMod A ′ ( P ))) for B n P r k t + \mathcal{B}_{\mathrm{nPr}_{k}^{t+}} B nPr k t + ; P = n P r k t + P=\mathrm{nPr}_{k}^{t+} P = nPr k t + , A ′ = ( n − 1 ) P r A t + A'=(n-1)\mathrm{Pr}_{A}^{t+} A ′ = ( n − 1 ) Pr A t + , etc. It suffices to prove that RMod B ( LMod C ( LMod A ′ ( P ) ) ) \operatorname{RMod}_{B}(\operatorname{LMod}_{C}(\operatorname{LMod}_{A'}(P))) RMod B ( LMod C ( LMod A ′ ( P ))) is equivalent to LMod D ( P ) ( = RMod D o p ( P ) ) \operatorname{LMod}_{D}(P)(=\operatorname{RMod}_{D^{\mathrm{op}}}(P)) LMod D ( P ) ( = RMod D op ( P )) for some D ∈ Alg 1 ( P ) D\in\operatorname{Alg}_{1}(P) D ∈ Alg 1 ( P ) . Consider the sequence of adjunctions
P ⇄ f LMod A ′ ( P ) ⇄ g RMod B ( LMod A ′ ( P ) ) ⇄ h RMod B ( LMod C ( LMod A ′ ( P ) ) ) P \underset{\ }{\overset{f}{\rightleftarrows}} \operatorname{LMod}_{A'}(P) \underset{\ }{\overset{g}{\rightleftarrows}} \operatorname{RMod}_{B}(\operatorname{LMod}_{A'}(P)) \underset{\ }{\overset{h}{\rightleftarrows}} \operatorname{RMod}_{B}(\operatorname{LMod}_{C}(\operatorname{LMod}_{A'}(P))) P ⇄ f LMod A ′ ( P ) ⇄ g RMod B ( LMod A ′ ( P )) ⇄ h RMod B ( LMod C ( LMod A ′ ( P )))
such that f f f , g g g , and h h h are free functors, and left arrows are forgetful functors. The functor f f f is given informally by A ′ ⊗ ( − ) A'\otimes(-) A ′ ⊗ ( − ) . The functor g g g is given by B ⊗ A ′ ( − ) B\otimes_{A'}(-) B ⊗ A ′ ( − ) . The functor h h h is given by C ⊗ A ′ ( − ) C\otimes_{A'}(-) C ⊗ A ′ ( − ) . More precisely, we let α : LMod A ′ ( P ) → LMod C ( LMod A ′ ( P ) ) \alpha:\operatorname{LMod}_{A'}(P)\to\operatorname{LMod}_{C}(\operatorname{LMod}_{A'}(P)) α : LMod A ′ ( P ) → LMod C ( LMod A ′ ( P )) be the (unique) right LMod A ′ ( P ) \operatorname{LMod}_{A'}(P) LMod A ′ ( P ) -module map which sends A ′ A' A ′ to C C C . Applying RMod B \operatorname{RMod}_{B} RMod B to α \alpha α we get a P P P -module map h h h . We consider the composite of adjunctions
F P : P ⇄ B P = RMod B ( LMod C ( LMod A ′ ( P ) ) ) : G P . F_{P}:P\rightleftarrows B_{P}=\operatorname{RMod}_{B}(\operatorname{LMod}_{C}(\operatorname{LMod}_{A'}(P))):G_{P}. F P : P ⇄ B P = RMod B ( LMod C ( LMod A ′ ( P ))) : G P .
We let C ⊗ A ′ B C\otimes_{A'}B C ⊗ A ′ B be the image of the unit object of P P P under F P F_{P} F P , so that F P F_{P} F P is given by C ⊗ A ′ B ⊗ ( − ) C\otimes_{A'}B\otimes(-) C ⊗ A ′ B ⊗ ( − ) . The right adjoint G P G_{P} G P is described by using Hom objects in P P P ; G P ( M ) ≃ Hom B ( C ⊗ A ′ B , M ) G_{P}(M)\simeq\operatorname{Hom}_{B}(C\otimes_{A'}B,M) G P ( M ) ≃ Hom B ( C ⊗ A ′ B , M ) . We write X : = C ⊗ A ′ B X:=C\otimes_{A'}B X := C ⊗ A ′ B . Let E E E be the endomorphism algebra of X X X defined as an object of Alg 1 ( P ) \operatorname{Alg}_{1}(P) Alg 1 ( P ) . We write Q = Mod ( n − 1 ) P r k t + ( Cat ^ † ) Q=\operatorname{Mod}_{(n-1)\mathrm{Pr}_{k}^{t+}}(\widehat{\operatorname{Cat}}_{\dagger}) Q = Mod ( n − 1 ) Pr k t + ( Cat † ) . This symmetric monoidal ∞ \infty ∞ -category is a V V V -presentably symmetric monoidal ∞ \infty ∞ -category, and P ⊂ Q P\subset Q P ⊂ Q is closed under tensor products and small colimits. Then ( F , G ) (F,G) ( F , G ) is extended to
F : Q ⇄ B = RMod B ( LMod C ( LMod A ′ ( Q ) ) ) : G . F:Q\rightleftarrows B=\operatorname{RMod}_{B}(\operatorname{LMod}_{C}(\operatorname{LMod}_{A'}(Q))):G. F : Q ⇄ B = RMod B ( LMod C ( LMod A ′ ( Q ))) : G .
Since Q Q Q is a V V V -presentably symmetric monoidal ∞ \infty ∞ -category, and B \mathcal{B} B is a Q Q Q -module (Q Q Q -tensored), B \mathcal{B} B and Q Q Q are Q Q Q -enriched (cf. [24 4.2.1] , [14 7.4.13] ). We define F ′ : RMod E ( Q ) → B F':\operatorname{RMod}_{E}(Q)\to\mathcal{B} F ′ : RMod E ( Q ) → B by L ↦ L ⊗ E X L\mapsto L\otimes_{E}X L ↦ L ⊗ E X , which is determined by the universal left action on E E E on X X X . We define G ′ : B → RMod E ( Q ) G':\mathcal{B}\to\operatorname{RMod}_{E}(Q) G ′ : B → RMod E ( Q ) as a right adjoint of F ′ F' F ′ , which is given by M ↦ Hom B ( X , M ) M\mapsto\operatorname{Hom}_{\mathcal{B}}(X,M) M ↦ Hom B ( X , M ) . The canonical evaluation map Hom B ( X , M ) ⊗ X → M \operatorname{Hom}_{\mathcal{B}}(X,M)\otimes X\to M Hom B ( X , M ) ⊗ X → M induces F ′ ( G ′ ( M ) ) = Hom B ( X , M ) ⊗ E X → M F'(G'(M))=\operatorname{Hom}_{\mathcal{B}}(X,M)\otimes_{E}X\to M F ′ ( G ′ ( M )) = Hom B ( X , M ) ⊗ E X → M . Thus, we have ϵ : F ′ ∘ G ′ → id \epsilon:F'\circ G'\to\operatorname{id} ϵ : F ′ ∘ G ′ → id . The canonical map L ⊗ X → L ⊗ E X L\otimes X\to L\otimes_{E}X L ⊗ X → L ⊗ E X induces L → Hom B ( X , L ⊗ E X ) = G ′ ( F ′ ( L ) ) L\to\operatorname{Hom}_{\mathcal{B}}(X,L\otimes_{E}X)=G'(F'(L)) L → Hom B ( X , L ⊗ E X ) = G ′ ( F ′ ( L )) . Thus, we have η : id → G ′ ∘ F ′ \eta:\operatorname{id}\to G'\circ F' η : id → G ′ ∘ F ′ .
We show that η \eta η and ϵ \epsilon ϵ are equivalences. Note that G ′ G' G ′ preserves V V V -small colimits since the composite
B → G ′ RMod E ( Q ) → forget Q \mathcal{B}\xrightarrow{G'}\operatorname{RMod}_{E}(Q)\xrightarrow{\text{forget}}Q B G ′ RMod E ( Q ) forget Q
preserves V V V -small colimits, and the forgetful functor is conservative and preserves V V V -small colimits. In particular, G ′ ∘ F ′ G'\circ F' G ′ ∘ F ′ preserves geometric realizations. We may assume that L L L is a free right E E E -module T ⊗ E T\otimes E T ⊗ E for some T ∈ Q T\in Q T ∈ Q . Then T ⊗ Hom B ( X , X ) → Hom B ( X , T ⊗ E ⊗ E X ) T\otimes\operatorname{Hom}_{\mathcal{B}}(X,X)\to\operatorname{Hom}_{\mathcal{B}}(X,T\otimes E\otimes_{E}X) T ⊗ Hom B ( X , X ) → Hom B ( X , T ⊗ E ⊗ E X ) is equivalent to an identity T ⊗ X → T ⊗ X T\otimes X\to T\otimes X T ⊗ X → T ⊗ X . This shows that η \eta η is an equivalence. Next, we show that ϵ M : Hom B ( X , M ) ⊗ E X → M \epsilon_{M}:\operatorname{Hom}_{\mathcal{B}}(X,M)\otimes_{E}X\to M ϵ M : Hom B ( X , M ) ⊗ E X → M is an equivalence. As above, we may assume that M = T ⊗ X M=T\otimes X M = T ⊗ X . Then we have canonical equivalences Hom B ( X , T ⊗ X ) ⊗ E X ≃ T ⊗ Hom B ( X , X ) ⊗ E X ≃ T ⊗ X \operatorname{Hom}_{\mathcal{B}}(X,T\otimes X)\otimes_{E}X\simeq T\otimes\operatorname{Hom}_{\mathcal{B}}(X,X)\otimes_{E}X\simeq T\otimes X Hom B ( X , T ⊗ X ) ⊗ E X ≃ T ⊗ Hom B ( X , X ) ⊗ E X ≃ T ⊗ X , and ϵ M \epsilon_{M} ϵ M can be identified with the identity T ⊗ X → T ⊗ X T\otimes X\to T\otimes X T ⊗ X → T ⊗ X . Thus, we see that ϵ \epsilon ϵ is an equivalence. Thus, F ′ F' F ′ and G ′ G' G ′ induce mutually inverse equivalences RMod E ( Q ) ≃ RMod B ( LMod C ( LMod A ′ ( Q ) ) ) \operatorname{RMod}_{E}(Q)\simeq\operatorname{RMod}_{B}(\operatorname{LMod}_{C}(\operatorname{LMod}_{A'}(Q))) RMod E ( Q ) ≃ RMod B ( LMod C ( LMod A ′ ( Q ))) . (For the proof of equivalences of F ′ F' F ′ and G ′ G' G ′ , one may also use Barr–Beck–Lurie theorem by combining the adjunction ( F , G ) (F,G) ( F , G ) and the free-forget adjunction Q ⇄ RMod E ( Q ) Q\rightleftarrows\operatorname{RMod}_{E}(Q) Q ⇄ RMod E ( Q ) , cf. [24 4.7.3.16] ). By the definition of G ′ G' G ′ (or F ′ F' F ′ ), this induces an equivalence RMod E ( P ) ≃ RMod B ( LMod C ( LMod A ′ ( P ) ) ) \operatorname{RMod}_{E}(P)\simeq\operatorname{RMod}_{B}(\operatorname{LMod}_{C}(\operatorname{LMod}_{A'}(P))) RMod E ( P ) ≃ RMod B ( LMod C ( LMod A ′ ( P ))) . □
Categorified Koszul duality of t t t -structured ( ∞ , n ) (\infty,n) ( ∞ , n ) -categories
Definition 4.17. Let R R R be an augmented E n \mathrm{E}_{n} E n -algebra in Mod k \operatorname{Mod}_{k} Mod k . Let n n n be a natural number or ∞ \infty ∞ . The augmented E n \mathrm{E}_{n} E n -algebra R R R is Artin if the following properties hold:
R R R is connective,
⨁ i ∈ Z H i ( R ) \bigoplus_{i\in\mathbb{Z}}H^{i}(R) ⨁ i ∈ Z H i ( R ) is finite dimensional over k k k ,
if I I I is the radical of H 0 ( R ) H^{0}(R) H 0 ( R ) , the canonical map k → H 0 ( R ) / I k\to H^{0}(R)/I k → H 0 ( R ) / I is an isomorphism.
If the truncation τ ≥ − m R \tau_{\geq-m}R τ ≥ − m R is Artin for any nonnegative integer m ≥ 0 m \geq0 m ≥ 0 , we say that R R R is almost Artin. By convention, an E ∞ E_{\infty} E ∞ -algebra is a commutative algebra object in Mod k \operatorname{Mod}_k Mod k , and Alg ∞ ( Mod k ) = CAlg ( Mod k ) \operatorname{Alg}_{\infty}(\operatorname{Mod}_k)=\operatorname{CAlg}(\operatorname{Mod}_k) Alg ∞ ( Mod k ) = CAlg ( Mod k ) . We let Art n \operatorname{Art}_n Art n denote the full subcategory on Alg n + ( Mod k ) \operatorname{Alg}^{+}_n(\operatorname{Mod}_k) Alg n + ( Mod k ) spanned by Artin E n E_n E n -algebras. We let AArt n \operatorname{AArt}_n AArt n denote the full subcategory on Alg n + ( Mod k ) \operatorname{Alg}^{+}_n(\operatorname{Mod}_k) Alg n + ( Mod k ) spanned by almost Artin E n E_n E n -algebras.
Remark 4.18. In [25 ] , the terminology small is adopted instead of Artin.
Let R R R be an augmented E n + 1 E_{n+1} E n + 1 -algebra. Note that ( n − 1 ) Prt R + (n-1)\operatorname{Prt}^{+}_{R} ( n − 1 ) Prt R + and ( n − 1 ) Prt D n + 1 ( R ) + (n-1)\operatorname{Prt}^{+}_{\mathbb{D}_{n+1}(R)} ( n − 1 ) Prt D n + 1 ( R ) + are E 1 E_1 E 1 -algebra objects in n Prt k + n\operatorname{Prt}^{+}_k n Prt k + . We regard ( n − 1 ) Prt k + (n-1)\operatorname{Prt}^{+}_k ( n − 1 ) Prt k + as a left ( n − 1 ) Prt R + ⊗ ( n − 1 ) Prt k + ( n − 1 ) Prt D n + 1 ( R ) + (n-1)\operatorname{Prt}^{+}_{R}\otimes_{(n-1)\operatorname{Prt}^{+}_k}(n-1)\operatorname{Prt}^{+}_{\mathbb{D}_{n+1}(R)} ( n − 1 ) Prt R + ⊗ ( n − 1 ) Prt k + ( n − 1 ) Prt D n + 1 ( R ) + -module determined by ( n − 1 ) Prt R + ⊗ ( n − 1 ) Prt k + ( n − 1 ) Prt D n + 1 ( R ) + → ( n − 1 ) Prt k + (n-1)\operatorname{Prt}^{+}_{R}\otimes_{(n-1)\operatorname{Prt}^{+}_k}(n-1)\operatorname{Prt}^{+}_{\mathbb{D}_{n+1}(R)}\to(n-1)\operatorname{Prt}^{+}_k ( n − 1 ) Prt R + ⊗ ( n − 1 ) Prt k + ( n − 1 ) Prt D n + 1 ( R ) + → ( n − 1 ) Prt k + induced by the universal pairing R ⊗ k D n + 1 ( R ) → k R\otimes_k\mathbb{D}_{n+1}(R)\to k R ⊗ k D n + 1 ( R ) → k . We have
( − ) ⊗ ( n − 1 ) Prt R + ( n − 1 ) Prt k + : RMod ( n − 1 ) Prt R + ( n Prt k + ) ⟶ LMod Prt D n + 1 ( R ) + ( n Prt k + ) . (-)\otimes_{(n-1)\operatorname{Prt}^{+}_{R}}(n-1)\operatorname{Prt}^{+}_k:\operatorname{RMod}_{(n-1)\operatorname{Prt}^{+}_{R}}(n\operatorname{Prt}^{+}_k)\longrightarrow\operatorname{LMod}_{\operatorname{Prt}^{+}_{\mathbb{D}_{n+1}(R)}}(n\operatorname{Prt}^{+}_k). ( − ) ⊗ ( n − 1 ) Prt R + ( n − 1 ) Prt k + : RMod ( n − 1 ) Prt R + ( n Prt k + ) ⟶ LMod Prt D n + 1 ( R ) + ( n Prt k + ) .
We write Φ R , n \Phi_{R,n} Φ R , n for ( − ) ⊗ ( n − 1 ) Prt R + ( n − 1 ) Prt k + (-)\otimes_{(n-1)\operatorname{Prt}^{+}_{R}}(n-1)\operatorname{Prt}^{+}_k ( − ) ⊗ ( n − 1 ) Prt R + ( n − 1 ) Prt k + . Similarly, we set
Ψ R , n = ( n − 1 ) Prt k + ⊗ ( n − 1 ) Prt D n + 1 ( R ) + ( − ) : LMod ( n − 1 ) Prt D n + 1 ( R ) + ( n Prt k + ) ⟶ RMod ( n − 1 ) Prt R + ( n Prt k + ) . \Psi_{R,n}=(n-1)\operatorname{Prt}^{+}_k\otimes_{(n-1)\operatorname{Prt}^{+}_{\mathbb{D}_{n+1}(R)}}(-):\operatorname{LMod}_{(n-1)\operatorname{Prt}^{+}_{\mathbb{D}_{n+1}(R)}}(n\operatorname{Prt}^{+}_k)\longrightarrow\operatorname{RMod}_{(n-1)\operatorname{Prt}^{+}_{R}}(n\operatorname{Prt}^{+}_k). Ψ R , n = ( n − 1 ) Prt k + ⊗ ( n − 1 ) Prt D n + 1 ( R ) + ( − ) : LMod ( n − 1 ) Prt D n + 1 ( R ) + ( n Prt k + ) ⟶ RMod ( n − 1 ) Prt R + ( n Prt k + ) .
Theorem 4.19. Let R R R be an almost Artin E n + 1 E_{n+1} E n + 1 -algebra. Suppose that n ≥ 1 n\geq1 n ≥ 1 . Then Ψ R , n \Psi_{R,n} Ψ R , n and Φ R , n \Phi_{R,n} Φ R , n induce mutually inverse equivalences
Ψ ^ R , n : LMod ( n − 1 ) Prt D n + 1 ( R ) ± ( n Prt k ± ) → ≃ RMod ( n − 1 ) Prt R ± ( n Prt k ± ) : Φ ^ R , n \widehat{\Psi}_{R,n}:\operatorname{LMod}_{(n-1)\operatorname{Prt}^{\pm}_{\mathbb{D}_{n+1}(R)}}(n\operatorname{Prt}^{\pm}_k)\xrightarrow{\simeq}\operatorname{RMod}_{(n-1)\operatorname{Prt}^{\pm}_{R}}(n\operatorname{Prt}^{\pm}_k):\widehat{\Phi}_{R,n} Ψ R , n : LMod ( n − 1 ) Prt D n + 1 ( R ) ± ( n Prt k ± ) ≃ RMod ( n − 1 ) Prt R ± ( n Prt k ± ) : Φ R , n in ( n + 1 ) Prt k ± (n+1)\operatorname{Prt}^{\pm}_k ( n + 1 ) Prt k ± . The functor Ψ ^ R , n \widehat{\Psi}_{R,n} Ψ R , n sends E E E to ( n − 1 ) Prt k ± ⊗ ^ ( n − 1 ) Prt D n + 1 ( R ) ± E (n-1)\operatorname{Prt}^{\pm}_k\widehat{\otimes}_{(n-1)\operatorname{Prt}^{\pm}_{\mathbb{D}_{n+1}(R)}}E ( n − 1 ) Prt k ± ⊗ ( n − 1 ) Prt D n + 1 ( R ) ± E where ⊗ ^ ( n − 1 ) Prt D n + 1 ( R ) ± \widehat{\otimes}_{(n-1)\operatorname{Prt}^{\pm}_{\mathbb{D}_{n+1}(R)}} ⊗ ( n − 1 ) Prt D n + 1 ( R ) ± means the relative tensor product in n Prt k ± n\operatorname{Prt}^{\pm}_k n Prt k ± . The functor Φ ^ R , n \widehat{\Phi}_{R,n} Φ R , n sends D D D to D ⊗ ^ ( n − 1 ) Prt R ± ( n − 1 ) Prt k ± D\widehat{\otimes}_{(n-1)\operatorname{Prt}^{\pm}_{R}}(n-1)\operatorname{Prt}^{\pm}_k D ⊗ ( n − 1 ) Prt R ± ( n − 1 ) Prt k ± where ⊗ ^ ( n − 1 ) Prt R ± \widehat{\otimes}_{(n-1)\operatorname{Prt}^{\pm}_{R}} ⊗ ( n − 1 ) Prt R ± means the relative tensor product in n Prt k ± n\operatorname{Prt}^{\pm}_k n Prt k ± .
We deduce this theorem from Proposition 4.20 and Proposition 4.21.
Proposition 4.20. When n = 1 n=1 n = 1 , the completions of Φ R , 1 \Phi_{R,1} Φ R , 1 and Ψ R , 1 \Psi_{R,1} Ψ R , 1 induce mutually inverse equivalences
Φ ^ R , 1 : RMod LMod R ( Prt k ± ) → ≃ LMod LMod D 2 ( R ) ( Prt k ± ) , Ψ ^ R , 1 : LMod LMod D 2 ( R ) ( Prt k ± ) → ≃ RMod LMod R ( Prt k ± ) . \widehat{\Phi}_{R,1}:\operatorname{RMod}_{\operatorname{LMod}_{R}}(\operatorname{Prt}^{\pm}_k)\xrightarrow{\simeq}\operatorname{LMod}_{\operatorname{LMod}_{\mathbb{D}_2(R)}}(\operatorname{Prt}^{\pm}_k),\qquad\widehat{\Psi}_{R,1}:\operatorname{LMod}_{\operatorname{LMod}_{\mathbb{D}_2(R)}}(\operatorname{Prt}^{\pm}_k)\xrightarrow{\simeq}\operatorname{RMod}_{\operatorname{LMod}_{R}}(\operatorname{Prt}^{\pm}_k). Φ R , 1 : RMod LMod R ( Prt k ± ) ≃ LMod LMod D 2 ( R ) ( Prt k ± ) , Ψ R , 1 : LMod LMod D 2 ( R ) ( Prt k ± ) ≃ RMod LMod R ( Prt k ± ) . Here LMod R \operatorname{LMod}_{R} LMod R and LMod D 2 ( R ) \operatorname{LMod}_{\mathbb{D}_2(R)} LMod D 2 ( R ) are equipped with the standard t t t -structures (see Example 3.2 and Remark 3.3).
When R R R is Artin, this is one of the main results of [20 ] . Proposition 4.20 generalizes the result of [20 ] to the case of almost Artin algebras. We will prove Proposition 4.20 in the next section.
Proposition 4.21. We assume that for an almost Artin E n + 1 E_{n+1} E n + 1 -algebra R R R ,
Φ R , n : RMod ( n − 1 ) Prt R + ( n Prt k + ) ⟶ LMod ( n − 1 ) Prt D n + 1 ( R ) + ( n Prt k + ) \Phi_{R,n}:\operatorname{RMod}_{(n-1)\operatorname{Prt}^{+}_{R}}(n\operatorname{Prt}^{+}_k)\longrightarrow\operatorname{LMod}_{(n-1)\operatorname{Prt}^{+}_{\mathbb{D}_{n+1}(R)}}(n\operatorname{Prt}^{+}_k) Φ R , n : RMod ( n − 1 ) Prt R + ( n Prt k + ) ⟶ LMod ( n − 1 ) Prt D n + 1 ( R ) + ( n Prt k + ) and
Ψ R , n : LMod ( n − 1 ) Prt D n + 1 ( R ) + ( n Prt k + ) ⟶ RMod ( n − 1 ) Prt R + ( n Prt k + ) \Psi_{R,n}:\operatorname{LMod}_{(n-1)\operatorname{Prt}^{+}_{\mathbb{D}_{n+1}(R)}}(n\operatorname{Prt}^{+}_k)\longrightarrow\operatorname{RMod}_{(n-1)\operatorname{Prt}^{+}_{R}}(n\operatorname{Prt}^{+}_k) Ψ R , n : LMod ( n − 1 ) Prt D n + 1 ( R ) + ( n Prt k + ) ⟶ RMod ( n − 1 ) Prt R + ( n Prt k + ) are mutually inverse equivalences after the completion. Then for an almost Artin E n + 2 E_{n+2} E n + 2 -algebra R R R , the functors Ψ R , n + 1 : LMod n Prt D n + 2 ( R ) + ( ( n + 1 ) Prt k + ) ⟶ RMod n Prt R + ( ( n + 1 ) Prt k + ) \Psi_{R,n+1}:\operatorname{LMod}_{n\operatorname{Prt}^{+}_{\mathbb{D}_{n+2}(R)}}((n+1)\operatorname{Prt}^{+}_k)\longrightarrow\operatorname{RMod}_{n\operatorname{Prt}^{+}_{R}}((n+1)\operatorname{Prt}^{+}_k) Ψ R , n + 1 : LMod n Prt D n + 2 ( R ) + (( n + 1 ) Prt k + ) ⟶ RMod n Prt R + (( n + 1 ) Prt k + ) and Φ R , n + 1 : RMod n Prt R + ( ( n + 1 ) Prt k + ) ⟶ LMod n Prt D n + 2 ( R ) + ( ( n + 1 ) Prt k + ) \Phi_{R,n+1}:\operatorname{RMod}_{n\operatorname{Prt}^{+}_{R}}((n+1)\operatorname{Prt}^{+}_k)\longrightarrow\operatorname{LMod}_{n\operatorname{Prt}^{+}_{\mathbb{D}_{n+2}(R)}}((n+1)\operatorname{Prt}^{+}_k) Φ R , n + 1 : RMod n Prt R + (( n + 1 ) Prt k + ) ⟶ LMod n Prt D n + 2 ( R ) + (( n + 1 ) Prt k + ) induce mutually inverse equivalences after the completion.
The rest of this section is devoted to the proof of Proposition 4.21.
Construction 4.22. Let R R R be an E n + 2 E_{n+2} E n + 2 -algebra. We write D = D n + 2 ( R ) D=\mathbb{D}_{n+2}(R) D = D n + 2 ( R ) . We define a right n Prt D + n\operatorname{Prt}^{+}_{D} n Prt D + -module map
ξ : n Prt D + ⟶ n Prt k + ⊗ ( n Prt R + ) opp n Prt k + . \xi:n\operatorname{Prt}^{+}_{D}\longrightarrow n\operatorname{Prt}^{+}_{k}\otimes_{(n\operatorname{Prt}^{+}_{R})^{\operatorname{opp}}}n\operatorname{Prt}^{+}_{k}. ξ : n Prt D + ⟶ n Prt k + ⊗ ( n Prt R + ) opp n Prt k + . The superscript opp \operatorname{opp} opp in ( n Prt R + ) opp (n\operatorname{Prt}^{+}_{R})^{\operatorname{opp}} ( n Prt R + ) opp indicates the opposite algebra of n Prt R + n\operatorname{Prt}^{+}_{R} n Prt R + in Alg 1 ( ( n + 1 ) Prt k + ) \operatorname{Alg}_1((n+1)\operatorname{Prt}^{+}_k) Alg 1 (( n + 1 ) Prt k + ) . This ξ \xi ξ is induced by the algebraization functor n Prt D + → End n Prt R + l ( n Prt k + ) n\operatorname{Prt}^{+}_{D}\to\operatorname{End}^{l}_{n\operatorname{Prt}^{+}_{R}}(n\operatorname{Prt}^{+}_k) n Prt D + → End n Prt R + l ( n Prt k + ) (see Definition 4.3 and Proposition 4.5) and
m : End n Prt R + l ( n Prt k + ) opp ≃ End n Prt R + r ( n Prt k + ) ⟶ n Prt k + ⊗ ( n Prt R + ) opp n Prt k + m:\operatorname{End}^{l}_{n\operatorname{Prt}^{+}_{R}}(n\operatorname{Prt}^{+}_k)^{\operatorname{opp}}\simeq\operatorname{End}^{r}_{n\operatorname{Prt}^{+}_{R}}(n\operatorname{Prt}^{+}_k)\longrightarrow n\operatorname{Prt}^{+}_k\otimes_{(n\operatorname{Prt}^{+}_{R})^{\operatorname{opp}}}n\operatorname{Prt}^{+}_k m : End n Prt R + l ( n Prt k + ) opp ≃ End n Prt R + r ( n Prt k + ) ⟶ n Prt k + ⊗ ( n Prt R + ) opp n Prt k +
defined as follows: The tautological left module action of End n Pr R t + r ( n Pr k t + ) \operatorname{End}_{n\operatorname{Pr}_{R}^{t+}}^{\mathrm{r}}(n\operatorname{Pr}_{k}^{t+}) End n Pr R t + r ( n Pr k t + ) on n Pr k t + n\operatorname{Pr}_{k}^{t+} n Pr k t + on the right-hand side in n Pr k t + ⊗ ( n Pr R t + ) o p p n Pr k t + n\operatorname{Pr}_{k}^{t+}\otimes_{(n\operatorname{Pr}_{R}^{t+})^{\mathrm{opp}}}n\operatorname{Pr}_{k}^{t+} n Pr k t + ⊗ ( n Pr R t + ) opp n Pr k t + induces
End n Pr R t + r ( n Pr k t + ) × n Pr k t + ⊗ ( n Pr R t + ) o p p n Pr k t + → n Pr k t + ⊗ ( n Pr R t + ) o p p n Pr k t + \operatorname{End}_{n\operatorname{Pr}_{R}^{t+}}^{\mathrm{r}}(n\operatorname{Pr}_{k}^{t+})\times n\operatorname{Pr}_{k}^{t+}\otimes_{(n\operatorname{Pr}_{R}^{t+})^{\mathrm{opp}}}n\operatorname{Pr}_{k}^{t+}\to n\operatorname{Pr}_{k}^{t+}\otimes_{(n\operatorname{Pr}_{R}^{t+})^{\mathrm{opp}}}n\operatorname{Pr}_{k}^{t+} End n Pr R t + r ( n Pr k t + ) × n Pr k t + ⊗ ( n Pr R t + ) opp n Pr k t + → n Pr k t + ⊗ ( n Pr R t + ) opp n Pr k t +
whose evaluation at ( n − 1 ) Pr k t + ⊗ ( n − 1 ) Pr k t + ∈ n Pr k t + ⊗ ( n Pr R t + ) o p p n Pr k t + (n-1)\operatorname{Pr}_{k}^{t+}\otimes(n-1)\operatorname{Pr}_{k}^{t+}\in n\operatorname{Pr}_{k}^{t+}\otimes_{(n\operatorname{Pr}_{R}^{t+})^{\mathrm{opp}}}n\operatorname{Pr}_{k}^{t+} ( n − 1 ) Pr k t + ⊗ ( n − 1 ) Pr k t + ∈ n Pr k t + ⊗ ( n Pr R t + ) opp n Pr k t + determines
m : End n Pr R t + r ( n Pr k t + ) = End n Pr R t + r ( n Pr k t + ) × { ( n − 1 ) Pr k t + ⊗ ( n − 1 ) Pr k t + } → n Pr k t + ⊗ ( n Pr R t + ) o p p n Pr k t + , m:\operatorname{End}_{n\operatorname{Pr}_{R}^{t+}}^{\mathrm{r}}(n\operatorname{Pr}_{k}^{t+})=\operatorname{End}_{n\operatorname{Pr}_{R}^{t+}}^{\mathrm{r}}(n\operatorname{Pr}_{k}^{t+})\times\{(n-1)\operatorname{Pr}_{k}^{t+}\otimes(n-1)\operatorname{Pr}_{k}^{t+}\}\to n\operatorname{Pr}_{k}^{t+}\otimes_{(n\operatorname{Pr}_{R}^{t+})^{\mathrm{opp}}}n\operatorname{Pr}_{k}^{t+}, m : End n Pr R t + r ( n Pr k t + ) = End n Pr R t + r ( n Pr k t + ) × {( n − 1 ) Pr k t + ⊗ ( n − 1 ) Pr k t + } → n Pr k t + ⊗ ( n Pr R t + ) opp n Pr k t + ,
where ( n − 1 ) Pr k t + ⊗ ( n − 1 ) Pr k t + (n-1)\operatorname{Pr}_{k}^{t+}\otimes(n-1)\operatorname{Pr}_{k}^{t+} ( n − 1 ) Pr k t + ⊗ ( n − 1 ) Pr k t + denotes the image of ( n − 1 ) Pr k t + ⊗ ( n − 1 ) Pr k t + ∈ n Pr k t + ⊗ n Pr k t + n Pr k t + (n-1)\operatorname{Pr}_{k}^{t+}\otimes(n-1)\operatorname{Pr}_{k}^{t+}\in n\operatorname{Pr}_{k}^{t+}\otimes_{n\operatorname{Pr}_{k}^{t+}}n\operatorname{Pr}_{k}^{t+} ( n − 1 ) Pr k t + ⊗ ( n − 1 ) Pr k t + ∈ n Pr k t + ⊗ n Pr k t + n Pr k t + under the functor n Pr k t + ⊗ n Pr k t + n Pr k t + → n Pr k t + ⊗ ( n Pr R t + ) o p p n Pr k t + n\operatorname{Pr}_{k}^{t+}\otimes_{n\operatorname{Pr}_{k}^{t+}}n\operatorname{Pr}_{k}^{t+}\to n\operatorname{Pr}_{k}^{t+}\otimes_{(n\operatorname{Pr}_{R}^{t+})^{\mathrm{opp}}}n\operatorname{Pr}_{k}^{t+} n Pr k t + ⊗ n Pr k t + n Pr k t + → n Pr k t + ⊗ ( n Pr R t + ) opp n Pr k t + induced by n Pr k t + → n Pr R t + n\operatorname{Pr}_{k}^{t+}\to n\operatorname{Pr}_{R}^{t+} n Pr k t + → n Pr R t + . The functor sends the unit object of End n Pr R t + r ( n Pr k t + ) \operatorname{End}_{n\operatorname{Pr}_{R}^{t+}}^{\mathrm{r}}(n\operatorname{Pr}_{k}^{t+}) End n Pr R t + r ( n Pr k t + ) (the identity functor) to ( n − 1 ) Pr k t + ⊗ ( n − 1 ) Pr k t + ∈ n Pr k t + ⊗ ( n Pr R t + ) o p p n Pr k t + (n-1)\operatorname{Pr}_{k}^{t+}\otimes(n-1)\operatorname{Pr}_{k}^{t+}\in n\operatorname{Pr}_{k}^{t+}\otimes_{(n\operatorname{Pr}_{R}^{t+})^{\mathrm{opp}}}n\operatorname{Pr}_{k}^{t+} ( n − 1 ) Pr k t + ⊗ ( n − 1 ) Pr k t + ∈ n Pr k t + ⊗ ( n Pr R t + ) opp n Pr k t + . The algebraization functor and m m m determine a right module action of n Pr D t + n\operatorname{Pr}_{D}^{t+} n Pr D t + on n Pr k t + ⊗ ( n Pr R t + ) o p p n Pr k t + n\operatorname{Pr}_{k}^{t+}\otimes_{(n\operatorname{Pr}_{R}^{t+})^{\mathrm{opp}}}n\operatorname{Pr}_{k}^{t+} n Pr k t + ⊗ ( n Pr R t + ) opp n Pr k t + . Then ξ \xi ξ is defined to be a (unique) right n Pr D t + n\operatorname{Pr}_{D}^{t+} n Pr D t + -linear functor n Pr D t + → n Pr k t + ⊗ ( n Pr R t + ) o p p n Pr k t + n\operatorname{Pr}_{D}^{t+}\to n\operatorname{Pr}_{k}^{t+}\otimes_{(n\operatorname{Pr}_{R}^{t+})^{\mathrm{opp}}}n\operatorname{Pr}_{k}^{t+} n Pr D t + → n Pr k t + ⊗ ( n Pr R t + ) opp n Pr k t + , which sends ( n − 1 ) Pr D t + (n-1)\operatorname{Pr}_{D}^{t+} ( n − 1 ) Pr D t + to ( n − 1 ) Pr k t + ⊗ ( n − 1 ) Pr k t + (n-1)\operatorname{Pr}_{k}^{t+}\otimes(n-1)\operatorname{Pr}_{k}^{t+} ( n − 1 ) Pr k t + ⊗ ( n − 1 ) Pr k t + .
Construction 4.23
Let R R R be an E n + 2 \mathbb{E}_{n+2} E n + 2 -algebra. We write D = D n + 2 ( R ) D=\mathbb{D}_{n+2}(R) D = D n + 2 ( R ) . We define a left n Pr R t + n\operatorname{Pr}_{R}^{t+} n Pr R t + -module map
η : n Pr k t + ⊗ ( n Pr D t + ) o p p n Pr k t + ≃ n Pr k t + ⊗ ( n Pr D t + ) n Pr k t + → RMod ( n − 1 ) Pr R t + ( n Pr k t + ) . \eta:n\operatorname{Pr}_{k}^{t+}\otimes_{(n\operatorname{Pr}_{D}^{t+})^{\mathrm{opp}}}n\operatorname{Pr}_{k}^{t+}\simeq n\operatorname{Pr}_{k}^{t+}\otimes_{(n\operatorname{Pr}_{D}^{t+})}n\operatorname{Pr}_{k}^{t+}\to\operatorname{RMod}_{(n-1)\operatorname{Pr}_{R}^{t+}}(n\operatorname{Pr}_{k}^{t+}). η : n Pr k t + ⊗ ( n Pr D t + ) opp n Pr k t + ≃ n Pr k t + ⊗ ( n Pr D t + ) n Pr k t + → RMod ( n − 1 ) Pr R t + ( n Pr k t + ) .
This η \eta η is defined as follows. If we look at n Pr k t + n\operatorname{Pr}_{k}^{t+} n Pr k t + on the right-hand side in n Pr k t + ⊗ ( n Pr D t + ) o p p n Pr k t + n\operatorname{Pr}_{k}^{t+}\otimes_{(n\operatorname{Pr}_{D}^{t+})^{\mathrm{opp}}}n\operatorname{Pr}_{k}^{t+} n Pr k t + ⊗ ( n Pr D t + ) opp n Pr k t + , this has the ( n Pr D t + ) o p p (n\operatorname{Pr}_{D}^{t+})^{\mathrm{opp}} ( n Pr D t + ) opp -n Pr R t + n\operatorname{Pr}_{R}^{t+} n Pr R t + -bimodule structure arising from the universal pairing D ⊗ k R → k D\otimes_{k}R\to k D ⊗ k R → k . In particular, n Pr k t + ⊗ ( n Pr D t + ) o p p n Pr k t + n\operatorname{Pr}_{k}^{t+}\otimes_{(n\operatorname{Pr}_{D}^{t+})^{\mathrm{opp}}}n\operatorname{Pr}_{k}^{t+} n Pr k t + ⊗ ( n Pr D t + ) opp n Pr k t + inherits a right n Pr R t + n\operatorname{Pr}_{R}^{t+} n Pr R t + -module structure. We have a (unique) right n Pr R t + n\operatorname{Pr}_{R}^{t+} n Pr R t + -linear functor
η ′ : n Pr R t + → n Pr k t + ⊗ ( n Pr D t + ) o p p n Pr k t + ≃ RMod ( n − 1 ) Pr k ⊗ D k t + ( n Pr k t + ) , \eta':n\operatorname{Pr}_{R}^{t+}\to n\operatorname{Pr}_{k}^{t+}\otimes_{(n\operatorname{Pr}_{D}^{t+})^{\mathrm{opp}}}n\operatorname{Pr}_{k}^{t+}\simeq\operatorname{RMod}_{(n-1)\operatorname{Pr}_{k\otimes_{D}k}^{t+}}(n\operatorname{Pr}_{k}^{t+}), η ′ : n Pr R t + → n Pr k t + ⊗ ( n Pr D t + ) opp n Pr k t + ≃ RMod ( n − 1 ) Pr k ⊗ D k t + ( n Pr k t + ) ,
which sends ( n − 1 ) Pr R t + (n-1)\operatorname{Pr}_{R}^{t+} ( n − 1 ) Pr R t + to ( n − 1 ) Pr k t + ∈ RMod ( n − 1 ) Pr k ⊗ D k t + ( n Pr k t + ) (n-1)\operatorname{Pr}_{k}^{t+}\in\operatorname{RMod}_{(n-1)\operatorname{Pr}_{k\otimes_{D}k}^{t+}}(n\operatorname{Pr}_{k}^{t+}) ( n − 1 ) Pr k t + ∈ RMod ( n − 1 ) Pr k ⊗ D k t + ( n Pr k t + ) (the ( n − 1 ) Pr k ⊗ D k t + (n-1)\operatorname{Pr}_{k\otimes_{D}k}^{t+} ( n − 1 ) Pr k ⊗ D k t + -module structure on ( n − 1 ) Pr k t + (n-1)\operatorname{Pr}_{k}^{t+} ( n − 1 ) Pr k t + is given by the augmentation ( n − 1 ) Pr k ⊗ D k t + → ( n − 1 ) Pr k t + (n-1)\operatorname{Pr}_{k\otimes_{D}k}^{t+}\to(n-1)\operatorname{Pr}_{k}^{t+} ( n − 1 ) Pr k ⊗ D k t + → ( n − 1 ) Pr k t + ). Taking duals Fun n Pr k t + ( − , n Pr k t + ) \operatorname{Fun}_{n\operatorname{Pr}_{k}^{t+}}(-,n\operatorname{Pr}_{k}^{t+}) Fun n Pr k t + ( − , n Pr k t + ) we have
η : LMod ( n − 1 ) Pr k ⊗ D k t + ( n Pr k t + ) ≃ n Pr k t + ⊗ n Pr D t + n Pr k t + → p RMod ( n − 1 ) Pr R t + ( n Pr k t + ) \eta:\operatorname{LMod}_{(n-1)\operatorname{Pr}_{k\otimes_{D}k}^{t+}}(n\operatorname{Pr}_{k}^{t+})\simeq n\operatorname{Pr}_{k}^{t+}\otimes_{n\operatorname{Pr}_{D}^{t+}}n\operatorname{Pr}_{k}^{t+}\xrightarrow{p}\operatorname{RMod}_{(n-1)\operatorname{Pr}_{R}^{t+}}(n\operatorname{Pr}_{k}^{t+}) η : LMod ( n − 1 ) Pr k ⊗ D k t + ( n Pr k t + ) ≃ n Pr k t + ⊗ n Pr D t + n Pr k t + p RMod ( n − 1 ) Pr R t + ( n Pr k t + )
which is left n Pr R t + n\operatorname{Pr}_{R}^{t+} n Pr R t + -linear. See Lemma 4.16 for the equivalence.
The relevant feature of ξ \xi ξ and η \eta η is that ξ \xi ξ is a right n Pr D t + n\operatorname{Pr}_{D}^{t+} n Pr D t + -module map, and η \eta η is a left n Pr R t + n\operatorname{Pr}_{R}^{t+} n Pr R t + -module map.
Lemma 4.24
Let ξ : n Pr D n + 2 ( R ) t + → n Pr k t + ⊗ ( n Pr R t + ) o p p n Pr k t + ≃ RMod ( n − 1 ) Pr k ⊗ R k t + ( n Pr k t + ) \xi:n\operatorname{Pr}_{\mathbb{D}_{n+2}(R)}^{t+}\to n\operatorname{Pr}_{k}^{t+}\otimes_{(n\operatorname{Pr}_{R}^{t+})^{\mathrm{opp}}}n\operatorname{Pr}_{k}^{t+}\simeq\operatorname{RMod}_{(n-1)\operatorname{Pr}_{k\otimes_{R}k}^{t+}}(n\operatorname{Pr}_{k}^{t+}) ξ : n Pr D n + 2 ( R ) t + → n Pr k t + ⊗ ( n Pr R t + ) opp n Pr k t + ≃ RMod ( n − 1 ) Pr k ⊗ R k t + ( n Pr k t + ) be the morphism in Construction 4.22. See Lemma 4.16 for the latter equivalence. We consider
Ψ k ⊗ R k , n : n Pr D n + 1 ( k ⊗ R k ) t + = LMod ( n − 1 ) Pr D n + 1 ( k ⊗ R k ) t + ( n Pr k t + ) → RMod ( n − 1 ) Pr k ⊗ R k t + ( n Pr k t + ) \Psi_{k\otimes_{R}k,n}:n\operatorname{Pr}_{\mathbb{D}_{n+1}(k\otimes_{R}k)}^{t+}=\operatorname{LMod}_{(n-1)\operatorname{Pr}_{\mathbb{D}_{n+1}(k\otimes_{R}k)}^{t+}}(n\operatorname{Pr}_{k}^{t+})\to\operatorname{RMod}_{(n-1)\operatorname{Pr}_{k\otimes_{R}k}^{t+}}(n\operatorname{Pr}_{k}^{t+}) Ψ k ⊗ R k , n : n Pr D n + 1 ( k ⊗ R k ) t + = LMod ( n − 1 ) Pr D n + 1 ( k ⊗ R k ) t + ( n Pr k t + ) → RMod ( n − 1 ) Pr k ⊗ R k t + ( n Pr k t + )
given informally by ( n − 1 ) Pr k t + ⊗ ( n − 1 ) Pr D n + 1 ( k ⊗ R k ) t + ( − ) (n-1)\operatorname{Pr}_{k}^{t+}\otimes_{(n-1)\operatorname{Pr}_{\mathbb{D}_{n+1}(k\otimes_{R}k)}^{t+}}(-) ( n − 1 ) Pr k t + ⊗ ( n − 1 ) Pr D n + 1 ( k ⊗ R k ) t + ( − ) . Here ( n − 1 ) Pr k t + (n-1)\operatorname{Pr}_{k}^{t+} ( n − 1 ) Pr k t + is regarded as a ( ( n − 1 ) Pr D n + 1 ( k ⊗ R k ) t + ) o p p ((n-1)\operatorname{Pr}_{\mathbb{D}_{n+1}(k\otimes_{R}k)}^{t+})^{\mathrm{opp}} (( n − 1 ) Pr D n + 1 ( k ⊗ R k ) t + ) opp -( n − 1 ) Pr k ⊗ R k t + (n-1)\operatorname{Pr}_{k\otimes_{R}k}^{t+} ( n − 1 ) Pr k ⊗ R k t + -bimodule determined by ( n − 1 ) Pr k t + ⊗ ( n − 1 ) Pr D n + 1 ( k ⊗ R k ) t + ( n − 1 ) Pr k t + → ( n − 1 ) Pr k t + (n-1)\operatorname{Pr}_{k}^{t+}\otimes_{(n-1)\operatorname{Pr}_{\mathbb{D}_{n+1}(k\otimes_{R}k)}^{t+}}(n-1)\operatorname{Pr}_{k}^{t+}\to(n-1)\operatorname{Pr}_{k}^{t+} ( n − 1 ) Pr k t + ⊗ ( n − 1 ) Pr D n + 1 ( k ⊗ R k ) t + ( n − 1 ) Pr k t + → ( n − 1 ) Pr k t + arising from the universal pairing ( k ⊗ R k ) ⊗ k D n + 1 ( k ⊗ R k ) → k (k\otimes_{R}k)\otimes_{k}\mathbb{D}_{n+1}(k\otimes_{R}k)\to k ( k ⊗ R k ) ⊗ k D n + 1 ( k ⊗ R k ) → k . Then ξ \xi ξ is equivalent to Ψ k ⊗ R k , n \Psi_{k\otimes_{R}k,n} Ψ k ⊗ R k , n as a morphism in ( n + 1 ) Pr k t + (n+1)\operatorname{Pr}_{k}^{t+} ( n + 1 ) Pr k t + .
Proof. Write D = D n + 2 ( R ) D=\mathbb{D}_{n+2}(R) D = D n + 2 ( R ) and Ψ = Ψ k ⊗ R k , n \Psi=\Psi_{k\otimes_{R}k,n} Ψ = Ψ k ⊗ R k , n . We first note that ξ \xi ξ carries ( n − 1 ) Pr D t + (n-1)\operatorname{Pr}_{D}^{t+} ( n − 1 ) Pr D t + to ( n − 1 ) Pr k t + ⊗ ( n − 1 ) Pr k t + (n-1)\operatorname{Pr}_{k}^{t+}\otimes(n-1)\operatorname{Pr}_{k}^{t+} ( n − 1 ) Pr k t + ⊗ ( n − 1 ) Pr k t + . By definition Ψ \Psi Ψ also carries ( n − 1 ) Pr D t + (n-1)\operatorname{Pr}_{D}^{t+} ( n − 1 ) Pr D t + to ( n − 1 ) Pr k t + ⊗ ( n − 1 ) Pr k t + (n-1)\operatorname{Pr}_{k}^{t+}\otimes(n-1)\operatorname{Pr}_{k}^{t+} ( n − 1 ) Pr k t + ⊗ ( n − 1 ) Pr k t + . To prove the assertion, we look at the endomorphism algebras. Recall the sequence of adjoint pairs from Lemma 4.8:
Alg n ( θ Pr R t + ) ⟵ ⟶ Alg n + 1 ( Mod k ) ⟵ ⟶ Alg n ( Pr k t + ) → U ← V ( ( n + 1 ) Pr k t + ) n Pr k t + o p . \operatorname{Alg}_{n}(\theta_{\operatorname{Pr}_{R}^{t+}})\ \substack{\longleftarrow\\[-0.3em]\longrightarrow}\ \operatorname{Alg}_{n+1}(\operatorname{Mod}_{k})\ \substack{\longleftarrow\\[-0.3em]\longrightarrow}\ \operatorname{Alg}_{n}(\operatorname{Pr}_{k}^{t+})\ \substack{\xrightarrow{U}\\[-0.3em]\xleftarrow{V}}\ ((n+1)\operatorname{Pr}_{k}^{t+})^{\mathrm{op}}_{n\operatorname{Pr}_{k}^{t+}}. Alg n ( θ Pr R t + ) ⟵ ⟶ Alg n + 1 ( Mod k ) ⟵ ⟶ Alg n ( Pr k t + ) U V (( n + 1 ) Pr k t + ) n Pr k t + op .
See Definition 3.22 for θ 0 P r k t + , 0 \theta_{0}^{\mathrm{Pr}^{t+}_{k},0} θ 0 Pr k t + , 0 . We regard RMod ( n − 1 ) P r k ⊗ R k t + ( n P r k t + ) \operatorname{RMod}_{(n-1)\mathrm{Pr}^{t+}_{k\otimes_{R}k}}(n\mathrm{Pr}^{t+}_{k}) RMod ( n − 1 ) Pr k ⊗ R k t + ( n Pr k t + ) with the specified object ( n − 1 ) P r k t + (n-1)\mathrm{Pr}^{t+}_{k} ( n − 1 ) Pr k t + as an object of ( ( n + 1 ) P r k t + ) n P r k t + / \left((n+1)\mathrm{Pr}^{t+}_{k}\right)_{n\mathrm{Pr}^{t+}_{k}/} ( ( n + 1 ) Pr k t + ) n Pr k t + / (see Lemma 4.11 ). Let B B B be the image of P = ( RMod ( n − 1 ) P r k ⊗ R k t + ( n P r k t + ) , ( n − 1 ) P r k t + ) P=\left(\operatorname{RMod}_{(n-1)\mathrm{Pr}^{t+}_{k\otimes_{R}k}}(n\mathrm{Pr}^{t+}_{k}),(n-1)\mathrm{Pr}^{t+}_{k}\right) P = ( RMod ( n − 1 ) Pr k ⊗ R k t + ( n Pr k t + ) , ( n − 1 ) Pr k t + ) in Alg n + 1 ( Mod k ) \operatorname{Alg}_{n+1}(\operatorname{Mod}_{k}) Alg n + 1 ( Mod k ) under the composite of right adjoint functors. By the adjoint pairs, we have a canonical equivalence of mapping spaces
α : Map Alg n + 1 ( Mod k ) ( C , B ) ≃ Map ( n + 1 ) P r k t + ) n P r k t + / ( ( n P r C t + , ( n − 1 ) P r C t + ) , P ) . \alpha:\operatorname{Map}_{\operatorname{Alg}_{n+1}(\operatorname{Mod}_{k})}(C,B)\simeq\operatorname{Map}_{\left(n+1)\mathrm{Pr}^{t+}_{k}\right)_{n\mathrm{Pr}^{t+}_{k}/}}\left((n\mathrm{Pr}^{t+}_{C},(n-1)\mathrm{Pr}^{t+}_{C}),P\right). α : Map Alg n + 1 ( Mod k ) ( C , B ) ≃ Map ( n + 1 ) Pr k t + ) n Pr k t + / ( ( n Pr C t + , ( n − 1 ) Pr C t + ) , P ) .
By Corollary 4.15, there exists an equivalence B ≃ D n + 1 ( k ⊗ R k ) ≃ D n + 2 ( R ) = D B\simeq\mathbb{D}_{n+1}(k\otimes_{R}k)\simeq\mathbb{D}_{n+2}(R)=D B ≃ D n + 1 ( k ⊗ R k ) ≃ D n + 2 ( R ) = D in Alg n + 1 ( Mod k ) \operatorname{Alg}_{n+1}(\operatorname{Mod}_{k}) Alg n + 1 ( Mod k ) . The universal map ( n P r D t + , ( n − 1 ) P r D t + ) → P (n\mathrm{Pr}^{t+}_{D},(n-1)\mathrm{Pr}^{t+}_{D})\to P ( n Pr D t + , ( n − 1 ) Pr D t + ) → P is induced by the universal morphism
u : ( ( n − 1 ) P r D t + ) ⊗ o p p ⟶ End P ( ( n − 1 ) P r k t + ) u:((n-1)\mathrm{Pr}^{t+}_{D})^{\otimes\mathrm{opp}}\longrightarrow\operatorname{End}_{P}((n-1)\mathrm{Pr}^{t+}_{k}) u : (( n − 1 ) Pr D t + ) ⊗ opp ⟶ End P (( n − 1 ) Pr k t + )
in Alg 1 ( n P r k t + ) \operatorname{Alg}_{1}(n\mathrm{Pr}^{t+}_{k}) Alg 1 ( n Pr k t + ) , which induces an equivalence
Map Alg n + 1 ( Mod k ) ( C , D ) ≃ Map Alg 1 ( n P r k t + ) ( ( ( n − 1 ) P r C t + ) ⊗ o p p , End P ( ( n − 1 ) P r k t + ) ) . \operatorname{Map}_{\operatorname{Alg}_{n+1}(\operatorname{Mod}_{k})}(C,D)\simeq\operatorname{Map}_{\operatorname{Alg}_{1}(n\mathrm{Pr}^{t+}_{k})}\left(((n-1)\mathrm{Pr}^{t+}_{C})^{\otimes\mathrm{opp}},\operatorname{End}_{P}((n-1)\mathrm{Pr}^{t+}_{k})\right). Map Alg n + 1 ( Mod k ) ( C , D ) ≃ Map Alg 1 ( n Pr k t + ) ( (( n − 1 ) Pr C t + ) ⊗ opp , End P (( n − 1 ) Pr k t + ) ) .
given by f ↦ u ∘ ( n − 1 ) P r f t + f\mapsto u\circ(n-1)\mathrm{Pr}^{t+}_{f} f ↦ u ∘ ( n − 1 ) Pr f t + . As explained in Remark 4.14, u u u corresponds to the pairing ( n − 1 ) P r k ⊗ R k t + ⊗ ( n − 1 ) P r k t + ( n − 1 ) P r D t + → ( n − 1 ) P r k t + (n-1)\mathrm{Pr}^{t+}_{k\otimes_{R}k}\otimes_{(n-1)\mathrm{Pr}^{t+}_{k}}(n-1)\mathrm{Pr}^{t+}_{D}\to(n-1)\mathrm{Pr}^{t+}_{k} ( n − 1 ) Pr k ⊗ R k t + ⊗ ( n − 1 ) Pr k t + ( n − 1 ) Pr D t + → ( n − 1 ) Pr k t + induced by the universal pairing ( k ⊗ R k ) ⊗ k D n + 1 ( k ⊗ R k ) → k (k\otimes_{R}k)\otimes_{k}\mathbb{D}_{n+1}(k\otimes_{R}k)\to k ( k ⊗ R k ) ⊗ k D n + 1 ( k ⊗ R k ) → k . Thus, u ≃ Ψ u\simeq\Psi u ≃ Ψ since u u u is the functor defined by the integral kernel ( n − 1 ) P r k t + (n-1)\mathrm{Pr}^{t+}_{k} ( n − 1 ) Pr k t + whose ( ( n − 1 ) P r D t + ) ⊗ o p p ((n-1)\mathrm{Pr}^{t+}_{D})^{\otimes\mathrm{opp}} (( n − 1 ) Pr D t + ) ⊗ opp -( ( n − 1 ) P r k ⊗ R k t + ) ((n-1)\mathrm{Pr}^{t+}_{k\otimes_{R}k}) (( n − 1 ) Pr k ⊗ R k t + ) -bimodule structure is given by ( n − 1 ) P r k ⊗ R k t + ⊗ ( n − 1 ) P r k t + ( n − 1 ) P r k t + → ( n − 1 ) P r D t + ⊗ ( n − 1 ) P r k t + (n-1)\mathrm{Pr}^{t+}_{k\otimes_{R}k}\otimes_{(n-1)\mathrm{Pr}^{t+}_{k}}(n-1)\mathrm{Pr}^{t+}_{k}\to(n-1)\mathrm{Pr}^{t+}_{D}\otimes(n-1)\mathrm{Pr}^{t+}_{k} ( n − 1 ) Pr k ⊗ R k t + ⊗ ( n − 1 ) Pr k t + ( n − 1 ) Pr k t + → ( n − 1 ) Pr D t + ⊗ ( n − 1 ) Pr k t + .
We note that ξ \xi ξ is the composite of the algebraization functor n P r D t + → End n P r R t + l ( n P r k t + ) n\mathrm{Pr}^{t+}_{D}\to\operatorname{End}^{l}_{n\mathrm{Pr}^{t+}_{R}}(n\mathrm{Pr}^{t+}_{k}) n Pr D t + → End n Pr R t + l ( n Pr k t + ) and the equivalence m : End n P r R t + l ( n P r k t + ) ≃ n P r k t + ⊗ n P r R t + n P r k t + m:\operatorname{End}^{l}_{n\mathrm{Pr}^{t+}_{R}}(n\mathrm{Pr}^{t+}_{k})\simeq n\mathrm{Pr}^{t+}_{k}\otimes_{n\mathrm{Pr}^{t+}_{R}}n\mathrm{Pr}^{t+}_{k} m : End n Pr R t + l ( n Pr k t + ) ≃ n Pr k t + ⊗ n Pr R t + n Pr k t + in ( n + 1 ) P r k t + (n+1)\mathrm{Pr}^{t+}_{k} ( n + 1 ) Pr k t + . The equivalence m m m sends the identity functor to ( n − 1 ) P r k t + (n-1)\mathrm{Pr}^{t+}_{k} ( n − 1 ) Pr k t + . From the construction of the algebraization functor, if we regard ξ \xi ξ as a morphism ( n P r D t + , ( n − 1 ) P r D t + ) → P (n\mathrm{Pr}^{t+}_{D},(n-1)\mathrm{Pr}^{t+}_{D})\to P ( n Pr D t + , ( n − 1 ) Pr D t + ) → P in ( ( n + 1 ) P r k t + ) n P r k t + / \left((n+1)\mathrm{Pr}^{t+}_{k}\right)_{n\mathrm{Pr}^{t+}_{k}/} ( ( n + 1 ) Pr k t + ) n Pr k t + / , then it corresponds to the identity map D → D D\to D D → D in Alg n + 1 ( Mod k ) \operatorname{Alg}_{n+1}(\operatorname{Mod}_{k}) Alg n + 1 ( Mod k ) through α \alpha α . Consequently, ξ ≃ Ψ \xi\simeq\Psi ξ ≃ Ψ .
Lemma 4.25
Lemma 4.25. Let η : LMod ( n − 1 ) P r k ⊗ D k t + ( n P r k t + ) → RMod ( n − 1 ) P r R t + ( n P r k t + ) \eta:\operatorname{LMod}_{(n-1)\mathrm{Pr}^{t+}_{k\otimes_{D}k}}(n\mathrm{Pr}^{t+}_{k})\to\operatorname{RMod}_{(n-1)\mathrm{Pr}^{t+}_{R}}(n\mathrm{Pr}^{t+}_{k}) η : LMod ( n − 1 ) Pr k ⊗ D k t + ( n Pr k t + ) → RMod ( n − 1 ) Pr R t + ( n Pr k t + ) be the morphism in Construction 4.23. We consider
Ψ R , n : LMod ( n − 1 ) P r k ⊗ D k t + ( n P r k t + ) → RMod ( n − 1 ) P r R t + ( n P r k t + ) \Psi_{R,n}:\operatorname{LMod}_{(n-1)\mathrm{Pr}^{t+}_{k\otimes_{D}k}}(n\mathrm{Pr}^{t+}_{k})\to\operatorname{RMod}_{(n-1)\mathrm{Pr}^{t+}_{R}}(n\mathrm{Pr}^{t+}_{k}) Ψ R , n : LMod ( n − 1 ) Pr k ⊗ D k t + ( n Pr k t + ) → RMod ( n − 1 ) Pr R t + ( n Pr k t + )
given informally by ( − ) ⊗ ( n − 1 ) P r ( k ⊗ D k ) t + ( n − 1 ) P r k t + (-)\otimes_{(n-1)\mathrm{Pr}^{t+}_{(k\otimes_{D}k)}}(n-1)\mathrm{Pr}^{t+}_{k} ( − ) ⊗ ( n − 1 ) Pr ( k ⊗ D k ) t + ( n − 1 ) Pr k t + . We note that k ⊗ D k = k ⊗ D n + 2 ( R ) k ≃ D n + 1 ( R ) k\otimes_{D}k=k\otimes_{\mathbb{D}_{n+2}(R)}k\simeq\mathbb{D}_{n+1}(R) k ⊗ D k = k ⊗ D n + 2 ( R ) k ≃ D n + 1 ( R ) (see Proposition 5.1). Here ( n − 1 ) P r k t + (n-1)\mathrm{Pr}^{t+}_{k} ( n − 1 ) Pr k t + is regarded as a ( n − 1 ) P r ( k ⊗ D k ) t + (n-1)\mathrm{Pr}^{t+}_{(k\otimes_{D}k)} ( n − 1 ) Pr ( k ⊗ D k ) t + -( n − 1 ) ( P r R t + ) ⊗ o p p (n-1)(\mathrm{Pr}^{t+}_{R})^{\otimes\mathrm{opp}} ( n − 1 ) ( Pr R t + ) ⊗ opp -bimodule determined by ( n − 1 ) P r R t + ⊗ ( n − 1 ) P r k t + ( n − 1 ) P r k t + → ( n − 1 ) P r k t + (n-1)\mathrm{Pr}^{t+}_{R}\otimes_{(n-1)\mathrm{Pr}^{t+}_{k}}(n-1)\mathrm{Pr}^{t+}_{k}\to(n-1)\mathrm{Pr}^{t+}_{k} ( n − 1 ) Pr R t + ⊗ ( n − 1 ) Pr k t + ( n − 1 ) Pr k t + → ( n − 1 ) Pr k t + arising from the universal pairing R ⊗ k D n + 1 ( R ) → k R\otimes_{k}\mathbb{D}_{n+1}(R)\to k R ⊗ k D n + 1 ( R ) → k . Then η \eta η is equivalent to Ψ R , n \Psi_{R,n} Ψ R , n as a morphism in ( n + 1 ) P r k t + (n+1)\mathrm{Pr}^{t+}_{k} ( n + 1 ) Pr k t + .
Proof. Let η ′ : n P r R t + → RMod ( n − 1 ) P r k ⊗ D k t + ( n P r k t + ) \eta':n\mathrm{Pr}^{t+}_{R}\to\operatorname{RMod}_{(n-1)\mathrm{Pr}^{t+}_{k\otimes_{D}k}}(n\mathrm{Pr}^{t+}_{k}) η ′ : n Pr R t + → RMod ( n − 1 ) Pr k ⊗ D k t + ( n Pr k t + ) be a (unique) right n P r R t + n\mathrm{Pr}^{t+}_{R} n Pr R t + -linear functor which sends ( n − 1 ) P r R t + (n-1)\mathrm{Pr}^{t+}_{R} ( n − 1 ) Pr R t + to ( n − 1 ) P r k t + (n-1)\mathrm{Pr}^{t+}_{k} ( n − 1 ) Pr k t + . The morphism η \eta η is defined to be its n P r k t + n\mathrm{Pr}^{t+}_{k} n Pr k t + -linear dual. By Lemma 4.24, η ′ \eta' η ′ is determined by n P r k t + ⊗ n P r R t + ( − ) n\mathrm{Pr}^{t+}_{k}\otimes_{n\mathrm{Pr}^{t+}_{R}}(-) n Pr k t + ⊗ n Pr R t + ( − ) , where the bimodule structure on n P r k t + n\mathrm{Pr}^{t+}_{k} n Pr k t + is induced by the universal E n + 1 \mathbf{E}_{n+1} E n + 1 -Koszul dual pairing
R ⊗ k D n + 1 ( R ) ≃ D n + 1 ( D n + 1 ( R ) ) ⊗ D n + 1 ( R ) → k . R\otimes_{k}\mathbb{D}_{n+1}(R)\simeq\mathbb{D}_{n+1}(\mathbb{D}_{n+1}(R))\otimes\mathbb{D}_{n+1}(R)\to k. R ⊗ k D n + 1 ( R ) ≃ D n + 1 ( D n + 1 ( R )) ⊗ D n + 1 ( R ) → k .
(R R R is almost Artin and R ≃ D n + 1 ( D n + 1 ( R ) ) R\simeq\mathbb{D}_{n+1}(\mathbb{D}_{n+1}(R)) R ≃ D n + 1 ( D n + 1 ( R )) ). See Proposition 5.3 Now our assertion follows from a general observation of Lemma 4.26.
Lemma 4.26
Lemma 4.26. Let C ⊗ \mathcal{C}^{\otimes} C ⊗ be an object of CAlg ( Cat ^ ∞ ) \operatorname{CAlg}(\widehat{\operatorname{Cat}}_{\infty}) CAlg ( Cat ∞ ) , that is, a symmetric monoidal ∞ \infty ∞ -category such that C \mathcal{C} C has small colimits and the tensor product preserves small colimits separately in each variable. Let A A A and B B B be E 1 \mathbf{E}_{1} E 1 -algebra objects in C ⊗ \mathcal{C}^{\otimes} C ⊗ . Let LMod A ( C ) \operatorname{LMod}_{A}(\mathcal{C}) LMod A ( C ) and LMod B ( C ) \operatorname{LMod}_{B}(\mathcal{C}) LMod B ( C ) be the ∞ \infty ∞ -categories of left A A A -modules and left B B B -modules, respectively. Let F : LMod A ( C ) → LMod B ( C ) F:\operatorname{LMod}_{A}(\mathcal{C})\to\operatorname{LMod}_{B}(\mathcal{C}) F : LMod A ( C ) → LMod B ( C ) be the functor determined by a B B B -A A A -bimodule B M A B M_{A} B M A . Namely, F F F is given by P ↦ B M A ⊗ A P P\mapsto{}_{B}M_{A}\otimes_{A}P P ↦ B M A ⊗ A P . Let Fun ( LMod A ( C ) , C ) \operatorname{Fun}(\operatorname{LMod}_{A}(\mathcal{C}),\mathcal{C}) Fun ( LMod A ( C ) , C ) and Fun C ( LMod B ( C ) , C ) \operatorname{Fun}_{\mathcal{C}}(\operatorname{LMod}_{B}(\mathcal{C}),\mathcal{C}) Fun C ( LMod B ( C ) , C ) denote the C \mathcal{C} C -linear functor categories. Recall the equivalences RMod A ( C ) ≃ Fun C ( LMod A ( C ) , C ) \operatorname{RMod}_{A}(\mathcal{C}) \simeq\operatorname{Fun}_{\mathcal{C}}(\operatorname{LMod}_{A}(\mathcal{C}),\mathcal{C}) RMod A ( C ) ≃ Fun C ( LMod A ( C ) , C ) and RMod B ( C ) ≃ Fun C ( LMod B ( C ) , C ) \operatorname{RMod}_{B}(\mathcal{C}) \simeq\operatorname{Fun}_{\mathcal{C}}(\operatorname{LMod}_{B}(\mathcal{C}),\mathcal{C}) RMod B ( C ) ≃ Fun C ( LMod B ( C ) , C ) . We consider
RMod B ( C ) ≃ Fun C ( LMod B ( C ) , C ) → Fun C ( LMod A ( C ) , C ) ≃ RMod A ( C ) . \begin{aligned}
\operatorname{RMod}_{B}(\mathcal{C}) &\simeq\operatorname{Fun}_{\mathcal{C}}(\operatorname{LMod}_{B}(\mathcal{C}),\mathcal{C})\\
&\to\operatorname{Fun}_{\mathcal{C}}(\operatorname{LMod}_{A}(\mathcal{C}),\mathcal{C})\\
&\simeq\operatorname{RMod}_{A}(\mathcal{C}).
\end{aligned} RMod B ( C ) ≃ Fun C ( LMod B ( C ) , C ) → Fun C ( LMod A ( C ) , C ) ≃ RMod A ( C ) .
Then the composite F ∨ : RMod B ( C ) → RMod A ( C ) F^{\vee}:\operatorname{RMod}_{B}(\mathcal{C})\to\operatorname{RMod}_{A}(\mathcal{C}) F ∨ : RMod B ( C ) → RMod A ( C ) is determined by the B B B -A A A -bimodule B M A {}_{B}M_{A} B M A . Namely, F ∨ F^{\vee} F ∨ is given by Q ↦ Q ⊗ B B M A Q\mapsto Q\otimes_{B}{}_{B}M_{A} Q ↦ Q ⊗ B B M A .
Proof. For N B ∈ RMod B ( C ) N_{B}\in\operatorname{RMod}_{B}(\mathcal{C}) N B ∈ RMod B ( C ) , the corresponding functor LMod B ( C ) → C \operatorname{LMod}_{B}(\mathcal{C})\to\mathcal{C} LMod B ( C ) → C is given informally by N B ⊗ B ( − ) N_{B}\otimes_{B}(-) N B ⊗ B ( − ) . Its composition with F F F is N B ⊗ B B M A ⊗ A ( − ) : LMod A ( C ) → C N_{B}\otimes_{B}{}_{B}M_{A}\otimes_{A}(-):\operatorname{LMod}_{A}(\mathcal{C})\to\mathcal{C} N B ⊗ B B M A ⊗ A ( − ) : LMod A ( C ) → C . Thus, F ∨ F^{\vee} F ∨ is given informally by N B ↦ N B ⊗ B B M A N_{B}\mapsto N_{B}\otimes_{B}{}_{B}M_{A} N B ↦ N B ⊗ B B M A . Namely, F ∨ F^{\vee} F ∨ is determined by ( − ) ⊗ B B M A (-)\otimes_{B}{}_{B}M_{A} ( − ) ⊗ B B M A . This is our assertion.
Proof of Proposition 4.21
We write D = D n + 2 ( R ) D=\mathbb{D}_{n+2}(R) D = D n + 2 ( R ) . For E ∈ LMod n Prt D n + 2 ( R ) ± ( ( n + 1 ) Prt ± ) \mathcal{E}\in\operatorname{LMod}_{n\operatorname{Prt}^{\pm}_{\mathbb{D}_{n+2}(R)}((n+1)\operatorname{Prt}^{\pm})} E ∈ LMod n Prt D n + 2 ( R ) ± (( n + 1 ) Prt ± ) we define
U E : E → ( n Prt k + ⊗ n Prt D + E ) ⊗ n Prt R + n Prt R + ≃ ( n Prt k + ⊗ ( n Prt R + ) o p p n Prt k + ) ⊗ n Prt D + E U_{\mathcal{E}}:\mathcal{E}\to(n\operatorname{Prt}^{+}_{k}\otimes_{n\operatorname{Prt}^{+}_{D}}\mathcal{E})\otimes_{n\operatorname{Prt}^{+}_{R}}n\operatorname{Prt}^{+}_{R}\simeq(n\operatorname{Prt}^{+}_{k}\otimes_{(n\operatorname{Prt}^{+}_{R})^{\mathrm{opp}}}n\operatorname{Prt}^{+}_{k})\otimes_{n\operatorname{Prt}^{+}_{D}}\mathcal{E} U E : E → ( n Prt k + ⊗ n Prt D + E ) ⊗ n Prt R + n Prt R + ≃ ( n Prt k + ⊗ ( n Prt R + ) opp n Prt k + ) ⊗ n Prt D + E
to be the morphism obtained from ξ : n Prt D + → n Prt k + ⊗ ( n Prt R + ) o p p n Prt k + \xi:n\operatorname{Prt}^{+}_{D}\to n\operatorname{Prt}^{+}_{k}\otimes_{(n\operatorname{Prt}^{+}_{R})^{\mathrm{opp}}}n\operatorname{Prt}^{+}_{k} ξ : n Prt D + → n Prt k + ⊗ ( n Prt R + ) opp n Prt k + by taking the tensor product over n Prt D + n\operatorname{Prt}^{+}_{D} n Prt D + . For D ∈ RMod n Prt R + ( ( n + 1 ) Prt + ) \mathcal{D}\in\operatorname{RMod}_{n\operatorname{Prt}^{+}_{R}}((n+1)\operatorname{Prt}^{+}) D ∈ RMod n Prt R + (( n + 1 ) Prt + ) we define
V D : n Prt k + ⊗ n Prt R + ( D ⊗ n Prt R + n Prt k + ) ≃ D ⊗ n Prt R + ( n Prt k + ⊗ ( n Prt D + ) o p p n Prt k + ) → D V_{\mathcal{D}}:n\operatorname{Prt}^{+}_{k}\otimes_{n\operatorname{Prt}^{+}_{R}}(\mathcal{D}\otimes_{n\operatorname{Prt}^{+}_{R}}n\operatorname{Prt}^{+}_{k})\simeq\mathcal{D}\otimes_{n\operatorname{Prt}^{+}_{R}}(n\operatorname{Prt}^{+}_{k}\otimes_{(n\operatorname{Prt}^{+}_{D})^{\mathrm{opp}}}n\operatorname{Prt}^{+}_{k})\to\mathcal{D} V D : n Prt k + ⊗ n Prt R + ( D ⊗ n Prt R + n Prt k + ) ≃ D ⊗ n Prt R + ( n Prt k + ⊗ ( n Prt D + ) opp n Prt k + ) → D
to be the morphism obtained from η : n Prt k + ⊗ ( n Prt D + ) o p p n Prt k + ≃ n Prt k + ⊗ RMod ( n − 1 ) Prt R + ( n Prt + ) n Prt k + → n Prt R + \eta:n\operatorname{Prt}^{+}_{k}\otimes_{(n\operatorname{Prt}^{+}_{D})^{\mathrm{opp}}}n\operatorname{Prt}^{+}_{k}\simeq n\operatorname{Prt}^{+}_{k}\otimes_{\operatorname{RMod}_{(n-1)\operatorname{Prt}^{+}_{R}}(n\operatorname{Prt}^{+})}n\operatorname{Prt}^{+}_{k}\to n\operatorname{Prt}^{+}_{R} η : n Prt k + ⊗ ( n Prt D + ) opp n Prt k + ≃ n Prt k + ⊗ RMod ( n − 1 ) Prt R + ( n Prt + ) n Prt k + → n Prt R + by taking the tensor product over n Prt R + n\operatorname{Prt}^{+}_{R} n Prt R + .
We set Φ = Φ R , n + 1 \Phi=\Phi_{R,n+1} Φ = Φ R , n + 1 and Ψ = Ψ R , n + 1 \Psi=\Psi_{R,n+1} Ψ = Ψ R , n + 1 . They determine the natural transformations U : id → Φ ∘ Ψ U:\operatorname{id}\to\Phi\circ\Psi U : id → Φ ∘ Ψ and V : Ψ ∘ Φ → id V:\Psi\circ\Phi\to\operatorname{id} V : Ψ ∘ Φ → id . We will prove that both U U U and V V V are equivalences after the completion. Suppose that E ∈ LMod n Prt D n + 2 ( R ) ± ( ( n + 1 ) Prt ± ) \mathcal{E}\in\operatorname{LMod}_{n\operatorname{Prt}^{\pm}_{\mathbb{D}_{n+2}(R)}((n+1)\operatorname{Prt}^{\pm})} E ∈ LMod n Prt D n + 2 ( R ) ± (( n + 1 ) Prt ± ) . The completion of U E U_{\mathcal{E}} U E determines
U ^ E : E = n Prt D ± ⊗ ^ n Prt D ± E → ( n Prt k ± ⊗ ^ ( n Prt R ± ) o p p n Prt k ± ) ⊗ ^ n Prt D ± E . \widehat{U}_{\mathcal{E}}:\mathcal{E}=n\operatorname{Prt}^{\pm}_{D}\widehat{\otimes}_{n\operatorname{Prt}^{\pm}_{D}}\mathcal{E}\to(n\operatorname{Prt}^{\pm}_{k}\widehat{\otimes}_{(n\operatorname{Prt}^{\pm}_{R})^{\mathrm{opp}}}n\operatorname{Prt}^{\pm}_{k})\widehat{\otimes}_{n\operatorname{Prt}^{\pm}_{D}}\mathcal{E}. U E : E = n Prt D ± ⊗ n Prt D ± E → ( n Prt k ± ⊗ ( n Prt R ± ) opp n Prt k ± ) ⊗ n Prt D ± E .
According to Lemma 4.24 and the inductive assumption, the completion of ξ \xi ξ induces an equivalence n Prt D ± ≃ n Prt k ± ⊗ ^ ( n Prt R ± ) o p p n Prt k ± n\operatorname{Prt}^{\pm}_{D}\simeq n\operatorname{Prt}^{\pm}_{k}\widehat{\otimes}_{(n\operatorname{Prt}^{\pm}_{R})^{\mathrm{opp}}}n\operatorname{Prt}^{\pm}_{k} n Prt D ± ≃ n Prt k ± ⊗ ( n Prt R ± ) opp n Prt k ± . Thus, U ^ E \widehat{U}_{\mathcal{E}} U E is an equivalence. Next, we suppose that D ∈ RMod n Prt R + ( ( n + 1 ) Prt ± ) \mathcal{D}\in\operatorname{RMod}_{n\operatorname{Prt}^{+}_{R}}((n+1)\operatorname{Prt}^{\pm}) D ∈ RMod n Prt R + (( n + 1 ) Prt ± ) . The completion of V D V_{\mathcal{D}} V D determines
V ^ D : D ⊗ ^ n Prt R ± ( n Prt k ± ⊗ ^ ( n Prt D ± ) o p p n Prt k ± ) → D ⊗ ^ n Prt R ± RMod ( n − 1 ) Prt R ± ( n Prt ± ) ≃ D . \widehat{V}_{\mathcal{D}}:\mathcal{D}\widehat{\otimes}_{n\operatorname{Prt}^{\pm}_{R}}(n\operatorname{Prt}^{\pm}_{k}\widehat{\otimes}_{(n\operatorname{Prt}^{\pm}_{D})^{\mathrm{opp}}}n\operatorname{Prt}^{\pm}_{k})\to\mathcal{D}\widehat{\otimes}_{n\operatorname{Prt}^{\pm}_{R}}\operatorname{RMod}_{(n-1)\operatorname{Prt}^{\pm}_{R}}(n\operatorname{Prt}^{\pm})\simeq\mathcal{D}. V D : D ⊗ n Prt R ± ( n Prt k ± ⊗ ( n Prt D ± ) opp n Prt k ± ) → D ⊗ n Prt R ± RMod ( n − 1 ) Prt R ± ( n Prt ± ) ≃ D .
For the right equivalence we note that ( n − 1 ) Prt R ± (n-1)\operatorname{Prt}^{\pm}_{R} ( n − 1 ) Prt R ± is E 2 \mathrm{E}_{2} E 2 -monoidal so that RMod ( n − 1 ) Prt R ± ( n Prt ± ) \operatorname{RMod}_{(n-1)\operatorname{Prt}^{\pm}_{R}}(n\operatorname{Prt}^{\pm}) RMod ( n − 1 ) Prt R ± ( n Prt ± ) is the opposite algebra of LMod ( n − 1 ) Prt R ± ( n Prt ± ) = n Prt R ± \operatorname{LMod}_{(n-1)\operatorname{Prt}^{\pm}_{R}}(n\operatorname{Prt}^{\pm})=n\operatorname{Prt}^{\pm}_{R} LMod ( n − 1 ) Prt R ± ( n Prt ± ) = n Prt R ± in Alg 1 ( ( n + 1 ) Prt ± ) \operatorname{Alg}_{1}((n+1)\operatorname{Prt}^{\pm}) Alg 1 (( n + 1 ) Prt ± ) (see [20 Proposition 2.15] ). According to Lemma 4.25 and the inductive assumption, the completion of η \eta η induces an equivalence
n Prt k ± ⊗ ^ ( n Prt D ± ) o p p n Prt k ± ≃ n Prt k ± ⊗ ^ n Prt D ± n Prt k ± ≃ n Prt R ± n\operatorname{Prt}^{\pm}_{k}\widehat{\otimes}_{(n\operatorname{Prt}^{\pm}_{D})^{\mathrm{opp}}}n\operatorname{Prt}^{\pm}_{k}\simeq n\operatorname{Prt}^{\pm}_{k}\widehat{\otimes}_{n\operatorname{Prt}^{\pm}_{D}}n\operatorname{Prt}^{\pm}_{k}\simeq n\operatorname{Prt}^{\pm}_{R} n Prt k ± ⊗ ( n Prt D ± ) opp n Prt k ± ≃ n Prt k ± ⊗ n Prt D ± n Prt k ± ≃ n Prt R ±
so that V ^ D \widehat{V}_{\mathcal{D}} V D is an equivalence.
The case of n = 1 n=1 n = 1
We treat the case of n = 1 n=1 n = 1 . Proposition 4.20 says that there exist mutually inverse equivalences
LMod LMod D 2 ( R ) ( Prt k ± ) ≃ RMod LMod R ( Prt k ± ) ( Prt k ± ) \operatorname{LMod}_{\operatorname{LMod}_{\mathbb{D}_{2}(R)}(\operatorname{Prt}^{\pm}_{k})}\simeq\operatorname{RMod}_{\operatorname{LMod}_{R}(\operatorname{Prt}^{\pm}_{k})}(\operatorname{Prt}^{\pm}_{k}) LMod LMod D 2 ( R ) ( Prt k ± ) ≃ RMod LMod R ( Prt k ± ) ( Prt k ± )
which are induced by Φ R , 1 \Phi_{R,1} Φ R , 1 (the left arrow) and Ψ R , 1 \Psi_{R,1} Ψ R , 1 (the right arrow) together with the left completion functor. Precisely, LMod LMod D 2 ( R ) ( Prt k ± ) ( Prt k ± ) → RMod LMod R ( Prt k ± ) ( Prt k ± ) \operatorname{LMod}_{\operatorname{LMod}_{\mathbb{D}_{2}(R)}(\operatorname{Prt}^{\pm}_{k})}(\operatorname{Prt}^{\pm}_{k})\to\operatorname{RMod}_{\operatorname{LMod}_{R}(\operatorname{Prt}^{\pm}_{k})}(\operatorname{Prt}^{\pm}_{k}) LMod LMod D 2 ( R ) ( Prt k ± ) ( Prt k ± ) → RMod LMod R ( Prt k ± ) ( Prt k ± ) is defined by L ^ ( Mod k ⊗ LMod D 2 ( R ) ( − ) ) \widehat{L}(\operatorname{Mod}_{k}\otimes_{\operatorname{LMod}_{\mathbb{D}_{2}(R)}(-)}) L ( Mod k ⊗ LMod D 2 ( R ) ( − ) ) , and RMod LMod R ( Prt k ± ) ( Prt k ± ) → LMod LMod D 2 ( R ) ( Prt k ± ) ( Prt k ± ) \operatorname{RMod}_{\operatorname{LMod}_{R}(\operatorname{Prt}^{\pm}_{k})}(\operatorname{Prt}^{\pm}_{k})\to\operatorname{LMod}_{\operatorname{LMod}_{\mathbb{D}_{2}(R)}(\operatorname{Prt}^{\pm}_{k})}(\operatorname{Prt}^{\pm}_{k}) RMod LMod R ( Prt k ± ) ( Prt k ± ) → LMod LMod D 2 ( R ) ( Prt k ± ) ( Prt k ± ) is given by ( − ) ⊗ LMod R Mod k (-)\otimes_{\operatorname{LMod}_{R}}\operatorname{Mod}_{k} ( − ) ⊗ LMod R Mod k . When R R R is Artin, this is one of the main results of [20 ] . We give a slight generalization to the case when R R R is almost Artin. The argument of [20 Section 3] goes through without essential change, with the following exceptions:
Proposition 5.1. If R R R is an almost Artin E 2 \mathrm{E}_{2} E 2 -algebra, there exists an equivalence k ⊗ D 2 ( R ) k ≃ D 1 ( R ) k\otimes_{\mathbb{D}_{2}(R)}k\simeq\mathbb{D}_{1}(R) k ⊗ D 2 ( R ) k ≃ D 1 ( R ) in Alg 1 ( Mod k ) \operatorname{Alg}_{1}(\operatorname{Mod}_{k}) Alg 1 ( Mod k ) .
Proposition 5.2. Suppose that R R R is an almost Artin E 2 \mathbb{E}_2 E 2 -algebra. There exists a left completion functor η : Mod k ⊗ R Mod E 2 ( R ) Mod k = RMod k ⊗ E 2 ( R ) k → LMod R \eta:\operatorname{Mod}_{k}\otimes_{R\operatorname{Mod}_{\mathbb{E}_2(R)}}\operatorname{Mod}_{k}=\operatorname{RMod}_{k\otimes_{\mathbb{E}_2(R)}k}\to\operatorname{LMod}_{R} η : Mod k ⊗ R Mod E 2 ( R ) Mod k = RMod k ⊗ E 2 ( R ) k → LMod R . Moreover, the target LMod R \operatorname{LMod}_{R} LMod R is a free LMod R \operatorname{LMod}_{R} LMod R -module of rank one.
Proposition 5.1 follows from the proof in [20 Lemma 3.13] if we prove the following:
Proposition 5.3. Let R R R be an almost Artin E 2 \mathbb{E}_2 E 2 -algebra. Then the biduality morphisms R → D 1 D 1 ( R ) R\to\mathbb{D}_1\mathbb{D}_1(R) R → D 1 D 1 ( R ) and R → D 2 D 2 ( R ) R\to\mathbb{D}_2\mathbb{D}_2(R) R → D 2 D 2 ( R ) are equivalences. Moreover, D 1 ( R ) \mathbb{D}_1(R) D 1 ( R ) and D 2 ( R ) \mathbb{D}_2(R) D 2 ( R ) are locally finite. Here regarding R R R as an E 1 \mathbb{E}_1 E 1 -algebra we define D 1 ( R ) \mathbb{D}_1(R) D 1 ( R ) .
Lemma 5.4. Let l l l be a natural number l ≥ 1 l\geq1 l ≥ 1 . Let f : A ′ → A f:A'\to A f : A ′ → A be a morphism of Artin E 1 \mathbb{E}_1 E 1 -algebras. Let n n n be a nonnegative integer. Assume that f f f exhibits A A A as the truncation τ ≥ − n ( A ′ ) \tau^{\geq-n}(A') τ ≥ − n ( A ′ ) . Suppose that H i ( A ′ ) = 0 H^i(A')=0 H i ( A ′ ) = 0 for any i < − n − 1 i<-n-1 i < − n − 1 . Then the induced morphism H i ( k ⊗ A ′ k ) → H i ( k ⊗ A k ) H^i(k\otimes_{A'}k)\to H^i(k\otimes_Ak) H i ( k ⊗ A ′ k ) → H i ( k ⊗ A k ) is an isomorphism of finite dimensional vector spaces for any − n ≤ i ≤ 0 -n\leq i\leq0 − n ≤ i ≤ 0 .
Proof. Under our assumption, A ′ A' A ′ is a square zero extension of A A A by H − n − 1 ( A ′ ) [ n + 1 ] H^{-n-1}(A')[n+1] H − n − 1 ( A ′ ) [ n + 1 ] (cf. [24 7.4.1.28] ) so that f f f is a pullback of the canonical morphism A → A ⊕ H − n − 1 ( A ′ ) [ n + 2 ] A\to A\oplus H^{-n-1}(A')[n+2] A → A ⊕ H − n − 1 ( A ′ ) [ n + 2 ] . Thus, there exists a sequence
A ′ = A ( 0 ) → A ( 1 ) → ⋯ → A ( m ) = A A'=A(0)\to A(1)\to\cdots\to A(m)=A A ′ = A ( 0 ) → A ( 1 ) → ⋯ → A ( m ) = A
in Art l \operatorname{Art}_l Art l such that, for each 0 ≤ i ≤ m − 1 0\leq i\leq m-1 0 ≤ i ≤ m − 1 , there exists a pullback square
A ( i ) → A ( i + 1 ) ↓ ↓ k → k ⊕ k [ n + 2 ] \begin{CD}
A(i) @>>> A(i+1) \\
@VVV @VVV \\
k @>>> k\oplus k[n+2]
\end{CD} A ( i ) ↓ ⏐ k A ( i + 1 ) ↓ ⏐ k ⊕ k [ n + 2 ]
in Art l \operatorname{Art}_l Art l . We may and will assume that A ′ A' A ′ is k × k ⊕ k [ n + 2 ] A k\times_{k\oplus k[n+2]}A k × k ⊕ k [ n + 2 ] A . Then we have the fiber sequence
k [ n + 1 ] → A ′ → A k[n+1]\to A'\to A k [ n + 1 ] → A ′ → A
in Mod A ′ \operatorname{Mod}_{A'} Mod A ′ . We note that k ⊗ A ′ k ≃ ( k ⊗ A A ) ⊗ A ′ k ≃ k ⊗ A ( A ⊗ A ′ k ) k\otimes_{A'}k\simeq(k\otimes_AA)\otimes_{A'}k\simeq k\otimes_A(A\otimes_{A'}k) k ⊗ A ′ k ≃ ( k ⊗ A A ) ⊗ A ′ k ≃ k ⊗ A ( A ⊗ A ′ k ) . There exists a (co)fiber sequence
k ⊗ A ′ k [ n + 1 ] → k = A ′ ⊗ A ′ k → A ⊗ A ′ k . k\otimes_{A'}k[n+1]\to k=A'\otimes_{A'}k\to A\otimes_{A'}k. k ⊗ A ′ k [ n + 1 ] → k = A ′ ⊗ A ′ k → A ⊗ A ′ k .
We note that H i ( k ⊗ A ′ k [ n + 1 ] ) = 0 H^i(k\otimes_{A'}k[n+1])=0 H i ( k ⊗ A ′ k [ n + 1 ]) = 0 for i ≥ − n i\geq-n i ≥ − n . Taking into account the long exact sequence of the (co)fiber sequence, we conclude that H i ( k ) → H i ( A ⊗ A ′ k ) H^i(k)\to H^i(A\otimes_{A'}k) H i ( k ) → H i ( A ⊗ A ′ k ) is an isomorphism for any − n ≤ i ≤ 0 -n\leq i\leq0 − n ≤ i ≤ 0 . This means that the amplitude of H i ( A ⊗ A ′ k ) H^i(A\otimes_{A'}k) H i ( A ⊗ A ′ k ) is concentrated in cohomological degrees i < − n − 1 i<-n-1 i < − n − 1 except H 0 ( A ⊗ A ′ k ) = k H^0(A\otimes_{A'}k)=k H 0 ( A ⊗ A ′ k ) = k . The fiber of the canonical morphism A ⊗ A ′ k → k A\otimes_{A'}k\to k A ⊗ A ′ k → k of A A A -modules is concentrated in the cohomological degree i < − n − 1 i<-n-1 i < − n − 1 . Thus, the fiber of k ⊗ A ′ k → k ⊗ A k k\otimes_{A'}k\to k\otimes_Ak k ⊗ A ′ k → k ⊗ A k is concentrated in cohomological degrees i < − n − 1 i<-n-1 i < − n − 1 . The induced morphism H i ( k ⊗ A ′ k ) → H i ( k ⊗ A k ) H^i(k\otimes_{A'}k)\to H^i(k\otimes_Ak) H i ( k ⊗ A ′ k ) → H i ( k ⊗ A k ) is an isomorphism of vector spaces for any − n ≤ i ≤ 0 -n\leq i\leq0 − n ≤ i ≤ 0 . It remains to show that H i ( k ⊗ A ′ k ) H^i(k\otimes_{A'}k) H i ( k ⊗ A ′ k ) and H i ( k ⊗ A k ) H^i(k\otimes_Ak) H i ( k ⊗ A k ) are finite dimensional. This is a consequence of the observation that for an augmented connective locally finite E 1 \mathbb{E}_1 E 1 -algebra R → k R\to k R → k , k ⊗ R k k\otimes_Rk k ⊗ R k is also connective and locally finite (a direct computation or the Tor spectral sequence). Thus, our assertion follows. □
Corollary 5.5. Let f : A ′ → A f:A'\to A f : A ′ → A be a morphism of Artin E 1 \mathbb{E}_1 E 1 -algebras. Let n n n be a nonnegative integer. Assume that f f f exhibits A A A as τ ≥ − n ( A ′ ) \tau^{\geq-n}(A') τ ≥ − n ( A ′ ) . Suppose that H i ( A ′ ) = 0 H^i(A')=0 H i ( A ′ ) = 0 for any i < − n − 1 i<-n-1 i < − n − 1 . Then the induced morphism H j ( D l ( A ) ) → H j ( D l ( A ′ ) ) H^j(\mathbb{D}_l(A))\to H^j(\mathbb{D}_l(A')) H j ( D l ( A )) → H j ( D l ( A ′ )) is an isomorphism of finite dimensional vector spaces for any 0 ≤ j ≤ n 0\leq j\leq n 0 ≤ j ≤ n .
Proof. For an augmented algebra A → k A\to k A → k , we let Bar ( A ) \operatorname{Bar}(A) Bar ( A ) denote the bar construction k ⊗ A k k\otimes_Ak k ⊗ A k . This construction induces a functor Bar : Alg l ( Mod k ) / k → Alg l − 1 ( Mod k ) / k \operatorname{Bar}:\operatorname{Alg}_l(\operatorname{Mod}_k)_{/k}\to\operatorname{Alg}_{l-1}(\operatorname{Mod}_k)_{/k} Bar : Alg l ( Mod k ) / k → Alg l − 1 ( Mod k ) / k . The l l l -times iteration yields Bar l : Alg l ( Mod k ) / k → Mod k \operatorname{Bar}^l:\operatorname{Alg}_l(\operatorname{Mod}_k)_{/k}\to\operatorname{Mod}_k Bar l : Alg l ( Mod k ) / k → Mod k . Then D l ( A ) \mathbb{D}_l(A) D l ( A ) can be identified with the k k k -linear dual of Bar l ( A ) \operatorname{Bar}^l(A) Bar l ( A ) in Mod k \operatorname{Mod}_k Mod k (see e.g. Corollary 4.15). Thus, it is enough to show that H j ( Bar l ( A ′ ) ) → H j ( Bar l ( A ) ) H^j(\operatorname{Bar}^l(A'))\to H^j(\operatorname{Bar}^l(A)) H j ( Bar l ( A ′ )) → H j ( Bar l ( A )) is an isomorphism of finite dimensional vector spaces for any − n ≤ j ≤ 0 -n\leq j\leq0 − n ≤ j ≤ 0 . This follows from Lemma 5.4. □
Proof of Proposition 5.3. Let R R R be either an almost Artin E 1 \mathbb{E}_1 E 1 -algebra or an almost Artin E 2 \mathbb{E}_2 E 2 -algebra. When R R R is an almost Artin E 2 \mathbb{E}_2 E 2 -algebra, the argument in the E 1 \mathbb{E}_1 E 1 -case works, so we focus on the case when R R R is an almost Artin E 1 \mathbb{E}_1 E 1 -algebra. We write R n R_n R n for τ ≥ − n R \tau^{\geq-n}R τ ≥ − n R for n ≥ 0 n\geq0 n ≥ 0 . We regard R R R as the sequential limit lim n ≥ 0 R n \lim_{n\geq0} R_n lim n ≥ 0 R n of Artin E 1 \mathbf{E}_1 E 1 -algebras. Then the Koszul duality functor carries small colimits to limits so that
D 1 ( c o l i m n ≥ 0 D 1 ( R n ) ) ≃ lim n ≥ 0 D 1 ( D 1 ( R n ) ) ≃ lim n ≥ 0 R n ≃ R \mathbb{D}_1\left(\mathop{\mathrm{colim}}_{n\geq0}\mathbb{D}_1(R_n)\right)\simeq\lim_{n\geq0}\mathbb{D}_1(\mathbb{D}_1(R_n))\simeq\lim_{n\geq0}R_n\simeq R D 1 ( colim n ≥ 0 D 1 ( R n ) ) ≃ n ≥ 0 lim D 1 ( D 1 ( R n )) ≃ n ≥ 0 lim R n ≃ R
where the right equivalence is the biduality equivalence for Artin E 1 \mathbf{E}_1 E 1 -algebras (cf. [25 X, Section 3] ). From Corollary 5.5 and the fact that D 1 ( R n ) \mathbb{D}_1(R_n) D 1 ( R n ) is coconnective and locally finite (cf. [25 X Lemma 4.5.9] ), we see that c o l i m n ≥ 0 D 1 ( R n ) \mathop{\mathrm{colim}}_{n\geq0}\mathbb{D}_1(R_n) colim n ≥ 0 D 1 ( R n ) is coconnective and locally finite. Also, D 1 ( R ) \mathbb{D}_1(R) D 1 ( R ) is coconnective and locally finite (as it is the linear dual of k ⊗ R k k\otimes_R k k ⊗ R k ). Let Alg + ( Mod k ) c c l f \operatorname{Alg}^{+}(\operatorname{Mod}_k)^{\mathrm{cclf}} Alg + ( Mod k ) cclf be the full subcategory of Alg + ( Mod k ) \operatorname{Alg}^{+}(\operatorname{Mod}_k) Alg + ( Mod k ) spanned by coconnective and locally finite algebras. According to [25 X, Corollary 3.1.15] , (the restriction of) the Koszul duality functor D 1 : Alg + ( Mod k ) c c l f → Alg + ( Mod k ) o p \mathbb{D}_1:\operatorname{Alg}^{+}(\operatorname{Mod}_k)^{\mathrm{cclf}}\to\operatorname{Alg}^{+}(\operatorname{Mod}_k)^{\mathrm{op}} D 1 : Alg + ( Mod k ) cclf → Alg + ( Mod k ) op is fully faithful. Since D 1 ( c o l i m n ≥ 0 D 1 ( R n ) ) ≃ R \mathbb{D}_1(\mathop{\mathrm{colim}}_{n\geq0}\mathbb{D}_1(R_n))\simeq R D 1 ( colim n ≥ 0 D 1 ( R n )) ≃ R , it follows that the canonical morphism c o l i m n ≥ 0 D 1 ( R n ) → D 1 ( R ) \mathop{\mathrm{colim}}_{n\geq0}\mathbb{D}_1(R_n)\to\mathbb{D}_1(R) colim n ≥ 0 D 1 ( R n ) → D 1 ( R ) is an equivalence. Consequently, we have R ≃ D 1 ( D 1 ( R ) ) R\simeq\mathbb{D}_1(\mathbb{D}_1(R)) R ≃ D 1 ( D 1 ( R )) .
From the above proof, we see:
Corollary 5.6. Let R R R be an almost Artin E 2 \mathbf{E}_2 E 2 -algebra. We write R n R_n R n for τ ≥ − n R \tau^{\geq-n}R τ ≥ − n R for n ≥ 0 n\geq0 n ≥ 0 . Then the canonical morphism c o l i m n ≥ 0 D 2 ( R n ) → D 2 ( R ) \mathop{\mathrm{colim}}_{n\geq0}\mathbb{D}_2(R_n)\to\mathbb{D}_2(R) colim n ≥ 0 D 2 ( R n ) → D 2 ( R ) is an equivalence.
Similarly, we let R R R be an almost Artin E 1 \mathbf{E}_1 E 1 -algebra. We write R n R_n R n for τ ≥ − n R \tau^{\geq-n}R τ ≥ − n R for n ≥ 0 n\geq0 n ≥ 0 . Then the canonical morphism c o l i m n ≥ 0 D 1 ( R n ) → D 1 ( R ) \mathop{\mathrm{colim}}_{n\geq0}\mathbb{D}_1(R_n)\to\mathbb{D}_1(R) colim n ≥ 0 D 1 ( R n ) → D 1 ( R ) is an equivalence.
For a morphism u : A → B u:A\to B u : A → B of almost Artin E 1 \mathbf{E}_1 E 1 -algebras, the pushforward functor along u u u is defined to be the k k k -linear (colimit-preserving) functor Ind ( LCoh ( B ) ) → Ind ( LCoh ( A ) ) \operatorname{Ind}(\operatorname{LCoh}(B))\to\operatorname{Ind}(\operatorname{LCoh}(A)) Ind ( LCoh ( B )) → Ind ( LCoh ( A )) that extends the restriction functor LCoh ( B ) → LCoh ( A ) \operatorname{LCoh}(B)\to\operatorname{LCoh}(A) LCoh ( B ) → LCoh ( A ) . See Example 3.4 for LCoh ( A ) \operatorname{LCoh}(A) LCoh ( A ) . The !-pullback functor along u u u is defined as the right adjoint Ind ( LCoh ( A ) ) → Ind ( LCoh ( B ) ) \operatorname{Ind}(\operatorname{LCoh}(A))\to\operatorname{Ind}(\operatorname{LCoh}(B)) Ind ( LCoh ( A )) → Ind ( LCoh ( B )) of the pushforward functor (see [21 Appendix] for details). For an almost Artin E 1 \mathbf{E}_1 E 1 -algebra A A A , the dualizing object ω A ∈ Ind ( LCoh ( A ) ) \omega_A\in\operatorname{Ind}(\operatorname{LCoh}(A)) ω A ∈ Ind ( LCoh ( A )) of A A A is defined to be the !-pullback of k ∈ Ind ( Coh ( k ) ) = Mod k k\in\operatorname{Ind}(\operatorname{Coh}(k))=\operatorname{Mod}_k k ∈ Ind ( Coh ( k )) = Mod k along k → A k\to A k → A .
Suppose that R R R is an almost Artin E 2 \mathbf{E}_2 E 2 -algebra. We write R n R_n R n for τ ≥ − n R \tau^{\geq-n}R τ ≥ − n R for n ≥ 0 n\geq0 n ≥ 0 . Recall Ind ( LCoh ( R ) ) \operatorname{Ind}(\operatorname{LCoh}(R)) Ind ( LCoh ( R )) and Ind ( LCoh ( R n ) ) \operatorname{Ind}(\operatorname{LCoh}(R_n)) Ind ( LCoh ( R n )) from Example 3.4. There is a fully faithful monoidal functor
Υ R n : LMod R n → Ind ( LCoh ( R n ) ) , \Upsilon_{R_n}:\operatorname{LMod}_{R_n}\to\operatorname{Ind}(\operatorname{LCoh}(R_n)), Υ R n : LMod R n → Ind ( LCoh ( R n )) ,
see [9 Corollary 9.6.3] , [21 ] . In particular, Υ R n \Upsilon_{R_n} Υ R n sends R n R_n R n to the dualizing object ω R n ∈ LCoh ( R n ) \omega_{R_n}\in\operatorname{LCoh}(R_n) ω R n ∈ LCoh ( R n ) . The functors Υ R n \Upsilon_{R_n} Υ R n and Υ R n + 1 \Upsilon_{R_{n+1}} Υ R n + 1 are compatible with the (usual derived) pullback functor LMod R n + 1 → LMod R n \operatorname{LMod}_{R_{n+1}}\to\operatorname{LMod}_{R_n} LMod R n + 1 → LMod R n and the !-pullback functor Ind ( LCoh ( R n + 1 ) ) → Ind ( LCoh ( R n ) ) \operatorname{Ind}(\operatorname{LCoh}(R_{n+1}))\to\operatorname{Ind}(\operatorname{LCoh}(R_n)) Ind ( LCoh ( R n + 1 )) → Ind ( LCoh ( R n )) .
Passing to the limits, we have a monoidal fully faithful functor
Υ ^ R = lim n Υ R n : lim n LMod R n → lim n Ind ( LCoh ( R n ) ) . \widehat{\Upsilon}_R=\lim_n\Upsilon_{R_n}:\lim_n\operatorname{LMod}_{R_n}\to\lim_n\operatorname{Ind}(\operatorname{LCoh}(R_n)). Υ R = n lim Υ R n : n lim LMod R n → n lim Ind ( LCoh ( R n )) .
Lemma 5.7. The !-pullback functors Ind ( LCoh ( R ) ) → Ind ( LCoh ( R n ) ) \operatorname{Ind}(\operatorname{LCoh}(R))\to\operatorname{Ind}(\operatorname{LCoh}(R_n)) Ind ( LCoh ( R )) → Ind ( LCoh ( R n )) induce an equivalence
Ind ( LCoh ( R ) ) ≃ lim n Ind ( LCoh ( R n ) ) . \operatorname{Ind}(\operatorname{LCoh}(R))\simeq\lim_n\operatorname{Ind}(\operatorname{LCoh}(R_n)). Ind ( LCoh ( R )) ≃ n lim Ind ( LCoh ( R n )) . Proof. When R R R arises from a commutative algebra object, the assertion follows from a more general result [11 Vol. I, Chapter 4, Proposition 6.4.3] . When R R R is a connective E 2 \mathbf{E}_2 E 2 -algebra, there exists an equivalence Ind ( LCoh ( R n ) ) ≃ RMod D 1 ( R n ) \operatorname{Ind}(\operatorname{LCoh}(R_n))\simeq\operatorname{RMod}_{\mathbb{D}_1(R_n)} Ind ( LCoh ( R n )) ≃ RMod D 1 ( R n ) (cf. [21 Lemma 2.6 (1)] , [20 proof of Lemma 3.11] , [25 X, Theorem 3.5.1] ). This also gives an equivalence LCoh ( R n ) ≃ RPerf D 1 ( R n ) \operatorname{LCoh}(R_n)\simeq\operatorname{RPerf}_{\mathbb{D}_1(R_n)} LCoh ( R n ) ≃ RPerf D 1 ( R n ) of full subcategories where RPerf D 1 ( R n ) \operatorname{RPerf}_{\mathbb{D}_1(R_n)} RPerf D 1 ( R n ) denotes the full subcategory of compact objects. We regard R R R as the sequential limit lim n ≥ 0 R n \lim_{n\geq0}R_n lim n ≥ 0 R n of Artin E 1 \mathbf{E}_1 E 1 -algebras. By [21 Lemma 2.6 (2)] , the diagram
Ind ( LCoh ( R n + 1 ) ) → RMod D 1 ( R n + 1 ) ↓ ↓ Ind ( LCoh ( R n ) ) → RMod D 1 ( R n ) \begin{CD}
\operatorname{Ind}(\operatorname{LCoh}(R_{n+1})) @>>> \operatorname{RMod}_{\mathbb{D}_1(R_{n+1})}\\
@VVV @VVV\\
\operatorname{Ind}(\operatorname{LCoh}(R_n)) @>>> \operatorname{RMod}_{\mathbb{D}_1(R_n)}
\end{CD} Ind ( LCoh ( R n + 1 )) ↓ ⏐ Ind ( LCoh ( R n )) RMod D 1 ( R n + 1 ) ↓ ⏐ RMod D 1 ( R n ) commutes, where the vertical functor on the left-hand side is the !-pullback functor, and the vertical functor on the right-hand side is the restriction functor along D 1 ( R n ) → D 1 ( R n + 1 ) \mathbb{D}_1(R_n)\to\mathbb{D}_1(R_{n+1}) D 1 ( R n ) → D 1 ( R n + 1 ) . To prove the assertion, it will suffice to show that the functor RMod D 1 ( R ) → lim n RMod D 1 ( R n ) \operatorname{RMod}_{\mathbb{D}_1(R)}\to\lim_n\operatorname{RMod}_{\mathbb{D}_1(R_n)} RMod D 1 ( R ) → lim n RMod D 1 ( R n ) induced by the restriction functors along D 1 ( R n ) → D 1 ( R ) \mathbb{D}_1(R_n)\to\mathbb{D}_1(R) D 1 ( R n ) → D 1 ( R ) is an equivalence. The functor Θ : Alg 1 ( Mod k ) → Pr \Theta:\operatorname{Alg}_1(\operatorname{Mod}_k)\to\operatorname{Pr} Θ : Alg 1 ( Mod k ) → Pr sending A A A to RMod A \operatorname{RMod}_A RMod A preserves filtered colimits (note that Θ Mod k : Alg 1 ( Mod k ) → ( Pr k ) Mod k / \Theta_{\operatorname{Mod}_k}:\operatorname{Alg}_1(\operatorname{Mod}_k)\to(\operatorname{Pr}_k)_{\operatorname{Mod}_k/} Θ Mod k : Alg 1 ( Mod k ) → ( Pr k ) Mod k / is a left adjoint (see Section 3.4 for Θ Mod k \Theta_{\operatorname{Mod}_k} Θ Mod k ), and the forgetful functor ( Pr k ) Mod k / → Pr (\operatorname{Pr}_k)_{\operatorname{Mod}_k/}\to\operatorname{Pr} ( Pr k ) Mod k / → Pr preserves sifted colimits). By Corollary 5.6,
we have a canonical equivalence RMod D 1 ( R ) ≃ colim n RMod D 1 ( R n ) \operatorname{RMod}_{\mathbb{D}_1}(R) \simeq\operatorname*{colim}_{n} \operatorname{RMod}_{\mathbb{D}_1}(R_n) RMod D 1 ( R ) ≃ colim n RMod D 1 ( R n ) in P r \mathrm{Pr} Pr such that colim n RMod D 1 ( R n ) \operatorname*{colim}_{n} \operatorname{RMod}_{\mathbb{D}_1}(R_n) colim n RMod D 1 ( R n ) is a colimit in P r \mathrm{Pr} Pr of the diagram RMod D 1 ( R 0 ) → RMod D 1 ( R 1 ) → ⋯ \operatorname{RMod}_{\mathbb{D}_1}(R_0) \to\operatorname{RMod}_{\mathbb{D}_1}(R_1) \to\cdots RMod D 1 ( R 0 ) → RMod D 1 ( R 1 ) → ⋯ induced by the base change functors. Passing to right adjoint functors we identify colim n RMod D 1 ( R n ) \operatorname*{colim}_{n} \operatorname{RMod}_{\mathbb{D}_1}(R_n) colim n RMod D 1 ( R n ) with lim n RMod D 1 ( R n ) \lim_{n} \operatorname{RMod}_{\mathbb{D}_1}(R_n) lim n RMod D 1 ( R n ) (i.e., the limit of ∞ \infty ∞ -categories). Thus, we conclude that RMod D 1 ( R ) → lim n RMod D 1 ( R n ) \operatorname{RMod}_{\mathbb{D}_1}(R) \to\lim_{n} \operatorname{RMod}_{\mathbb{D}_1}(R_n) RMod D 1 ( R ) → lim n RMod D 1 ( R n ) is an equivalence. □ \square □
By Lemma 5.7, we have
Corollary 5.8. The diagram
LMod R → Υ R Ind ( LCoh ( R ) ) ↓ ↓ ≃ lim n LMod R n → Υ ^ R lim n Ind ( LCoh ( R n + 1 ) ) \begin{CD}
\operatorname{LMod}_{R} @>{\Upsilon_R}>> \operatorname{Ind}(\operatorname{LCoh}(R)) \\
@VVV @VV{\simeq}V \\
\lim_n \operatorname{LMod}_{R_n} @>{\widehat{\Upsilon}_R}>> \lim_n \operatorname{Ind}(\operatorname{LCoh}(R_{n+1}))
\end{CD} LMod R ↓ ⏐ n lim LMod R n Υ R Υ R Ind ( LCoh ( R )) ↓ ⏐ ≃ n lim Ind ( LCoh ( R n + 1 )) commutes up to canonical homotopy. In particular, Υ R : LMod R → Ind ( LCoh ( R ) ) \Upsilon_R : \operatorname{LMod}_{R} \to\operatorname{Ind}(\operatorname{LCoh}(R)) Υ R : LMod R → Ind ( LCoh ( R )) can be identified with the canonical functor LMod R → lim n LMod R n \operatorname{LMod}_{R} \to\lim_n \operatorname{LMod}_{R_n} LMod R → lim n LMod R n induced by base changes through the embedding lim n LMod R n ⊂ lim n Ind ( LCoh ( R n + 1 ) ) ≃ Ind ( LCoh ( R ) ) \lim_n \operatorname{LMod}_{R_n} \subset\lim_n \operatorname{Ind}(\operatorname{LCoh}(R_{n+1})) \simeq\operatorname{Ind}(\operatorname{LCoh}(R)) lim n LMod R n ⊂ lim n Ind ( LCoh ( R n + 1 )) ≃ Ind ( LCoh ( R )) .
Corollary 5.9. Let ω R ∈ Ind ( LCoh ( R ) ) \omega_R \in\operatorname{Ind}(\operatorname{LCoh}(R)) ω R ∈ Ind ( LCoh ( R )) be the dualizing object, which is a unit object of the monoidal ∞ \infty ∞ -category Ind ( LCoh ( R ) ) \operatorname{Ind}(\operatorname{LCoh}(R)) Ind ( LCoh ( R )) . The (opposite of the) endomorphism E 2 E_2 E 2 -algebra of ω R \omega_R ω R is R R R . In particular, Υ R \Upsilon_R Υ R is fully faithful if we restrict the domain to the single object R R R .
Proof. By the equivalences in the diagram of Corollary 5.8, the (opposite of the) endomorphism E 2 E_2 E 2 -algebra of ω R \omega_R ω R is the limit of the (opposite of the) endomorphism E 2 E_2 E 2 -algebra of ω R n \omega_{R_n} ω R n , which is canonically equivalent to lim n R n \lim_n R_n lim n R n . Namely, there exist equivalences
Hom Ind ( LCoh ( R ) ) ( ω R , ω R ) ≃ lim n Hom Ind ( LCoh ( R n ) ) ( ω R n , ω R n ) ≃ lim n R n ≃ R \operatorname{Hom}_{\operatorname{Ind}(\operatorname{LCoh}(R))}(\omega_R,\omega_R) \simeq\lim_n \operatorname{Hom}_{\operatorname{Ind}(\operatorname{LCoh}(R_n))}(\omega_{R_n},\omega_{R_n}) \simeq\lim_n R_n \simeq R Hom Ind ( LCoh ( R )) ( ω R , ω R ) ≃ n lim Hom Ind ( LCoh ( R n )) ( ω R n , ω R n ) ≃ n lim R n ≃ R
as desired. Here Hom \operatorname{Hom} Hom indicates the Hom object defined in Mod k \operatorname{Mod}_k Mod k (note that Ind ( LCoh ( R ) ) \operatorname{Ind}(\operatorname{LCoh}(R)) Ind ( LCoh ( R )) is a k k k -linear presentable ∞ \infty ∞ -category). □ \square □
Definition 5.10. We define R ′ R' R ′ in Ind ( LCoh ( R ) ) \operatorname{Ind}(\operatorname{LCoh}(R)) Ind ( LCoh ( R )) as the limit of ⋯ → R n → ⋯ → R 1 → R 0 \cdots\to R_n \to\cdots\to R_1 \to R_0 ⋯ → R n → ⋯ → R 1 → R 0 taken in Ind ( LCoh ( R ) ) \operatorname{Ind}(\operatorname{LCoh}(R)) Ind ( LCoh ( R )) . Here R n ∈ LCoh ( R ) ⊂ Ind ( LCoh ( R ) ) R_n \in\operatorname{LCoh}(R) \subset\operatorname{Ind}(\operatorname{LCoh}(R)) R n ∈ LCoh ( R ) ⊂ Ind ( LCoh ( R )) .
More generally, for M ∈ LMod R M \in\operatorname{LMod}_{R} M ∈ LMod R we define M ′ ∈ Ind ( LCoh ( R ) ) M' \in\operatorname{Ind}(\operatorname{LCoh}(R)) M ′ ∈ Ind ( LCoh ( R )) as the limit of ⋯ → τ ≥ − n M → ⋯ → τ ≥ − 1 M → τ ≥ 0 M \cdots\to\tau^{\geq-n}M \to\cdots\to\tau^{\geq-1}M \to\tau^{\geq0}M ⋯ → τ ≥ − n M → ⋯ → τ ≥ − 1 M → τ ≥ 0 M taken in Ind ( LCoh ( R ) ) \operatorname{Ind}(\operatorname{LCoh}(R)) Ind ( LCoh ( R )) . Here τ ≥ − n M \tau^{\geq-n}M τ ≥ − n M is regarded as an object of Ind ( LCoh ( R ) ) ≥ − n ≃ LMod R ≥ − n \operatorname{Ind}(\operatorname{LCoh}(R))^{\geq-n} \simeq\operatorname{LMod}_{R}^{\geq-n} Ind ( LCoh ( R ) ) ≥ − n ≃ LMod R ≥ − n .
Let ψ R : Ind ( LCoh ( R ) ) → LMod R \psi_R : \operatorname{Ind}(\operatorname{LCoh}(R)) \to\operatorname{LMod}_{R} ψ R : Ind ( LCoh ( R )) → LMod R be the left Kan extension of the inclusion LCoh ( R ) → LMod R \operatorname{LCoh}(R) \to\operatorname{LMod}_{R} LCoh ( R ) → LMod R . We remark that in [9 ] and [21 ] the notation Ψ R \Psi_R Ψ R is used instead of ψ R \psi_R ψ R . We note that Ind ( LCoh ( R ) ) → LMod R \operatorname{Ind}(\operatorname{LCoh}(R)) \to\operatorname{LMod}_{R} Ind ( LCoh ( R )) → LMod R induces an equivalence Ind ( LCoh ( R ) ) ≥ 0 ≃ LMod R ≥ 0 \operatorname{Ind}(\operatorname{LCoh}(R))^{\geq0} \simeq\operatorname{LMod}_{R}^{\geq0} Ind ( LCoh ( R ) ) ≥ 0 ≃ LMod R ≥ 0 (see [9 Proposition 1.2.4] for the commutative case. The proof of the case of E 2 E_2 E 2 -algebras is similar, cf. [20 the proof of Lemma 3.11] ). The functor LMod R → Ind ( LCoh ( R ) ) \operatorname{LMod}_{R} \to\operatorname{Ind}(\operatorname{LCoh}(R)) LMod R → Ind ( LCoh ( R )) given by M ↦ M ′ M \mapsto M' M ↦ M ′ is nothing but the right adjoint of ψ R \psi_R ψ R . The quasi-inverse LMod R ≥ 0 → Ind ( LCoh ( R ) ) ≥ 0 \operatorname{LMod}_{R}^{\geq0} \to\operatorname{Ind}(\operatorname{LCoh}(R))^{\geq0} LMod R ≥ 0 → Ind ( LCoh ( R ) ) ≥ 0 of Ind ( LCoh ( R ) ) ≥ 0 ≃ LMod R ≥ 0 \operatorname{Ind}(\operatorname{LCoh}(R))^{\geq0} \simeq\operatorname{LMod}_{R}^{\geq0} Ind ( LCoh ( R ) ) ≥ 0 ≃ LMod R ≥ 0 is given by the restriction of the right adjoint.
Proposition 5.11. There exists a canonical equivalence Hom Ind ( LCoh ( R ) ) ( R ′ , M ′ ) ≃ M \operatorname{Hom}_{\operatorname{Ind}(\operatorname{LCoh}(R))}(R',M') \simeq M Hom Ind ( LCoh ( R )) ( R ′ , M ′ ) ≃ M in Mod k \operatorname{Mod}_k Mod k . Here Hom \operatorname{Hom} Hom indicates the Hom object defined in Mod k \operatorname{Mod}_k Mod k .
Proof. By the definition of M ′ M' M ′ , there exists a canonical equivalence
Hom Ind ( LCoh ( R ) ) ( R ′ , M ′ ) ≃ lim n Hom Ind ( LCoh ( R ) ) ( R ′ , τ ≥ − n M ) . \operatorname{Hom}_{\operatorname{Ind}(\operatorname{LCoh}(R))}(R',M') \simeq\lim_n \operatorname{Hom}_{\operatorname{Ind}(\operatorname{LCoh}(R))}(R',\tau^{\geq-n}M). Hom Ind ( LCoh ( R )) ( R ′ , M ′ ) ≃ n lim Hom Ind ( LCoh ( R )) ( R ′ , τ ≥ − n M ) . If we denote by τ Ind ≤ ∗ \tau_{\operatorname{Ind}}^{\leq*} τ Ind ≤ ∗ and τ Ind ≥ ∗ \tau_{\operatorname{Ind}}^{\geq*} τ Ind ≥ ∗ the truncations with respect to the t t t -structure on Ind ( LCoh ( R ) ) \operatorname{Ind}(\operatorname{LCoh}(R)) Ind ( LCoh ( R )) , then τ Ind ≥ − n R ′ ≃ τ Ind ≥ − n R \tau_{\operatorname{Ind}}^{\geq-n}R' \simeq\tau_{\operatorname{Ind}}^{\geq-n}R τ Ind ≥ − n R ′ ≃ τ Ind ≥ − n R . Indeed, there exists a cofiber sequence
τ Ind ≤ − n ( R ′ ) → R ′ → τ Ind ≥ − n ( R ′ ) \tau_{\operatorname{Ind}}^{\leq-n}(R') \to R' \to\tau_{\operatorname{Ind}}^{\geq-n}(R') τ Ind ≤ − n ( R ′ ) → R ′ → τ Ind ≥ − n ( R ′ ) in Ind ( LCoh ( R ) ) \operatorname{Ind}(\operatorname{LCoh}(R)) Ind ( LCoh ( R )) . Since τ Ind ≤ − n \tau_{\operatorname{Ind}}^{\leq-n} τ Ind ≤ − n preserves small limits we have a cofiber sequence
lim m ( τ Ind ≤ − n τ ≥ − m R ) → lim m R m → τ Ind ≥ − n ( lim m R m ) . \lim_m(\tau_{\operatorname{Ind}}^{\leq-n}\tau^{\geq-m}R) \to\lim_m R_m \to\tau_{\operatorname{Ind}}^{\geq-n}(\lim_m R_m). m lim ( τ Ind ≤ − n τ ≥ − m R ) → m lim R m → τ Ind ≥ − n ( m lim R m ) .
Since τ I n d ≤ − n τ ≥ − m R ≃ τ ≤ − n τ ≥ − m R ≃ τ ≥ − m τ ≤ − n R \tau_{\mathrm{Ind}}^{\leq-n}\tau^{\geq-m}R \simeq\tau^{\leq-n}\tau^{\geq-m}R \simeq\tau^{\geq-m}\tau^{\leq-n}R τ Ind ≤ − n τ ≥ − m R ≃ τ ≤ − n τ ≥ − m R ≃ τ ≥ − m τ ≤ − n R , it follows that lim m ( τ ≥ − m τ ≤ − n R ) ≃ τ I n d ≤ − n ( R ′ ) \lim_m(\tau^{\geq-m}\tau^{\leq-n}R) \simeq\tau_{\mathrm{Ind}}^{\leq-n}(R') lim m ( τ ≥ − m τ ≤ − n R ) ≃ τ Ind ≤ − n ( R ′ ) . Hence, the above cofiber sequence is obtained from the cofiber sequence
τ ≤ − n R ⟶ R ⟶ τ > − n R \tau^{\leq-n}R \longrightarrow R \longrightarrow\tau^{>-n}R τ ≤ − n R ⟶ R ⟶ τ > − n R
by taking τ ≥ − m \tau^{\geq-m} τ ≥ − m and passing to the limit m → ∞ m \to\infty m → ∞ . It follows that τ I n d > − n ( lim m R m ) ≃ τ > − n R \tau_{\mathrm{Ind}}^{>-n}(\lim_m R_m) \simeq\tau^{>-n}R τ Ind > − n ( lim m R m ) ≃ τ > − n R . Consequently, we have
Hom Ind ( LCoh ( R ) ) ( R ′ , τ ≥ − n M ) ≃ Hom Ind ( LCoh ( R ) ) ( τ I n d ≥ − n R ′ , τ ≥ − n M ) ≃ Hom Ind ( LCoh ( R ) ) ( τ ≥ − n R , τ ≥ − n M ) . \operatorname{Hom}_{\operatorname{Ind}(\operatorname{LCoh}(R))}(R',\tau^{\geq-n}M) \simeq\operatorname{Hom}_{\operatorname{Ind}(\operatorname{LCoh}(R))}(\tau_{\mathrm{Ind}}^{\geq-n}R',\tau^{\geq-n}M) \simeq\operatorname{Hom}_{\operatorname{Ind}(\operatorname{LCoh}(R))}(\tau^{\geq-n}R,\tau^{\geq-n}M). Hom Ind ( LCoh ( R )) ( R ′ , τ ≥ − n M ) ≃ Hom Ind ( LCoh ( R )) ( τ Ind ≥ − n R ′ , τ ≥ − n M ) ≃ Hom Ind ( LCoh ( R )) ( τ ≥ − n R , τ ≥ − n M ) .
Recall the equivalence Ind ( LCoh ( R ) ) ≥ − n ≃ LMod R ≥ − n \operatorname{Ind}(\operatorname{LCoh}(R))^{\geq-n} \simeq\operatorname{LMod}_{R}^{\geq-n} Ind ( LCoh ( R ) ) ≥ − n ≃ LMod R ≥ − n . There exists an equivalence
Hom Ind ( LCoh ( R ) ) ( τ ≥ − n R , τ ≥ − n M ) ≃ Hom LMod R ( τ ≥ − n R , τ ≥ − n M ) ≃ Hom LMod R ( R , τ ≥ − n M ) ≃ τ ≥ − n M \operatorname{Hom}_{\operatorname{Ind}(\operatorname{LCoh}(R))}(\tau^{\geq-n}R,\tau^{\geq-n}M) \simeq\operatorname{Hom}_{\operatorname{LMod}_{R}}(\tau^{\geq-n}R,\tau^{\geq-n}M) \simeq\operatorname{Hom}_{\operatorname{LMod}_{R}}(R,\tau^{\geq-n}M) \simeq\tau^{\geq-n}M Hom Ind ( LCoh ( R )) ( τ ≥ − n R , τ ≥ − n M ) ≃ Hom LMod R ( τ ≥ − n R , τ ≥ − n M ) ≃ Hom LMod R ( R , τ ≥ − n M ) ≃ τ ≥ − n M
in LMod R \operatorname{LMod}_{R} LMod R . Hence we conclude lim n Hom Ind ( LCoh ( R ) ) ( R ′ , τ ≥ − n M ) ≃ lim n τ ≥ − n M ≃ M \lim_n\operatorname{Hom}_{\operatorname{Ind}(\operatorname{LCoh}(R))}(R',\tau^{\geq-n}M) \simeq\lim_n\tau^{\geq-n}M \simeq M lim n Hom Ind ( LCoh ( R )) ( R ′ , τ ≥ − n M ) ≃ lim n τ ≥ − n M ≃ M . □ \square □
Corollary 5.12. The functor ψ R : Ind ( LCoh ( R ) ) → LMod R \psi_R : \operatorname{Ind}(\operatorname{LCoh}(R)) \to\operatorname{LMod}_R ψ R : Ind ( LCoh ( R )) → LMod R sends R ′ R' R ′ to R R R .
Corollary 5.13. The left completion functor ψ R \psi_R ψ R induces an equivalence Hom Ind ( LCoh ( R ) ) ( R ′ , R ′ ) → Hom LMod R ( R , R ) ≃ R \operatorname{Hom}_{\operatorname{Ind}(\operatorname{LCoh}(R))}(R',R') \to\operatorname{Hom}_{\operatorname{LMod}_R}(R,R) \simeq R Hom Ind ( LCoh ( R )) ( R ′ , R ′ ) → Hom LMod R ( R , R ) ≃ R .
Corollary 5.14. Let R R R be an almost Artin E 2 \mathbb{E}_2 E 2 -algebra. Then there exists a canonical equivalence
Ind ( LCoh ( R ) ) ≃ End LMod D 2 ( R ) ( Mod k ) \operatorname{Ind}(\operatorname{LCoh}(R)) \simeq\operatorname{End}_{\operatorname{LMod}_{\mathbb{D}_2(R)}}(\operatorname{Mod}_k) Ind ( LCoh ( R )) ≃ End LMod D 2 ( R ) ( Mod k )
in Alg 1 ( Pr k ) \operatorname{Alg}_1(\operatorname{Pr}_k) Alg 1 ( Pr k ) .
Proof. We note R ≃ lim n R n R \simeq\lim_n R_n R ≃ lim n R n . We consider the Koszul dual End Ind ( LCoh ( R n ) ) ( Mod k ) \operatorname{End}_{\operatorname{Ind}(\operatorname{LCoh}(R_n))}(\operatorname{Mod}_k) End Ind ( LCoh ( R n )) ( Mod k ) of Ind ( LCoh ( R n ) ) → Mod k \operatorname{Ind}(\operatorname{LCoh}(R_n)) \to\operatorname{Mod}_k Ind ( LCoh ( R n )) → Mod k (in Alg 1 + ( Pr k ) \operatorname{Alg}_1^+(\operatorname{Pr}_k) Alg 1 + ( Pr k ) ) and the universal pairing
u : Ind ( LCoh ( R n ) ) ⊗ End Ind ( LCoh ( R n ) ) ( Mod k ) ⟶ Mod k , u : \operatorname{Ind}(\operatorname{LCoh}(R_n)) \otimes\operatorname{End}_{\operatorname{Ind}(\operatorname{LCoh}(R_n))}(\operatorname{Mod}_k) \longrightarrow\operatorname{Mod}_k, u : Ind ( LCoh ( R n )) ⊗ End Ind ( LCoh ( R n )) ( Mod k ) ⟶ Mod k ,
which is a morphism in Alg 1 ( Pr k ) \operatorname{Alg}_1(\operatorname{Pr}_k) Alg 1 ( Pr k ) . This induces Ind ( LCoh ( R n ) ) → End End Ind ( LCoh ( R n ) ) ( Mod k ) ( Mod k ) \operatorname{Ind}(\operatorname{LCoh}(R_n)) \to\operatorname{End}_{\operatorname{End}_{\operatorname{Ind}(\operatorname{LCoh}(R_n))}(\operatorname{Mod}_k)}(\operatorname{Mod}_k) Ind ( LCoh ( R n )) → End End Ind ( LCoh ( R n )) ( Mod k ) ( Mod k ) in Alg 1 ( Pr k ) \operatorname{Alg}_1(\operatorname{Pr}_k) Alg 1 ( Pr k ) . Let LMod D 2 ( R n ) → End Ind ( LCoh ( R n ) ) ( Mod k ) \operatorname{LMod}_{\mathbb{D}_2(R_n)} \to\operatorname{End}_{\operatorname{Ind}(\operatorname{LCoh}(R_n))}(\operatorname{Mod}_k) LMod D 2 ( R n ) → End Ind ( LCoh ( R n )) ( Mod k ) be the algebraization functor (defined as a morphism of Alg 1 ( Pr k ) \operatorname{Alg}_1(\operatorname{Pr}_k) Alg 1 ( Pr k ) ) of End Ind ( LCoh ( R ) ) ( Mod k ) ∈ Alg 1 ( Pr k ) \operatorname{End}_{\operatorname{Ind}(\operatorname{LCoh}(R))}(\operatorname{Mod}_k) \in\operatorname{Alg}_1(\operatorname{Pr}_k) End Ind ( LCoh ( R )) ( Mod k ) ∈ Alg 1 ( Pr k ) (see [21 Lemma 3.3] ). This induces the restriction End End Ind ( LCoh ( R n ) ) ( Mod k ) ( Mod k ) → End LMod D 2 ( R n ) ( Mod k ) \operatorname{End}_{\operatorname{End}_{\operatorname{Ind}(\operatorname{LCoh}(R_n))}(\operatorname{Mod}_k)}(\operatorname{Mod}_k) \to\operatorname{End}_{\operatorname{LMod}_{\mathbb{D}_2(R_n)}}(\operatorname{Mod}_k) End End Ind ( LCoh ( R n )) ( Mod k ) ( Mod k ) → End LMod D 2 ( R n ) ( Mod k ) . We consider the composite functor
Ind ( LCoh ( R n ) ) ⟶ End End Ind ( LCoh ( R n ) ) ( Mod k ) ( Mod k ) ⟶ End LMod D 2 ( R n ) ( Mod k ) \operatorname{Ind}(\operatorname{LCoh}(R_n)) \longrightarrow\operatorname{End}_{\operatorname{End}_{\operatorname{Ind}(\operatorname{LCoh}(R_n))}(\operatorname{Mod}_k)}(\operatorname{Mod}_k) \longrightarrow\operatorname{End}_{\operatorname{LMod}_{\mathbb{D}_2(R_n)}}(\operatorname{Mod}_k) Ind ( LCoh ( R n )) ⟶ End End Ind ( LCoh ( R n )) ( Mod k ) ( Mod k ) ⟶ End LMod D 2 ( R n ) ( Mod k )
in Alg 1 ( Pr k ) \operatorname{Alg}_1(\operatorname{Pr}_k) Alg 1 ( Pr k ) . The sequence Ind ( LCoh ( R n ) ) → End End Ind ( LCoh ( R n ) ) ( Mod k ) ( Mod k ) → End LMod D 2 ( R n ) ( Mod k ) \operatorname{Ind}(\operatorname{LCoh}(R_n)) \to\operatorname{End}_{\operatorname{End}_{\operatorname{Ind}(\operatorname{LCoh}(R_n))}(\operatorname{Mod}_k)}(\operatorname{Mod}_k) \to\operatorname{End}_{\operatorname{LMod}_{\mathbb{D}_2(R_n)}}(\operatorname{Mod}_k) Ind ( LCoh ( R n )) → End End Ind ( LCoh ( R n )) ( Mod k ) ( Mod k ) → End LMod D 2 ( R n ) ( Mod k ) is covariantly functorial with respect to ⋯ → R 2 → R 1 → R 0 \cdots\to R_2 \to R_1 \to R_0 ⋯ → R 2 → R 1 → R 0 (cf. [21 Section 3] ). Thus, we have
lim n Ind ( LCoh ( R n ) ) ⟶ lim n End LMod D 2 ( R n ) ( Mod k ) ≃ lim n RMod k ⊗ D 2 ( R n ) k . \lim_n\operatorname{Ind}(\operatorname{LCoh}(R_n)) \longrightarrow\lim_n\operatorname{End}_{\operatorname{LMod}_{\mathbb{D}_2(R_n)}}(\operatorname{Mod}_k) \simeq\lim_n\operatorname{RMod}_{k\otimes_{\mathbb{D}_2(R_n)}k}. n lim Ind ( LCoh ( R n )) ⟶ n lim End LMod D 2 ( R n ) ( Mod k ) ≃ n lim RMod k ⊗ D 2 ( R n ) k .
The arrow on the left-hand side is a morphism in Alg 1 ( Pr k ) \operatorname{Alg}_1(\operatorname{Pr}_k) Alg 1 ( Pr k ) , and the equivalence on the right-hand side means an equivalence in Pr k \operatorname{Pr}_k Pr k . The limit on the left-hand side is the limit of the diagram induced by ! ! ! -pullback functors. The limit on the right-hand side is the limit of the diagram induced by the restriction functors along k ⊗ D 2 ( R 0 ) k → k ⊗ D 2 ( R 1 ) k → k ⊗ D 2 ( R 2 ) k → ⋯ k\otimes_{\mathbb{D}_2(R_0)}k \to k\otimes_{\mathbb{D}_2(R_1)}k \to k\otimes_{\mathbb{D}_2(R_2)}k \to\cdots k ⊗ D 2 ( R 0 ) k → k ⊗ D 2 ( R 1 ) k → k ⊗ D 2 ( R 2 ) k → ⋯ . From the proof of Lemma 5.7, we have
lim n RMod k ⊗ D 2 ( R n ) k ≃ lim n RMod D 1 ( R n ) ≃ RMod D 1 ( R ) . \lim_n\operatorname{RMod}_{k\otimes_{\mathbb{D}_2(R_n)}k} \simeq\lim_n\operatorname{RMod}_{\mathbb{D}_1(R_n)} \simeq\operatorname{RMod}_{\mathbb{D}_1(R)}. n lim RMod k ⊗ D 2 ( R n ) k ≃ n lim RMod D 1 ( R n ) ≃ RMod D 1 ( R ) .
Also, by Proposition 5.1 there exist equivalences End LMod D 2 ( R ) ( Mod k ) ≃ RMod k ⊗ D 2 ( R ) k ≃ RMod D 1 ( R ) \operatorname{End}_{\operatorname{LMod}_{\mathbb{D}_2(R)}}(\operatorname{Mod}_k) \simeq\operatorname{RMod}_{k\otimes_{\mathbb{D}_2(R)}k} \simeq\operatorname{RMod}_{\mathbb{D}_1(R)} End LMod D 2 ( R ) ( Mod k ) ≃ RMod k ⊗ D 2 ( R ) k ≃ RMod D 1 ( R ) in Pr k \operatorname{Pr}_k Pr k . Consequently, the canonical morphism End LMod D 2 ( R ) ( Mod k ) → lim n End LMod D 2 ( R n ) ( Mod k ) \operatorname{End}_{\operatorname{LMod}_{\mathbb{D}_2(R)}}(\operatorname{Mod}_k) \to\lim_n\operatorname{End}_{\operatorname{LMod}_{\mathbb{D}_2(R_n)}}(\operatorname{Mod}_k) End LMod D 2 ( R ) ( Mod k ) → lim n End LMod D 2 ( R n ) ( Mod k ) in Alg 1 ( Pr k ) \operatorname{Alg}_1(\operatorname{Pr}_k) Alg 1 ( Pr k ) is an equivalence. By Lemma 5.7, the canonical functor Ind ( LCoh ( R ) ) → lim n Ind ( LCoh ( R n ) ) \operatorname{Ind}(\operatorname{LCoh}(R)) \to\lim_n\operatorname{Ind}(\operatorname{LCoh}(R_n)) Ind ( LCoh ( R )) → lim n Ind ( LCoh ( R n )) is an equivalence. Thus, it is enough to prove that each Ind ( LCoh ( R n ) ) → End LMod D 2 ( R n ) ( Mod k ) \operatorname{Ind}(\operatorname{LCoh}(R_n)) \to\operatorname{End}_{\operatorname{LMod}_{\mathbb{D}_2(R_n)}}(\operatorname{Mod}_k) Ind ( LCoh ( R n )) → End LMod D 2 ( R n ) ( Mod k ) is an equivalence. This is proved in [21 Theorem 1.2] . □ \square □
The proof of the first assertion of Proposition 5.2 in [20 ] also works in the case when R R R is almost Artin. There exist equivalences RMod k ⊗ D 2 ( R ) k ≃ RMod D 1 ( R ) ≃ Ind ( LCoh ( R ) ) \operatorname{RMod}_{k\otimes_{\mathbb{D}_2(R)}k} \simeq\operatorname{RMod}_{\mathbb{D}_1(R)} \simeq\operatorname{Ind}(\operatorname{LCoh}(R)) RMod k ⊗ D 2 ( R ) k ≃ RMod D 1 ( R ) ≃ Ind ( LCoh ( R )) . For the first equivalence, we use Proposition 5.1 . Through these equivalences, the functor η : Mod k ⊗ RMod D 2 ( R ) Mod k → LMod R \eta: \operatorname{Mod}_k \otimes_{\operatorname{RMod}_{\mathbb{D}_2(R)}} \operatorname{Mod}_k \to\operatorname{LMod}_R η : Mod k ⊗ RMod D 2 ( R ) Mod k → LMod R can be identified with ψ R : Ind ( LCoh ( R ) ) → LMod R \psi_R : \operatorname{Ind}(\operatorname{LCoh}(R)) \to\operatorname{LMod}_R ψ R : Ind ( LCoh ( R )) → LMod R , which is a left completion. We also note that the left completion ψ R \psi_R ψ R can be identified with the composite
Ind ( LCoh ( R ) ) ⟶ lim n Ind ( LCoh ( R ) ) ≥ − n ≃ lim n LMod R ≥ − n ≃ LMod R . (2) \operatorname{Ind}(\operatorname{LCoh}(R)) \longrightarrow\lim_n\operatorname{Ind}(\operatorname{LCoh}(R))^{\geq-n} \simeq\lim_n\operatorname{LMod}_{R}^{\geq-n} \simeq\operatorname{LMod}_R. \tag*{(2)} Ind ( LCoh ( R )) ⟶ n lim Ind ( LCoh ( R ) ) ≥ − n ≃ n lim LMod R ≥ − n ≃ LMod R . (2)
The first functor is induced by the truncation functors τ I n d ≥ − n : Ind ( Loh ( R ) ) → Ind ( Loh ( R ) ) ≥ − n \tau_{\mathrm{Ind}}^{\geq-n} : \operatorname{Ind}(\operatorname{Loh}(R)) \to\operatorname{Ind}(\operatorname{Loh}(R))^{\geq-n} τ Ind ≥ − n : Ind ( Loh ( R )) → Ind ( Loh ( R ) ) ≥ − n . The middle equivalence is induced by equivalences Ind ( Loh ( R ) ) ≥ − n = Ind ( Loh ( R ) ≥ − n ) ≃ LMod R ≥ − n \operatorname{Ind}(\operatorname{Loh}(R))^{\geq-n} = \operatorname{Ind}(\operatorname{Loh}(R)^{\geq-n}) \simeq\operatorname{LMod}_{R}^{\geq-n} Ind ( Loh ( R ) ) ≥ − n = Ind ( Loh ( R ) ≥ − n ) ≃ LMod R ≥ − n . The right equivalence comes from the truncation functors on LMod R \operatorname{LMod}_{R} LMod R with respect to the left complete t t t -structure (R R R is connective). From the proof of Proposition 5.11 , R ′ R' R ′ maps to the { τ ≥ − n R } ∈ lim n LMod R ≥ − n \{\tau^{\geq-n}R\} \in\lim_n \operatorname{LMod}_{R}^{\geq-n} { τ ≥ − n R } ∈ lim n LMod R ≥ − n where { τ ≥ − n R } \{\tau^{\geq-n}R\} { τ ≥ − n R } informally indicates the evident object of lim n LMod R ≥ − n \lim_n \operatorname{LMod}_{R}^{\geq-n} lim n LMod R ≥ − n such that τ ≥ − n R ∈ LMod R ≥ − n \tau^{\geq-n}R \in\operatorname{LMod}_{R}^{\geq-n} τ ≥ − n R ∈ LMod R ≥ − n .
Proof of Proposition 5.2 . We prove the latter assertion: We will prove that the target LMod R \operatorname{LMod}_{R} LMod R is a free LMod R \operatorname{LMod}_{R} LMod R -module of rank one. Taking left completions, the action functor LMod R × RMod k ⊗ D 2 ( R ) k → RMod k ⊗ D 2 ( R ) k \operatorname{LMod}_{R} \times\operatorname{RMod}_{k \otimes_{\mathbb{D}_{2}(R)} k} \to\operatorname{RMod}_{k \otimes_{\mathbb{D}_{2}(R)} k} LMod R × RMod k ⊗ D 2 ( R ) k → RMod k ⊗ D 2 ( R ) k induces the module action functor LMod R × LMod R → LMod R \operatorname{LMod}_{R} \times\operatorname{LMod}_{R} \to\operatorname{LMod}_{R} LMod R × LMod R → LMod R , where LMod R \operatorname{LMod}_{R} LMod R is identified with a left completion of RMod k ⊗ D 2 ( R ) k \operatorname{RMod}_{k \otimes_{\mathbb{D}_{2}(R)} k} RMod k ⊗ D 2 ( R ) k . By the same argument as in [20 ] , we are reduced to showing that there exists an object E ∈ RMod k ⊗ D 2 ( R ) k E \in\operatorname{RMod}_{k \otimes_{\mathbb{D}_{2}(R)} k} E ∈ RMod k ⊗ D 2 ( R ) k such that the composite
ρ : LMod R × { E } → LMod R × RMod k ⊗ D 2 ( R ) k → RMod k ⊗ D 2 ( R ) k → η LMod R \rho: \operatorname{LMod}_{R} \times\{E\} \to\operatorname{LMod}_{R} \times\operatorname{RMod}_{k \otimes_{\mathbb{D}_{2}(R)} k} \to\operatorname{RMod}_{k \otimes_{\mathbb{D}_{2}(R)} k} \xrightarrow{\eta} \operatorname{LMod}_{R} ρ : LMod R × { E } → LMod R × RMod k ⊗ D 2 ( R ) k → RMod k ⊗ D 2 ( R ) k η LMod R
is an equivalence. We take E E E to be R ′ R' R ′ through the equivalences RMod k ⊗ D 2 ( R ) k ≃ Ind ( Loh ( R ) ) \operatorname{RMod}_{k \otimes_{\mathbb{D}_{2}(R)} k} \simeq\operatorname{Ind}(\operatorname{Loh}(R)) RMod k ⊗ D 2 ( R ) k ≃ Ind ( Loh ( R )) . By Corollary 5.12, ( R , E ) ∈ LMod R × { E } (R, E) \in\operatorname{LMod}_{R} \times\{E\} ( R , E ) ∈ LMod R × { E } maps to R ∈ LMod R R \in\operatorname{LMod}_{R} R ∈ LMod R . By Corollary 5.8 , Corollary 5.9 , Corollary 5.13 and Corollary 5.14, we see that ρ : LMod R → LMod R \rho: \operatorname{LMod}_{R} \to\operatorname{LMod}_{R} ρ : LMod R → LMod R is fully faithful after the restriction to the full subcategory consisting of the single object R R R . Since R R R is a single compact generator of LMod R \operatorname{LMod}_{R} LMod R , and ρ \rho ρ is a colimit-preserving functor, we conclude that ρ : LMod R → LMod R \rho: \operatorname{LMod}_{R} \to\operatorname{LMod}_{R} ρ : LMod R → LMod R is an equivalence of stable ∞ \infty ∞ -categories. ◻
Hochschild cohomology
We study Hochschild cohomology of t t t -structured stable presentable ∞ \infty ∞ -categories.
Let n n n be a nonnegative integer
Let k k k be a connective commutative ring spectrum. Let D \mathcal{D} D be any of an object of ( n + 1 ) Pr k (n+1)\operatorname{Pr}_{k} ( n + 1 ) Pr k or an object of ( n + 1 ) Pr k t + (n+1)\operatorname{Pr}_{k}^{t+} ( n + 1 ) Pr k t + or an object of ( n + 1 ) Pr k t ± (n+1)\operatorname{Pr}_{k}^{t\pm} ( n + 1 ) Pr k t ± . We will define Hochschild cohomology of D \mathcal{D} D .
Definition 6.1. Let P \mathcal{P} P be any of Pr k \operatorname{Pr}_{k} Pr k or Pr k t + \operatorname{Pr}_{k}^{t+} Pr k t + or Pr k t ± \operatorname{Pr}_{k}^{t\pm} Pr k t ± . Let D \mathcal{D} D be an object of ( n + 1 ) P (n+1)\mathcal{P} ( n + 1 ) P . For n ≥ 1 n \geq1 n ≥ 1 , we consider an endomorphism algebra object End ( D ) ∈ Alg 1 ( Mod n P ( Cat ^ † ) ) \operatorname{End}(\mathcal{D}) \in\operatorname{Alg}_{1}(\operatorname{Mod}_{n\mathcal{P}}(\widehat{\operatorname{Cat}}_{\dagger})) End ( D ) ∈ Alg 1 ( Mod n P ( Cat † )) (note that an internal Hom object Hom ( D , D ) \operatorname{Hom}(\mathcal{D},\mathcal{D}) Hom ( D , D ) does not generally exist in ( n + 1 ) P (n+1)\mathcal{P} ( n + 1 ) P but the internal Hom object Hom ( D , D ) \operatorname{Hom}(\mathcal{D},\mathcal{D}) Hom ( D , D ) exists in Mod n P ( Cat ^ † ) \operatorname{Mod}_{n\mathcal{P}}(\widehat{\operatorname{Cat}}_{\dagger}) Mod n P ( Cat † ) since Mod n P ( Cat ^ † ) \operatorname{Mod}_{n\mathcal{P}}(\widehat{\operatorname{Cat}}_{\dagger}) Mod n P ( Cat † ) is V \mathbb{V} V -presentable). We assume that End ( D ) \operatorname{End}(\mathcal{D}) End ( D ) has an algebraization functor n P C → End ( D ) n\mathcal{P}_{\mathcal{C}} \to\operatorname{End}(\mathcal{D}) n P C → End ( D ) (see Definition 4.3 ). In this situation, we say that D \mathcal{D} D has Hochschild cohomology, and we refer to C ∈ Alg n + 2 ( Mod k ) \mathcal{C} \in\operatorname{Alg}_{n+2}(\operatorname{Mod}_{k}) C ∈ Alg n + 2 ( Mod k ) as Hochschild cohomology of D \mathcal{D} D .
Suppose that n = 0 n=0 n = 0 . We consider the internal endomorphism algebra object End ( D ) ∈ Alg 1 ( Pr k ) \operatorname{End}(\mathcal{D}) \in\operatorname{Alg}_{1}(\operatorname{Pr}_{k}) End ( D ) ∈ Alg 1 ( Pr k ) when P = Pr k \mathcal{P}=\operatorname{Pr}_{k} P = Pr k . When P \mathcal{P} P is either Pr k t + \operatorname{Pr}_{k}^{t+} Pr k t + or Pr k t ± \operatorname{Pr}_{k}^{t\pm} Pr k t ± , we consider the internal endomorphism algebra object End ( D ) ∈ Alg 1 ( Pr k t ) \operatorname{End}(\mathcal{D}) \in\operatorname{Alg}_{1}(\operatorname{Pr}_{k}^{t}) End ( D ) ∈ Alg 1 ( Pr k t ) by regarding D \mathcal{D} D as an object of Pr k t \operatorname{Pr}_{k}^{t} Pr k t . Let LMod C → End ( D ) \operatorname{LMod}_{\mathcal{C}} \to\operatorname{End}(\mathcal{D}) LMod C → End ( D ) be an algebraization functor, which always exists in this case. We refer to C ∈ Alg n + 2 ( Mod k ) \mathcal{C} \in\operatorname{Alg}_{n+2}(\operatorname{Mod}_{k}) C ∈ Alg n + 2 ( Mod k ) as Hochschild cohomology of D \mathcal{D} D .
We write H H ( n + 1 ) P ( D / k ) \mathcal{HH}_{(n+1)\mathcal{P}}(\mathcal{D}/k) HH ( n + 1 ) P ( D / k ) for C \mathcal{C} C . It is common to refer to H H ( n + 1 ) P ( D / k ) \mathcal{HH}_{(n+1)\mathcal{P}}(\mathcal{D}/k) HH ( n + 1 ) P ( D / k ) as the Hochschild cochains etc., but unless stated otherwise we refer to it as Hochschild cohomology.
Warning 6.1.1. If ( n + 1 ) P (n+1)\mathcal{P} ( n + 1 ) P , in which D \mathcal{D} D is defined, is clear, we often write H H ∙ ( D / k ) \mathcal{HH}^{\bullet}(\mathcal{D}/k) HH ∙ ( D / k ) for H H ( n + 1 ) P ( D / k ) \mathcal{HH}_{(n+1)\mathcal{P}}(\mathcal{D}/k) HH ( n + 1 ) P ( D / k ) .
Example 6.2. We assume that End ( D ) \operatorname{End}(\mathcal{D}) End ( D ) has a good endomorphism algebra of a unit (see Definition 4.7). We consider E n ( End ( D ) ) E^{n}(\operatorname{End}(\mathcal{D})) E n ( End ( D )) , which is an E n + 1 E_{n+1} E n + 1 -algebra in P \mathcal{P} P (see Definition 4.7 for E n E^{n} E n ). In this situation, D \mathcal{D} D has Hochschild cohomology H H ( n + 1 ) P ( D / k ) ∈ Alg n + 2 ( Mod k ) \mathcal{HH}_{(n+1)\mathcal{P}}(\mathcal{D}/k) \in\operatorname{Alg}_{n+2}(\operatorname{Mod}_{k}) HH ( n + 1 ) P ( D / k ) ∈ Alg n + 2 ( Mod k ) , defined as the endomorphism algebra of a unit object of E n ( End ( D ) ) E^{n}(\operatorname{End}(\mathcal{D})) E n ( End ( D )) (see Lemma 4.8 ).
Remark 6.3. We consider the case of P = Pr k \mathcal{P}=\operatorname{Pr}_{k} P = Pr k . Let us attempt to define Hochschild cohomology. Let C \mathcal{C} C be an object of Mod i Pr k ( Cat ^ † ) \operatorname{Mod}_{i\operatorname{Pr}_{k}}(\widehat{\operatorname{Cat}}_{\dagger}) Mod i Pr k ( Cat † ) . Here we assume i ≥ 1 i \geq1 i ≥ 1 . For any two objects M , N M,N M , N of C \mathcal{C} C there exists a Hom object Hom ( M , N ) \operatorname{Hom}(M,N) Hom ( M , N ) defined as an object of Mod ( i − 1 ) Pr k ( Cat ^ † ) \operatorname{Mod}_{(i-1)\operatorname{Pr}_{k}}(\widehat{\operatorname{Cat}}_{\dagger}) Mod ( i − 1 ) Pr k ( Cat † ) (generally not in i Pr k ) s i n c e i\operatorname{Pr}_{k}) since i Pr k ) s in ce \operatorname{Mod}_{(i-1)\operatorname{Pr}_{k}}(\widehat{\operatorname{Cat}}_{\dagger})$ is κ 0 \kappa_{0} κ 0 -compactly generated such that i Pr k i\operatorname{Pr}_{k} i Pr k consists of κ 0 \kappa_{0} κ 0 -compact objects ([30 5.1.7] ). Thus, one obtains E n ( End ( D ) ) ∈ Alg n + 1 ( Mod Mod k ( Cat ^ † ) ) E^{n}(\operatorname{End}(\mathcal{D})) \in\operatorname{Alg}_{n+1}(\operatorname{Mod}_{\operatorname{Mod}_{k}}(\widehat{\operatorname{Cat}}_{\dagger})) E n ( End ( D )) ∈ Alg n + 1 ( Mod Mod k ( Cat † )) . However, for two objects M , N M,N M , N of Mod Mod k ( Cat ^ † ) \operatorname{Mod}_{\operatorname{Mod}_{k}}(\widehat{\operatorname{Cat}}_{\dagger}) Mod Mod k ( Cat † ) , a Hom object Hom ( M , N ) \operatorname{Hom}(M,N) Hom ( M , N ) does not generally exist in Mod k \operatorname{Mod}_{k} Mod k , while Hom ( M , N ) \operatorname{Hom}(M,N) Hom ( M , N ) exists as a k k k -module spectrum in the enlarged universe V \mathcal{V} V . This motivates us to study when Hochschild cohomology H H ∙ ( D / k ) \mathcal{HH}^{\bullet}(\mathcal{D}/k) HH ∙ ( D / k ) exists in Mod k \operatorname{Mod}_k Mod k (or Alg n + 2 ( Mod k ) \operatorname{Alg}_{n+2}(\operatorname{Mod}_k) Alg n + 2 ( Mod k ) ).
As observed in [1 Theorem C] , for n ≥ 3 n \ge3 n ≥ 3 stable monoidal presentable ( ∞ , n ) (\infty,n) ( ∞ , n ) -categories do not generally have a presentable endomorphism algebra of a unit. Thus, we need to consider when D \mathcal{D} D has Hochschild cohomology. Moreover, if D ∈ ( n + 1 ) Pr k t + \mathcal{D} \in(n+1)\operatorname{Pr}^{\mathrm{t}+}_k D ∈ ( n + 1 ) Pr k t + has Hochschild cohomology H H ∙ ( D / k ) \mathcal{HH}^{\bullet}(\mathcal{D}/k) HH ∙ ( D / k ) , one can ask whether H H ∙ ( D / k ) \mathcal{HH}^{\bullet}(\mathcal{D}/k) HH ∙ ( D / k ) coincides with the Hochschild cohomology H H ∙ ( D / k ) \mathcal{HH}^{\bullet}(\mathcal{D}/k) HH ∙ ( D / k ) of the underlying stable ( ∞ , n + 1 ) (\infty,n+1) ( ∞ , n + 1 ) -category D ∈ ( n + 1 ) Pr k \mathcal{D} \in(n+1)\operatorname{Pr}_k D ∈ ( n + 1 ) Pr k , when the latter exists. We remark that when n = 0 n=0 n = 0 Hochschild cohomology does not depend on t t t -structures, so that Hochschild cohomology is determined by the underlying k k k -linear presentable ∞ \infty ∞ -category (see [20 ] ). In the subsequent subsections, we study these issues.
6.2.
We consider the case of t t t -structured (stable) ( ∞ , 2 ) (\infty,2) ( ∞ , 2 ) -categories. The typical example is LMod M ⊗ ( Pr k t + ) \operatorname{LMod}_{\mathcal{M}^{\otimes}}(\operatorname{Pr}^{\mathrm{t}+}_k) LMod M ⊗ ( Pr k t + ) such that M ⊗ \mathcal{M}^{\otimes} M ⊗ is an object of Alg 1 ( Pr k t + ) \operatorname{Alg}_1(\operatorname{Pr}^{\mathrm{t}+}_k) Alg 1 ( Pr k t + ) , that is, a k k k -linear E 1 \mathbb{E}_1 E 1 -monoidal ∞ \infty ∞ -category with an accessible right complete t t t -structure.
Let P \mathcal{P} P be a commutative algebra object in Cat ^ † \widehat{\operatorname{Cat}}_{\dagger} Cat † . We consider a small category enriched over P \mathcal{P} P . There are several approaches to enriched ∞ \infty ∞ -categories, see e.g. [14 ] , [15 ] , [28 ] . We adopt the formalism developed in Hinich [15 ] . We briefly review the formulation of enriched categories.
For a small set X X X , we consider the functor category Fun ( X , P ) \operatorname{Fun}(X,\mathcal{P}) Fun ( X , P ) , which we regard as a P \mathcal{P} P -module. Let End ( Fun ( X , P ) ) \operatorname{End}(\operatorname{Fun}(X,\mathcal{P})) End ( Fun ( X , P )) denote the internal endomorphism algebra of the P \mathcal{P} P -module, which is defined as an object of Alg 1 ( Mod P ( Cat ^ † ) ) \operatorname{Alg}_1(\operatorname{Mod}_{\mathcal{P}}(\widehat{\operatorname{Cat}}_{\dagger})) Alg 1 ( Mod P ( Cat † )) . The underlying P \mathcal{P} P -module of End ( Fun ( X , P ) ) \operatorname{End}(\operatorname{Fun}(X,\mathcal{P})) End ( Fun ( X , P )) is equivalent to Fun ( X × X , P ) \operatorname{Fun}(X \times X,\mathcal{P}) Fun ( X × X , P ) . Indeed, Fun ( X , P ) ∈ Mod P ( Cat ^ † ) \operatorname{Fun}(X,\mathcal{P}) \in\operatorname{Mod}_{\mathcal{P}}(\widehat{\operatorname{Cat}}_{\dagger}) Fun ( X , P ) ∈ Mod P ( Cat † ) is a dualizable P \mathcal{P} P -module (in particular, Fun ( X , P ) ∈ 2 P \operatorname{Fun}(X,\mathcal{P}) \in2\mathcal{P} Fun ( X , P ) ∈ 2 P ) whose dual Fun ( X , P ) ∨ \operatorname{Fun}(X,\mathcal{P})^{\vee} Fun ( X , P ) ∨ is equivalent to Fun ( X , P ) \operatorname{Fun}(X,\mathcal{P}) Fun ( X , P ) . Thus, the Hom object Hom P ( Fun ( X , P ) , Fun ( X , P ) ) \operatorname{Hom}_{\mathcal{P}}(\operatorname{Fun}(X,\mathcal{P}),\operatorname{Fun}(X,\mathcal{P})) Hom P ( Fun ( X , P ) , Fun ( X , P )) exists in 2 P 2\mathcal{P} 2 P , and it is equivalent to Fun ( X , P ) ∨ ⊗ P Fun ( X , P ) ≃ Fun ( X × X , P ) \operatorname{Fun}(X,\mathcal{P})^{\vee} \otimes_{\mathcal{P}} \operatorname{Fun}(X,\mathcal{P}) \simeq\operatorname{Fun}(X \times X,\mathcal{P}) Fun ( X , P ) ∨ ⊗ P Fun ( X , P ) ≃ Fun ( X × X , P ) . Here the subscript P \mathcal{P} P in Hom P ( Fun ( X , P ) , Fun ( X , P ) ) \operatorname{Hom}_{\mathcal{P}}(\operatorname{Fun}(X,\mathcal{P}),\operatorname{Fun}(X,\mathcal{P})) Hom P ( Fun ( X , P ) , Fun ( X , P )) is not the ∞ \infty ∞ -category on which Fun ( X , P ) \operatorname{Fun}(X,\mathcal{P}) Fun ( X , P ) lives (but Fun ( X , P ) \operatorname{Fun}(X,\mathcal{P}) Fun ( X , P ) lies in 2 P 2\mathcal{P} 2 P ): The subscript P \mathcal{P} P means “P \mathcal{P} P -linear”. For A , B ∈ Fun ( X × X , P ) A,B \in\operatorname{Fun}(X \times X,\mathcal{P}) A , B ∈ Fun ( X × X , P ) , the tensor product A ⊗ B A \otimes B A ⊗ B is given informally by A ⊗ B ( x , z ) = colim y ∈ X A ( x , y ) ⊗ B ( y , z ) A \otimes B(x,z) = \operatorname*{colim}_{y \in X} A(x,y) \otimes B(y,z) A ⊗ B ( x , z ) = colim y ∈ X A ( x , y ) ⊗ B ( y , z ) . In this setting, a P \mathcal{P} P -enriched category A \mathcal{A} A having the prescribed set X X X of objects is an algebra object A ∈ Alg 1 ( End ( Fun ( X , P ) ) ) \mathcal{A} \in\operatorname{Alg}_1(\operatorname{End}(\operatorname{Fun}(X,\mathcal{P}))) A ∈ Alg 1 ( End ( Fun ( X , P ))) . We think of the underlying functor A : X × X → P \mathcal{A}: X \times X \to\mathcal{P} A : X × X → P as the assignment ( x , y ) ↦ Hom ( x , y ) ∈ P (x,y) \mapsto\operatorname{Hom}(x,y) \in\mathcal{P} ( x , y ) ↦ Hom ( x , y ) ∈ P , and the associative algebra structure on A \mathcal{A} A encodes the data of composition and units.
We note that Fun ( X , P ) \operatorname{Fun}(X,\mathcal{P}) Fun ( X , P ) has the tautological End ( Fun ( X , P ) ) \operatorname{End}(\operatorname{Fun}(X,\mathcal{P})) End ( Fun ( X , P )) -module structure in the natural way. We define the ∞ \infty ∞ -category of A \mathcal{A} A -module objects in Fun ( X , P ) \operatorname{Fun}(X,\mathcal{P}) Fun ( X , P ) :
P P ( A ) = LMod A ( Fun ( X , P ) ) . \mathcal{P}_{\mathcal{P}}(\mathcal{A}) = \operatorname{LMod}_{\mathcal{A}}(\operatorname{Fun}(X,\mathcal{P})). P P ( A ) = LMod A ( Fun ( X , P )) .
We think of an object F F F of P P ( A ) \mathcal{P}_{\mathcal{P}}(\mathcal{A}) P P ( A ) as a presheaf on the enriched category A \mathcal{A} A taking values in P \mathcal{P} P : The underlying functor F : X → P F:X \to\mathcal{P} F : X → P assigns each object x ∈ X x \in X x ∈ X to F ( x ) ∈ P F(x) \in\mathcal{P} F ( x ) ∈ P . Note that P P ( A ) \mathcal{P}_{\mathcal{P}}(\mathcal{A}) P P ( A ) is a P \mathcal{P} P -module, that is, an object of Mod P ( Cat ^ † ) \operatorname{Mod}_{\mathcal{P}}(\widehat{\operatorname{Cat}}_{\dagger}) Mod P ( Cat † ) .
Example 6.4. Suppose that X X X is a one-point category having a single object ∗ \ast ∗ such that Map X ( ∗ , ∗ ) \operatorname{Map}_{X}(\ast,\ast) Map X ( ∗ , ∗ ) is a contractible space. Then A \mathcal{A} A amounts to an object of Alg 1 ( P ) \operatorname{Alg}_1(\mathcal{P}) Alg 1 ( P ) , and P P ( A ) \mathcal{P}_{\mathcal{P}}(\mathcal{A}) P P ( A ) is LMod A ( P ) \operatorname{LMod}_{\mathcal{A}}(\mathcal{P}) LMod A ( P ) .
Proposition 6.5. P P ( A ) \mathcal{P}_{\mathcal{P}}(\mathcal{A}) P P ( A ) belongs to 2 P 2\mathcal{P} 2 P . Namely, it is κ 0 \kappa_0 κ 0 -compact in Mod P ( Cat ^ † ) \operatorname{Mod}_{\mathcal{P}}(\widehat{\operatorname{Cat}}_{\dagger}) Mod P ( Cat † ) . Moreover, P P ( A ) \mathcal{P}_{\mathcal{P}}(\mathcal{A}) P P ( A ) is dualizable in 2 P 2\mathcal{P} 2 P .
Proposition 6.6. When P \mathcal{P} P is any of Pr k \operatorname{Pr}_k Pr k or Pr k t + \operatorname{Pr}^{\mathrm{t}+}_k Pr k t + or Pr k t ± \operatorname{Pr}^{\mathrm{t}\pm}_k Pr k t ± , P P ( A ) \mathcal{P}_{\mathcal{P}}(\mathcal{A}) P P ( A ) has Hochschild cohomology H H 2 P ∙ ( P P ( A ) / k ) ∈ Alg 3 ( Mod k ) \mathcal{HH}^{\bullet}_{2\mathcal{P}}(\mathcal{P}_{\mathcal{P}}(\mathcal{A})/k) \in\operatorname{Alg}_3(\operatorname{Mod}_k) HH 2 P ∙ ( P P ( A ) / k ) ∈ Alg 3 ( Mod k ) .
For applications in any context depending on P \mathcal{P} P , it is important to understand the construction of Hochschild cohomology. We proceed step by step.
Proof of Proposition 6.5. The tensor product in Mod P ( Cat ^ † ) \operatorname{Mod}_{\mathcal{P}}(\widehat{\operatorname{Cat}}_{\dagger}) Mod P ( Cat † ) preserves colimits separately in each variable. It is enough to show that P P ( A ) \mathcal{P}_{\mathcal{P}}(\mathcal{A}) P P ( A ) is dualizable in Mod P ( Cat ^ † ) \operatorname{Mod}_{\mathcal{P}}(\widehat{\operatorname{Cat}}_{\dagger}) Mod P ( Cat † ) whose dual is RMod A ( Fun ( X , P ) ) \operatorname{RMod}_{\mathcal{A}}(\operatorname{Fun}(X,\mathcal{P})) RMod A ( Fun ( X , P )) . For this purpose, we define an evaluation map and a coevaluation map. For ease of notation, we write P X \mathcal{P}^{X} P X for Fun ( X , P ) \operatorname{Fun}(X,\mathcal{P}) Fun ( X , P ) . According to [24 4.8.4.6] , there exists a canonical equivalence RMod A ( End ( P X ) ) ⊗ End ( P X ) P X ≃ RMod A ( P X ) \operatorname{RMod}_{\mathcal{A}}(\operatorname{End}(\mathcal{P}^{X})) \otimes_{\operatorname{End}(\mathcal{P}^{X})} \mathcal{P}^{X} \simeq\operatorname{RMod}_{\mathcal{A}}(\mathcal{P}^{X}) RMod A ( End ( P X )) ⊗ End ( P X ) P X ≃ RMod A ( P X ) . Moreover, P X \mathcal{P}^{X} P X is dualizable and self-dual in Mod P ( Cat ^ † ) \operatorname{Mod}_{\mathcal{P}}(\widehat{\operatorname{Cat}}_{\dagger}) Mod P ( Cat † ) so that P X ∈ 2 P \mathcal{P}^{X} \in2\mathcal{P} P X ∈ 2 P . Thus we have equivalences
RMod A ( P X ) ⊗ P LMod A ( P X ) ≃ RMod A ( End ( P X ) ) ⊗ End ( P X ) P X ⊗ P P X ⊗ End ( P X ) LMod A ( End ( P X ) ) ≃ RMod A ( End ( P X ) ) ⊗ End ( P X ) Hom ( P X , P X ) ⊗ End ( P X ) LMod A ( End ( P X ) ) ≃ RMod A ( End ( P X ) ) ⊗ End ( P X ) LMod A ( End ( P X ) ) ≃ LMod A op ⊗ A ( End ( P X ) ) . \begin{aligned}
\operatorname{RMod}_{A}(P^{X})\otimes_{P}\operatorname{LMod}_{A}(P^{X})
&\simeq\operatorname{RMod}_{A}(\operatorname{End}(P^{X}))\otimes_{\operatorname{End}(P^{X})}P^{X}\otimes_{P}P^{X}\otimes_{\operatorname{End}(P^{X})}\operatorname{LMod}_{A}(\operatorname{End}(P^{X}))\\
&\simeq\operatorname{RMod}_{A}(\operatorname{End}(P^{X}))\otimes_{\operatorname{End}(P^{X})}\operatorname{Hom}(P^{X},P^{X})\otimes_{\operatorname{End}(P^{X})}\operatorname{LMod}_{A}(\operatorname{End}(P^{X}))\\
&\simeq\operatorname{RMod}_{A}(\operatorname{End}(P^{X}))\otimes_{\operatorname{End}(P^{X})}\operatorname{LMod}_{A}(\operatorname{End}(P^{X}))\\
&\simeq\operatorname{LMod}_{A^{\operatorname{op}}\otimes A}(\operatorname{End}(P^{X})).
\end{aligned} RMod A ( P X ) ⊗ P LMod A ( P X ) ≃ RMod A ( End ( P X )) ⊗ End ( P X ) P X ⊗ P P X ⊗ End ( P X ) LMod A ( End ( P X )) ≃ RMod A ( End ( P X )) ⊗ End ( P X ) Hom ( P X , P X ) ⊗ End ( P X ) LMod A ( End ( P X )) ≃ RMod A ( End ( P X )) ⊗ End ( P X ) LMod A ( End ( P X )) ≃ LMod A op ⊗ A ( End ( P X )) .
Using these equivalences, we define an evaluation map u : RMod A ( P X ) ⊗ P LMod A ( P X ) → P u:\operatorname{RMod}_{A}(P^{X})\otimes_{P}\operatorname{LMod}_{A}(P^{X})\to P u : RMod A ( P X ) ⊗ P LMod A ( P X ) → P as the composite of
LMod A op ⊗ A ( End ( P X ) ) ⟶ End ( P X ) \operatorname{LMod}_{A^{\operatorname{op}}\otimes A}(\operatorname{End}(P^{X}))\longrightarrow\operatorname{End}(P^{X}) LMod A op ⊗ A ( End ( P X )) ⟶ End ( P X )
given by E ↦ A Δ ⊗ A op ⊗ A E \mathcal{E}\mapsto A_{\Delta}\otimes_{A^{\operatorname{op}}\otimes A}\mathcal{E} E ↦ A Δ ⊗ A op ⊗ A E and the canonical evaluation map
End ( P X ) = ( P X ) ∨ ⊗ P P X ⟶ P . \operatorname{End}(P^{X})=(P^{X})^{\vee}\otimes_{P}P^{X}\longrightarrow P. End ( P X ) = ( P X ) ∨ ⊗ P P X ⟶ P .
Here A Δ A_{\Delta} A Δ denotes A A A endowed with the diagonal right A op ⊗ A A^{\operatorname{op}}\otimes A A op ⊗ A -module structure. The coevaluation map v : P → LMod A ( P X ) ⊗ P RMod A ( P X ) v:P\to\operatorname{LMod}_{A}(P^{X})\otimes_{P}\operatorname{RMod}_{A}(P^{X}) v : P → LMod A ( P X ) ⊗ P RMod A ( P X ) is defined to be a unique P P P -linear functor which carries the unit of P P P to the diagonal module A Δ ∈ LMod A op ⊗ A ( End ( P X ) ) ≃ LMod A ( P X ) ⊗ P RMod A ( P X ) A_{\Delta}\in\operatorname{LMod}_{A^{\operatorname{op}}\otimes A}(\operatorname{End}(P^{X}))\simeq\operatorname{LMod}_{A}(P^{X})\otimes_{P}\operatorname{RMod}_{A}(P^{X}) A Δ ∈ LMod A op ⊗ A ( End ( P X )) ≃ LMod A ( P X ) ⊗ P RMod A ( P X ) . It remains to prove that u u u and v v v satisfy the triangle identities. It suffices to show that
LMod A ( P X ) → v ⊗ id LMod A ( P X ) ⊗ P RMod A ( P X ) ⊗ P LMod A ( P X ) \operatorname{LMod}_{A}(P^{X})\xrightarrow{v\otimes\operatorname{id}}\operatorname{LMod}_{A}(P^{X})\otimes_{P}\operatorname{RMod}_{A}(P^{X})\otimes_{P}\operatorname{LMod}_{A}(P^{X}) LMod A ( P X ) v ⊗ id LMod A ( P X ) ⊗ P RMod A ( P X ) ⊗ P LMod A ( P X )
→ id ⊗ u LMod A ( P X ) \xrightarrow{\operatorname{id}\otimes u}\operatorname{LMod}_{A}(P^{X}) id ⊗ u LMod A ( P X )
is the identity functor (another triangle identity is similar). Indeed, unfolding the definitions, the composite is given by P ↦ ( A ⊗ P ) ⊗ ( A op ⊗ A ) A ≃ P P\mapsto(A\otimes P)\otimes_{(A^{\operatorname{op}}\otimes A)}A\simeq P P ↦ ( A ⊗ P ) ⊗ ( A op ⊗ A ) A ≃ P . Namely, the composite is equivalent to the identity. □
For ease of notation, we write P ( A ) = P P ( A ) P(A)=P_{P}(A) P ( A ) = P P ( A ) . By Proposition 6.5, P ( A ) ∈ 2 Pr k t + P(A)\in2\operatorname{Pr}^{\mathrm{t}+}_{k} P ( A ) ∈ 2 Pr k t + , and it is dualizable in 2 P 2P 2 P . Next we will consider End ( P ( A ) ) \operatorname{End}(P(A)) End ( P ( A )) . Note that P ( A ) = LMod A ( Fun ( X , P ) ) P(A)=\operatorname{LMod}_{A}(\operatorname{Fun}(X,P)) P ( A ) = LMod A ( Fun ( X , P )) , and its dual is RMod A ( Fun ( X , P ) ) \operatorname{RMod}_{A}(\operatorname{Fun}(X,P)) RMod A ( Fun ( X , P )) (see the proof of Proposition 6.5). Thus, the internal Hom object Hom P ( P ( A ) , P ( A ) ) \operatorname{Hom}_{P}(P(A),P(A)) Hom P ( P ( A ) , P ( A )) is equivalent to RMod A ( Fun ( X , P ) ) ⊗ P LMod A ( Fun ( X , P ) ) \operatorname{RMod}_{A}(\operatorname{Fun}(X,P))\otimes_{P}\operatorname{LMod}_{A}(\operatorname{Fun}(X,P)) RMod A ( Fun ( X , P )) ⊗ P LMod A ( Fun ( X , P )) . Thus, End ( P ( A ) ) \operatorname{End}(P(A)) End ( P ( A )) is defined as an object of Alg 1 ( 2 P ) \operatorname{Alg}_{1}(2P) Alg 1 ( 2 P ) . Its underlying ∞ \infty ∞ -category is A BMod A ( End ( Fun ( X , P ) ) ) {}_{A}\operatorname{BMod}_{A}(\operatorname{End}(\operatorname{Fun}(X,P))) A BMod A ( End ( Fun ( X , P ))) where the diagonal bimodule A A A is a unit object.
Lemma 6.7. Suppose that P P P has internal Hom objects and small limits. For E , F ∈ A BMod A ( End ( Fun ( X , P ) ) ) \mathcal{E},\mathcal{F}\in{}_{A}\operatorname{BMod}_{A}(\operatorname{End}(\operatorname{Fun}(X,P))) E , F ∈ A BMod A ( End ( Fun ( X , P ))) , there exists a Hom object
Hom A BMod A ( End ( Fun ( X , P ) ) ) ( E , F ) \operatorname{Hom}_{{}_{A}\operatorname{BMod}_{A}(\operatorname{End}(\operatorname{Fun}(X,P)))}(\mathcal{E},\mathcal{F}) Hom A BMod A ( End ( Fun ( X , P ))) ( E , F ) in P P P .
Proof. To observe this, using the comonad of the free-forget adjunction of bimodules
End ( Fun ( X , P ) ) ⇄ A ⊗ ( − ) ⊗ A forget A BMod A ( End ( Fun ( X , P ) ) ) \operatorname{End}(\operatorname{Fun}(X,P))
\underset{\operatorname{forget}}{\overset{A\otimes(-)\otimes A}{\rightleftarrows}}
{}_{A}\operatorname{BMod}_{A}(\operatorname{End}(\operatorname{Fun}(X,P))) End ( Fun ( X , P )) forget ⇄ A ⊗ ( − ) ⊗ A A BMod A ( End ( Fun ( X , P )))
we take a simplicial object R ∙ \mathcal{R}_{\bullet} R ∙ in A BMod A ( End ( Fun ( X , P ) ) ) {}_{A}\operatorname{BMod}_{A}(\operatorname{End}(\operatorname{Fun}(X,P))) A BMod A ( End ( Fun ( X , P ))) with an augmentation ϵ : R ∙ → E \epsilon:\mathcal{R}_{\bullet}\to\mathcal{E} ϵ : R ∙ → E such that each n n n -th term is a free bimodule A ⊗ E ( n ) ⊗ A A\otimes\mathcal{E}(n)\otimes A A ⊗ E ( n ) ⊗ A , and ϵ \epsilon ϵ exhibits E \mathcal{E} E as a colimit (see [24 Proposition 4.7.3.14, 4.7.3.13] ). (This simplicial object is usually called the bar resolution.) Then we have
Hom A BMod A ( End ( Fun ( X , P ) ) ) ( E , F ) ≃ lim [ n ] ∈ Δ Hom Fun ( X × X , P ) ( E ( n ) , F ) . \operatorname{Hom}_{{}_{A}\operatorname{BMod}_{A}(\operatorname{End}(\operatorname{Fun}(X,P)))}(\mathcal{E},\mathcal{F})\simeq\lim_{[n]\in\Delta}\operatorname{Hom}_{\operatorname{Fun}(X\times X,P)}(\mathcal{E}(n),\mathcal{F}). Hom A BMod A ( End ( Fun ( X , P ))) ( E , F ) ≃ [ n ] ∈ Δ lim Hom Fun ( X × X , P ) ( E ( n ) , F ) .
By the assumption, P P P admits internal Hom objects so that Fun ( X × X , P ) \operatorname{Fun}(X\times X,P) Fun ( X × X , P ) admits Hom objects in P P P . Thus each Hom Fun ( X × X , P ) ( E ( n ) , F ) \operatorname{Hom}_{\operatorname{Fun}(X\times X,P)}(\mathcal{E}(n),\mathcal{F}) Hom Fun ( X × X , P ) ( E ( n ) , F ) exists in P P P . Since P P P has small limits by our assumption, lim [ n ] ∈ Δ Hom Fun ( X × X , P ) ( E ( n ) , F ) \lim_{[n]\in\Delta}\operatorname{Hom}_{\operatorname{Fun}(X\times X,P)}(\mathcal{E}(n),\mathcal{F}) lim [ n ] ∈ Δ Hom Fun ( X × X , P ) ( E ( n ) , F ) exists in P P P . □
We now focus on the case when both E \mathcal{E} E and F \mathcal{F} F are the diagonal bimodule A A A .
Corollary 6.8. The following holds.
Suppose that P P P is either Pr k \operatorname{Pr}_{k} Pr k or Pr k t + \operatorname{Pr}^{\mathrm{t}+}_{k} Pr k t + . There exists
Hom A BMod A ( End ( Fun ( X , P ) ) ) ( A , A ) \operatorname{Hom}_{{}_{A}\operatorname{BMod}_{A}(\operatorname{End}(\operatorname{Fun}(X,P)))}(A,A) Hom A BMod A ( End ( Fun ( X , P ))) ( A , A ) in Alg 2 ( P ) \operatorname{Alg}_{2}(P) Alg 2 ( P ) .
(2) Suppose that P = Pr k t ± \mathcal{P}=\operatorname{Pr}_{k}^{\mathrm{t}\pm} P = Pr k t ± . There exists a Hom \operatorname{Hom} Hom object Hom A B M o d A ( End ( Fun ( X , Pr k t ± ) ) ) ( A , A ) \operatorname{Hom}_{\mathcal{A}\mathbin{\mathrm{BMod}}_{\mathcal{A}}(\operatorname{End}(\operatorname{Fun}(X,\operatorname{Pr}_{k}^{\mathrm{t}\pm})))}(\mathcal{A},\mathcal{A}) Hom A BMod A ( End ( Fun ( X , Pr k t ± ))) ( A , A ) in Pr k t + \operatorname{Pr}_{k}^{\mathrm{t}+} Pr k t + , where A B M o d A ( End ( Fun ( X , Pr k t ± ) ) ) \mathcal{A}\mathbin{\mathrm{BMod}}_{\mathcal{A}}(\operatorname{End}(\operatorname{Fun}(X,\operatorname{Pr}_{k}^{\mathrm{t}\pm}))) A BMod A ( End ( Fun ( X , Pr k t ± ))) is regarded as a Pr k t + \operatorname{Pr}_{k}^{\mathrm{t}+} Pr k t + -module.
Proof. Note first that Pr k t \operatorname{Pr}_{k}^{\mathrm{t}} Pr k t and Pr k t + \operatorname{Pr}_{k}^{\mathrm{t}+} Pr k t + admit internal Hom objects (cf. Remark 3.5): For P , Q ∈ Pr k t \mathcal{P},\mathcal{Q}\in\operatorname{Pr}_{k}^{\mathrm{t}} P , Q ∈ Pr k t , an internal Hom object Hom Pr k t ( P , Q ) \operatorname{Hom}_{\operatorname{Pr}_{k}^{\mathrm{t}}}(\mathcal{P},\mathcal{Q}) Hom Pr k t ( P , Q ) from P \mathcal{P} P to Q \mathcal{Q} Q in Pr k t \operatorname{Pr}_{k}^{\mathrm{t}} Pr k t is given by the pair ( Hom Pr k ( P , Q ) , Hom Pr k ( P , Q ) ≤ 0 ) (\operatorname{Hom}_{\operatorname{Pr}_{k}}(\mathcal{P},\mathcal{Q}),\operatorname{Hom}_{\operatorname{Pr}_{k}}(\mathcal{P},\mathcal{Q})^{\leq0}) ( Hom Pr k ( P , Q ) , Hom Pr k ( P , Q ) ≤ 0 ) such that Hom Pr k ( P , Q ) \operatorname{Hom}_{\operatorname{Pr}_{k}}(\mathcal{P},\mathcal{Q}) Hom Pr k ( P , Q ) is the internal Hom object from the underlying object P \mathcal{P} P to Q \mathcal{Q} Q , and Hom Pr k ( P , Q ) ≤ 0 \operatorname{Hom}_{\operatorname{Pr}_{k}}(\mathcal{P},\mathcal{Q})^{\leq0} Hom Pr k ( P , Q ) ≤ 0 is the full subcategory, spanned by right t t t -exact functors. Thus, Pr k t + \operatorname{Pr}_{k}^{\mathrm{t}+} Pr k t + also admits internal Hom objects (by taking right completions of internal Hom objects in Pr k t \operatorname{Pr}_{k}^{\mathrm{t}} Pr k t ). Moreover, Pr k t + \operatorname{Pr}_{k}^{\mathrm{t}+} Pr k t + has small limits (see Proposition 3.12). The ∞ \infty ∞ -category Pr k \operatorname{Pr}_{k} Pr k has small limits (see [23 5.5.3.13] , [24 4.2.3.3] ) and has internal Hom objects (see e.g., [17 Lemma 5.1] ). Thus, (1) follows from Lemma 6.7.
Next, we prove (2). From the proof of Lemma 6.7, we take ϵ : R ∙ → A \epsilon:\mathcal{R}_{\bullet}\to\mathcal{A} ϵ : R ∙ → A such that each n n n -th term is a free bimodule A ⊗ E ( n ) ⊗ A \mathcal{A}\otimes\mathcal{E}(n)\otimes\mathcal{A} A ⊗ E ( n ) ⊗ A , and ϵ \epsilon ϵ exhibits A \mathcal{A} A as a colimit. In this case, E ( n ) ≃ A ⊗ ^ 2 n + 1 \mathcal{E}(n)\simeq\mathcal{A}^{\widehat{\otimes}2n+1} E ( n ) ≃ A ⊗ 2 n + 1 where ⊗ ^ \widehat{\otimes} ⊗ indicates the tensor product in Pr k t ± \operatorname{Pr}_{k}^{\mathrm{t}\pm} Pr k t ± . We consider Hom Fun ( X × X , Pr k t ± ) ( A ⊗ ^ 2 n + 1 , A ) \operatorname{Hom}_{\operatorname{Fun}(X\times X,\operatorname{Pr}_{k}^{\mathrm{t}\pm})}(\mathcal{A}^{\widehat{\otimes}2n+1},\mathcal{A}) Hom Fun ( X × X , Pr k t ± ) ( A ⊗ 2 n + 1 , A ) . This is the Hom object defined as an object of Pr k t + \operatorname{Pr}_{k}^{\mathrm{t}+} Pr k t + (not of Pr k t ± \operatorname{Pr}_{k}^{\mathrm{t}\pm} Pr k t ± ), where we regard Fun ( X × X , Pr k t ± ) \operatorname{Fun}(X\times X,\operatorname{Pr}_{k}^{\mathrm{t}\pm}) Fun ( X × X , Pr k t ± ) as a Pr k t + \operatorname{Pr}_{k}^{\mathrm{t}+} Pr k t + -module. We have the limit lim [ n ] ∈ Δ Hom Fun ( X × X , Pr k t ± ) ( A ⊗ ^ 2 n + 1 , A ) \lim_{[n]\in\Delta}\operatorname{Hom}_{\operatorname{Fun}(X\times X,\operatorname{Pr}_{k}^{\mathrm{t}\pm})}(\mathcal{A}^{\widehat{\otimes}2n+1},\mathcal{A}) lim [ n ] ∈ Δ Hom Fun ( X × X , Pr k t ± ) ( A ⊗ 2 n + 1 , A ) taken in Pr k t + \operatorname{Pr}_{k}^{\mathrm{t}+} Pr k t + , so that the Hom object Hom A B M o d A ( End ( Fun ( X , Pr k t ± ) ) ) ( A , A ) \operatorname{Hom}_{\mathcal{A}\mathbin{\mathrm{BMod}}_{\mathcal{A}}(\operatorname{End}(\operatorname{Fun}(X,\operatorname{Pr}_{k}^{\mathrm{t}\pm})))}(\mathcal{A},\mathcal{A}) Hom A BMod A ( End ( Fun ( X , Pr k t ± ))) ( A , A ) defined in Pr k t + \operatorname{Pr}_{k}^{\mathrm{t}+} Pr k t + is given by lim [ n ] ∈ Δ Hom Fun ( X × X , Pr k t ± ) ( A ⊗ ^ 2 n + 1 , A ) \lim_{[n]\in\Delta}\operatorname{Hom}_{\operatorname{Fun}(X\times X,\operatorname{Pr}_{k}^{\mathrm{t}\pm})}(\mathcal{A}^{\widehat{\otimes}2n+1},\mathcal{A}) lim [ n ] ∈ Δ Hom Fun ( X × X , Pr k t ± ) ( A ⊗ 2 n + 1 , A ) .
We note that Hom A B M o d A ( End ( Fun ( X , P ) ) ) ( A , A ) \operatorname{Hom}_{\mathcal{A}\mathbin{\mathrm{BMod}}_{\mathcal{A}}(\operatorname{End}(\operatorname{Fun}(X,\mathcal{P})))}(\mathcal{A},\mathcal{A}) Hom A BMod A ( End ( Fun ( X , P ))) ( A , A ) is a categorical analogue of the Hochschild complex of A \mathcal{A} A , that is, a Drinfeld center. Indeed, we can compute it as lim [ n ] ∈ Δ Hom Fun ( X × X , P ) ( A ⊗ ^ 2 n + 3 , A ) \lim_{[n]\in\Delta}\operatorname{Hom}_{\operatorname{Fun}(X\times X,\mathcal{P})}(\mathcal{A}^{\widehat{\otimes}2n+3},\mathcal{A}) lim [ n ] ∈ Δ Hom Fun ( X × X , P ) ( A ⊗ 2 n + 3 , A ) , using the above simplicial object R ∙ \mathcal{R}_{\bullet} R ∙ with an augmentation R ∙ → A \mathcal{R}_{\bullet}\to\mathcal{A} R ∙ → A .
Suppose that P = Pr k t + \mathcal{P}=\operatorname{Pr}_{k}^{\mathrm{t}+} P = Pr k t + . We recall the adjoint pair θ 0 Pr k t + : Alg 1 ( Mod k ) ⇄ ( Pr k t + ) Mod k / : E Pr k t + \theta_{0\operatorname{Pr}_{k}^{\mathrm{t}+}}:\operatorname{Alg}_{1}(\operatorname{Mod}_{k})\rightleftarrows(\operatorname{Pr}_{k}^{\mathrm{t}+})_{\operatorname{Mod}_{k}/}:E_{\operatorname{Pr}_{k}^{\mathrm{t}+}} θ 0 Pr k t + : Alg 1 ( Mod k ) ⇄ ( Pr k t + ) Mod k / : E Pr k t + where the symmetric monoidal functor θ 0 Pr k t + \theta_{0\operatorname{Pr}_{k}^{\mathrm{t}+}} θ 0 Pr k t + sends B B B to ( LMod B , B ) (\operatorname{LMod}_{B},B) ( LMod B , B ) , while E + E^{+} E + sends ( A , A ) (\mathcal{A},\mathcal{A}) ( A , A ) to End ( A ) op \operatorname{End}(\mathcal{A})^{\operatorname{op}} End ( A ) op (see Lemma 4.8 and its proof). Applying Alg 2 ( − ) \operatorname{Alg}_{2}(-) Alg 2 ( − ) to the right adjoint, we have Alg 2 ( E Pr k t + + ) : Alg 2 ( Pr k t + ) → Alg 2 ( Alg 1 ( Mod k ) ) ≃ Alg 3 ( Mod k ) \operatorname{Alg}_{2}(E_{\operatorname{Pr}_{k}^{\mathrm{t}+}}^{+}):\operatorname{Alg}_{2}(\operatorname{Pr}_{k}^{\mathrm{t}+})\to\operatorname{Alg}_{2}(\operatorname{Alg}_{1}(\operatorname{Mod}_{k}))\simeq\operatorname{Alg}_{3}(\operatorname{Mod}_{k}) Alg 2 ( E Pr k t + + ) : Alg 2 ( Pr k t + ) → Alg 2 ( Alg 1 ( Mod k )) ≃ Alg 3 ( Mod k ) where the final equivalence is Dunn additivity theorem. Then Hochschild cohomology H H ∙ ( P ( A ) / k ) ∈ Alg 3 ( Mod k ) \mathcal{HH}^{\bullet}(\mathcal{P}(\mathcal{A})/k)\in\operatorname{Alg}_{3}(\operatorname{Mod}_{k}) HH ∙ ( P ( A ) / k ) ∈ Alg 3 ( Mod k ) is defined to be the image of Hom A B M o d A ( End ( Fun ( X , Pr k t + ) ) ) ( A , A ) \operatorname{Hom}_{\mathcal{A}\mathbin{\mathrm{BMod}}_{\mathcal{A}}(\operatorname{End}(\operatorname{Fun}(X,\operatorname{Pr}_{k}^{\mathrm{t}+})))}(\mathcal{A},\mathcal{A}) Hom A BMod A ( End ( Fun ( X , Pr k t + ))) ( A , A ) under Alg 2 ( E Pr k t + + ) \operatorname{Alg}_{2}(E_{\operatorname{Pr}_{k}^{\mathrm{t}+}}^{+}) Alg 2 ( E Pr k t + + ) , that is the endomorphism algebra of a unit object of Hom A B M o d A ( End ( Fun ( X , P ) ) ) ( A , A ) \operatorname{Hom}_{\mathcal{A}\mathbin{\mathrm{BMod}}_{\mathcal{A}}(\operatorname{End}(\operatorname{Fun}(X,\mathcal{P})))}(\mathcal{A},\mathcal{A}) Hom A BMod A ( End ( Fun ( X , P ))) ( A , A ) (cf. Example 6.4). By construction, it has the required property.
Suppose that P = Pr k t ± \mathcal{P}=\operatorname{Pr}_{k}^{\mathrm{t}\pm} P = Pr k t ± . We take into account the localization adjoint pair L ^ : Pr k t + ⇄ Pr k t ± \widehat{L}:\operatorname{Pr}_{k}^{\mathrm{t}+}\rightleftarrows\operatorname{Pr}_{k}^{\mathrm{t}\pm} L : Pr k t + ⇄ Pr k t ± . The Hochschild cohomology H H ∙ ( P ( A ) / k ) ∈ Alg 3 ( Mod k ) \mathcal{HH}^{\bullet}(\mathcal{P}(\mathcal{A})/k)\in\operatorname{Alg}_{3}(\operatorname{Mod}_{k}) HH ∙ ( P ( A ) / k ) ∈ Alg 3 ( Mod k ) is defined to be the image of Hom A B M o d A ( End ( Fun ( X , Pr k t ± ) ) ) ( A , A ) ∈ Alg 2 ( Pr k t + ) \operatorname{Hom}_{\mathcal{A}\mathbin{\mathrm{BMod}}_{\mathcal{A}}(\operatorname{End}(\operatorname{Fun}(X,\operatorname{Pr}_{k}^{\mathrm{t}\pm})))}(\mathcal{A},\mathcal{A})\in\operatorname{Alg}_{2}(\operatorname{Pr}_{k}^{\mathrm{t}+}) Hom A BMod A ( End ( Fun ( X , Pr k t ± ))) ( A , A ) ∈ Alg 2 ( Pr k t + ) under Alg 2 ( E Pr k t + + ) \operatorname{Alg}_{2}(E_{\operatorname{Pr}_{k}^{\mathrm{t}+}}^{+}) Alg 2 ( E Pr k t + + ) .
Suppose that P = Pr k \mathcal{P}=\operatorname{Pr}_{k} P = Pr k . Using the adjoint pair θ 0 Pr k , 0 : Alg 1 ( Mod k ) ⇄ ( Pr k ) Mod k / : E Pr k + \theta_{0\operatorname{Pr}_{k},0}:\operatorname{Alg}_{1}(\operatorname{Mod}_{k})\rightleftarrows(\operatorname{Pr}_{k})_{\operatorname{Mod}_{k}/}:E_{\operatorname{Pr}_{k}}^{+} θ 0 Pr k , 0 : Alg 1 ( Mod k ) ⇄ ( Pr k ) Mod k / : E Pr k + instead of ( θ 0 Pr k t + , 0 , E Pr k t + + ) (\theta_{0\operatorname{Pr}_{k}^{\mathrm{t}+},0},E_{\operatorname{Pr}_{k}^{\mathrm{t}+}}^{+}) ( θ 0 Pr k t + , 0 , E Pr k t + + ) , one can construct Hochschild cohomology in the same way as above.
Proof of Proposition 6.6. The above construction shows our assertion.
We prove that in this setting Hochschild cohomology is independent of the choice of P \mathcal{P} P .
Proposition 6.9. Let A \mathcal{A} A be a Pr k t + \operatorname{Pr}_{k}^{\mathrm{t}+} Pr k t + -enriched category having a prescribed set X X X of objects (i.e., an algebra object A ∈ Alg 1 ( Fun ( X , Pr k t + ) ) \mathcal{A}\in\operatorname{Alg}_{1}(\operatorname{Fun}(X,\operatorname{Pr}_{k}^{\mathrm{t}+})) A ∈ Alg 1 ( Fun ( X , Pr k t + )) ).
(1) Consider P Pr k t + ( A ) ∈ 2 Pr k t + P_{\operatorname{Pr}_{k}^{\mathrm{t}+}}(\mathcal{A})\in2\operatorname{Pr}_{k}^{\mathrm{t}+} P Pr k t + ( A ) ∈ 2 Pr k t + . Let P Pr k t + ( A ) ∈ 2 Pr k P_{\operatorname{Pr}_{k}^{\mathrm{t}+}}(\mathcal{A})\in2\operatorname{Pr}_{k} P Pr k t + ( A ) ∈ 2 Pr k denote its underlying k k k -linear stable presentable ( ∞ , 2 ) (\infty,2) ( ∞ , 2 ) -category P Pr k t + ( A ) = P Pr k t + ( A ) ⊗ Pr k t + Pr k P_{\operatorname{Pr}_{k}^{\mathrm{t}+}}(\mathcal{A})=P_{\operatorname{Pr}_{k}^{\mathrm{t}+}}(\mathcal{A})\otimes_{\operatorname{Pr}_{k}^{\mathrm{t}+}}\operatorname{Pr}_{k} P Pr k t + ( A ) = P Pr k t + ( A ) ⊗ Pr k t + Pr k . Then there exists a canonical equivalence H H 2 Pr k t + ∙ ( P Pr k t + ( A ) / k ) ≃ H H 2 Pr k ∙ ( P Pr k t + ( A ) / k ) \mathcal{HH}^{\bullet}_{2\operatorname{Pr}_{k}^{\mathrm{t}+}}(P_{\operatorname{Pr}_{k}^{\mathrm{t}+}}(\mathcal{A})/k)\simeq\mathcal{HH}^{\bullet}_{2\operatorname{Pr}_{k}}(P_{\operatorname{Pr}_{k}^{\mathrm{t}+}}(\mathcal{A})/k) HH 2 Pr k t + ∙ ( P Pr k t + ( A ) / k ) ≃ HH 2 Pr k ∙ ( P Pr k t + ( A ) / k ) in Alg 3 ( Mod k ) \operatorname{Alg}_{3}(\operatorname{Mod}_{k}) Alg 3 ( Mod k ) .
(2) Suppose that A = A ^ \mathcal{A}=\widehat{\mathcal{A}} A = A is the Pr k t + \operatorname{Pr}_{k}^{\mathrm{t}+} Pr k t + -enriched category obtained from a Pr k t ± \operatorname{Pr}_{k}^{\mathrm{t}\pm} Pr k t ± -enriched category A ^ \widehat{\mathcal{A}} A . Consider P Pr k t ± ( A ^ ) ∈ 2 Pr k t ± P_{\operatorname{Pr}_{k}^{\mathrm{t}\pm}}(\widehat{\mathcal{A}})\in2\operatorname{Pr}_{k}^{\mathrm{t}\pm} P Pr k t ± ( A ) ∈ 2 Pr k t ± and P Pr k t + ( A ^ ) ∈ 2 Pr k t + P_{\operatorname{Pr}_{k}^{\mathrm{t}+}}(\widehat{\mathcal{A}})\in2\operatorname{Pr}_{k}^{\mathrm{t}+} P Pr k t + ( A ) ∈ 2 Pr k t + . Then there exists a canonical equivalence H H 2 Pr k t ± ∙ ( P Pr k t ± ( A ^ ) / k ) ≃ H H 2 Pr k t + ∙ ( P Pr k t ± ( A ^ ) / k ) \mathcal{HH}^{\bullet}_{2\operatorname{Pr}_{k}^{\mathrm{t}\pm}}(P_{\operatorname{Pr}_{k}^{\mathrm{t}\pm}}(\widehat{\mathcal{A}})/k)\simeq\mathcal{HH}^{\bullet}_{2\operatorname{Pr}_{k}^{\mathrm{t}+}}(P_{\operatorname{Pr}_{k}^{\mathrm{t}\pm}}(\widehat{\mathcal{A}})/k) HH 2 Pr k t ± ∙ ( P Pr k t ± ( A ) / k ) ≃ HH 2 Pr k t + ∙ ( P Pr k t ± ( A ) / k ) in Alg 3 ( Mod k ) \operatorname{Alg}_{3}(\operatorname{Mod}_{k}) Alg 3 ( Mod k ) .
Proof. We prove (1). By Proposition 6.5, End ( P ( A ) ) ∈ Alg 1 ( 2 Pr k ) \operatorname{End}(P(\mathcal{A})) \in\operatorname{Alg}_{1}(2\operatorname{Pr}_{k}) End ( P ( A )) ∈ Alg 1 ( 2 Pr k ) is canonically equivalent to End ( P ( A ) ) ⊗ Pr k t + Pr k \operatorname{End}(P(\mathcal{A})) \otimes_{\operatorname{Pr}_{k}^{t+}} \operatorname{Pr}_{k} End ( P ( A )) ⊗ Pr k t + Pr k , where ( − ) ⊗ Pr k t + Pr k (-) \otimes_{\operatorname{Pr}_{k}^{t+}} \operatorname{Pr}_{k} ( − ) ⊗ Pr k t + Pr k indicates the symmetric monoidal base change functor 2 Pr k t + → 2 Pr k 2\operatorname{Pr}_{k}^{t+} \to2\operatorname{Pr}_{k} 2 Pr k t + → 2 Pr k . It gives rise to
Hom A BMod A ( End ( Fun ( X , Pr k t + ) ) , ( A , A ) ) → Hom A BMod A ( End ( Fun ( X , Pr k ) ) , ( A , A ) ) \operatorname{Hom}_{\mathcal{A}\operatorname{BMod}_{\mathcal{A}}}\left(\operatorname{End}(\operatorname{Fun}(X,\operatorname{Pr}_{k}^{t+})),(\mathcal{A},\mathcal{A})\right) \to\operatorname{Hom}_{\mathcal{A}\operatorname{BMod}_{\mathcal{A}}}\left(\operatorname{End}(\operatorname{Fun}(X,\operatorname{Pr}_{k})),(\mathcal{A},\mathcal{A})\right) Hom A BMod A ( End ( Fun ( X , Pr k t + )) , ( A , A ) ) → Hom A BMod A ( End ( Fun ( X , Pr k )) , ( A , A ) )
in Alg 2 ( Pr k ) \operatorname{Alg}_{2}(\operatorname{Pr}_{k}) Alg 2 ( Pr k ) . Here A ∈ Fun ( X × X , Pr k ) \mathcal{A} \in\operatorname{Fun}(X \times X,\operatorname{Pr}_{k}) A ∈ Fun ( X × X , Pr k ) is the Pr k \operatorname{Pr}_{k} Pr k -enriched category determined by Fun ( X × X , Pr k t + ) → Fun ( X × X , Pr k ) \operatorname{Fun}(X \times X,\operatorname{Pr}_{k}^{t+}) \to\operatorname{Fun}(X \times X,\operatorname{Pr}_{k}) Fun ( X × X , Pr k t + ) → Fun ( X × X , Pr k ) . It is enough to prove that it induces an equivalence between endomorphism algebra objects of the unit objects (the identity functors) in Alg 3 ( Mod k ) \operatorname{Alg}_{3}(\operatorname{Mod}_{k}) Alg 3 ( Mod k ) . We identify the left-hand side with
lim [ n ] ∈ Δ Hom Fun ( X × X , Pr k t + ) ( A ⊗ 2 n + 1 , A ) . \lim_{[n]\in\Delta}\operatorname{Hom}_{\operatorname{Fun}(X\times X,\operatorname{Pr}_{k}^{t+})}(\mathcal{A}^{\otimes2n+1},\mathcal{A}). [ n ] ∈ Δ lim Hom Fun ( X × X , Pr k t + ) ( A ⊗ 2 n + 1 , A ) .
This limit is taken in Pr k t + \operatorname{Pr}_{k}^{t+} Pr k t + . As for H H ∙ ( P ( A ) / k ) HH^\bullet(P(\mathcal{A})/k) H H ∙ ( P ( A ) / k ) we identify the right-hand side with
lim [ n ] ∈ Δ Hom Fun ( X × X , Pr k ) ( A ⊗ 2 n + 1 , A ) , \lim_{[n]\in\Delta}\operatorname{Hom}_{\operatorname{Fun}(X\times X,\operatorname{Pr}_{k})}(\mathcal{A}^{\otimes2n+1},\mathcal{A}), [ n ] ∈ Δ lim Hom Fun ( X × X , Pr k ) ( A ⊗ 2 n + 1 , A ) ,
where the limit is taken in Pr k \operatorname{Pr}_{k} Pr k . We consider an intermediate objects
lim [ n ] ∈ Δ Hom Fun ( X × X , Pr k t ) ( ( A ′ ) ⊗ 2 n + 1 , A ′ ) , \lim_{[n]\in\Delta}\operatorname{Hom}_{\operatorname{Fun}(X\times X,\operatorname{Pr}_{k}^{t})}((\mathcal{A}')^{\otimes2n+1},\mathcal{A}'), [ n ] ∈ Δ lim Hom Fun ( X × X , Pr k t ) (( A ′ ) ⊗ 2 n + 1 , A ′ ) ,
where A ′ ∈ Fun ( X × X , Pr k t ) \mathcal{A}' \in\operatorname{Fun}(X \times X,\operatorname{Pr}_{k}^{t}) A ′ ∈ Fun ( X × X , Pr k t ) is the Pr k t \operatorname{Pr}_{k}^{t} Pr k t -enriched category determined by Fun ( X × X , Pr k t + ) → Fun ( X × X , Pr k t ) \operatorname{Fun}(X \times X,\operatorname{Pr}_{k}^{t+}) \to\operatorname{Fun}(X \times X,\operatorname{Pr}_{k}^{t}) Fun ( X × X , Pr k t + ) → Fun ( X × X , Pr k t ) (induced by Pr k t + → Pr k t \operatorname{Pr}_{k}^{t+} \to\operatorname{Pr}_{k}^{t} Pr k t + → Pr k t ), and the limit is taken in Pr k t \operatorname{Pr}_{k}^{t} Pr k t . (We note that it “computes” Hom A ′ BMod A ′ ( End ( Fun ( X , Pr k t ) ) , ( A ′ , A ′ ) ) \operatorname{Hom}_{\mathcal{A}'\operatorname{BMod}_{\mathcal{A}'}}(\operatorname{End}(\operatorname{Fun}(X,\operatorname{Pr}_{k}^{t})),(\mathcal{A}',\mathcal{A}')) Hom A ′ BMod A ′ ( End ( Fun ( X , Pr k t )) , ( A ′ , A ′ )) .)
For P , Q ∈ Pr k t \mathcal{P},\mathcal{Q} \in\operatorname{Pr}_{k}^{t} P , Q ∈ Pr k t , an internal Hom object Hom Pr k t ( P , Q ) \operatorname{Hom}_{\operatorname{Pr}_{k}^{t}}(\mathcal{P},\mathcal{Q}) Hom Pr k t ( P , Q ) from P \mathcal{P} P to Q \mathcal{Q} Q in Pr k t \operatorname{Pr}_{k}^{t} Pr k t is given by the pair ( Hom Pr k ( P , Q ) , Hom Pr k ( P , Q ) ≤ 0 ) (\operatorname{Hom}_{\operatorname{Pr}_{k}}(\mathcal{P},\mathcal{Q}),\operatorname{Hom}_{\operatorname{Pr}_{k}}(\mathcal{P},\mathcal{Q})^{\leq0}) ( Hom Pr k ( P , Q ) , Hom Pr k ( P , Q ) ≤ 0 ) such that Hom Pr k ( P , Q ) ∈ Pr k \operatorname{Hom}_{\operatorname{Pr}_{k}}(\mathcal{P},\mathcal{Q}) \in\operatorname{Pr}_{k} Hom Pr k ( P , Q ) ∈ Pr k is the internal Hom object from the underlying object P \mathcal{P} P to Q \mathcal{Q} Q , and Hom Pr k ( P , Q ) ≤ 0 \operatorname{Hom}_{\operatorname{Pr}_{k}}(\mathcal{P},\mathcal{Q})^{\leq0} Hom Pr k ( P , Q ) ≤ 0 is the full subcategory, spanned by right t t t -exact functors (cf. the proof of Corollary 6.8). Moreover, the symmetric monoidal forgetful functor Pr k t → Pr k \operatorname{Pr}_{k}^{t} \to\operatorname{Pr}_{k} Pr k t → Pr k preserves small limits (see Proposition 3.12). Thus, we see that the canonical functor
lim [ n ] ∈ Δ Hom Fun ( X × X , Pr k t ) ( ( A ′ ) ⊗ 2 n + 1 , A ′ ) → lim [ n ] ∈ Δ Hom Fun ( X × X , Pr k ) ( A ⊗ 2 n + 1 , A ) \lim_{[n]\in\Delta}\operatorname{Hom}_{\operatorname{Fun}(X\times X,\operatorname{Pr}_{k}^{t})}((\mathcal{A}')^{\otimes2n+1},\mathcal{A}') \to\lim_{[n]\in\Delta}\operatorname{Hom}_{\operatorname{Fun}(X\times X,\operatorname{Pr}_{k})}(\mathcal{A}^{\otimes2n+1},\mathcal{A}) [ n ] ∈ Δ lim Hom Fun ( X × X , Pr k t ) (( A ′ ) ⊗ 2 n + 1 , A ′ ) → [ n ] ∈ Δ lim Hom Fun ( X × X , Pr k ) ( A ⊗ 2 n + 1 , A )
in Pr k \operatorname{Pr}_{k} Pr k is an equivalence. There also exists a canonical map
f : lim [ n ] ∈ Δ Hom Fun ( X × X , Pr k t + ) ( A ⊗ 2 n + 1 , A ) → lim [ n ] ∈ Δ Hom Fun ( X × X , Pr k t ) ( ( A ′ ) ⊗ 2 n + 1 , A ′ ) f:\lim_{[n]\in\Delta}\operatorname{Hom}_{\operatorname{Fun}(X\times X,\operatorname{Pr}_{k}^{t+})}(\mathcal{A}^{\otimes2n+1},\mathcal{A}) \to\lim_{[n]\in\Delta}\operatorname{Hom}_{\operatorname{Fun}(X\times X,\operatorname{Pr}_{k}^{t})}((\mathcal{A}')^{\otimes2n+1},\mathcal{A}') f : [ n ] ∈ Δ lim Hom Fun ( X × X , Pr k t + ) ( A ⊗ 2 n + 1 , A ) → [ n ] ∈ Δ lim Hom Fun ( X × X , Pr k t ) (( A ′ ) ⊗ 2 n + 1 , A ′ )
in Pr k t \operatorname{Pr}_{k}^{t} Pr k t , which exhibits the left-hand side as a right-completion of the right-hand side (note that the internal Hom objects in Pr k t + \operatorname{Pr}_{k}^{t+} Pr k t + are constructed as right completions of those in Pr k t \operatorname{Pr}_{k}^{t} Pr k t ). It is enough to prove that f f f induces an equivalence of endomorphism algebra objects of unit objects (identity functors). We consider the diagram
Alg 1 ( Mod k ) → θ Pr k t , 0 ( Pr k t ) Mod k / θ Pr k t + , 0 ↓ R ^ ↑ ( Pr k t + ) Mod k / → I + ( Pr k t ) Mod k / \begin{CD}
\operatorname{Alg}_{1}(\operatorname{Mod}_{k}) @>{\theta_{\operatorname{Pr}_{k}^{t},0}}>> (\operatorname{Pr}_{k}^{t})_{\operatorname{Mod}_{k}}/ \\
@V{\theta_{\operatorname{Pr}_{k}^{t+},0}}VV @A{\widehat{R}}AA \\
(\operatorname{Pr}_{k}^{t+})_{\operatorname{Mod}_{k}}/ @>{I^{+}}>> (\operatorname{Pr}_{k}^{t})_{\operatorname{Mod}_{k}}/
\end{CD} Alg 1 ( Mod k ) θ Pr k t + , 0 ↓ ⏐ ( Pr k t + ) Mod k / θ Pr k t , 0 I + ( Pr k t ) Mod k / R ⏐ ↑ ( Pr k t ) Mod k /
Both θ Pr k t , 0 \theta_{\operatorname{Pr}_{k}^{t},0} θ Pr k t , 0 and θ Pr k t + , 0 \theta_{\operatorname{Pr}_{k}^{t+},0} θ Pr k t + , 0 send A \mathcal{A} A to ( ( LMod A , LMod A ≤ 0 ) , A ∈ LMod A ) ((\operatorname{LMod}_{\mathcal{A}},\operatorname{LMod}_{\mathcal{A}}^{\leq0}),\mathcal{A}\in\operatorname{LMod}_{\mathcal{A}}) (( LMod A , LMod A ≤ 0 ) , A ∈ LMod A ) . The functor I + I^{+} I + is the inclusion, while R ^ \widehat{R} R is the right completion functor. Note that I + ∘ θ Pr k t + , 0 = θ Pr k t , 0 I^{+}\circ\theta_{\operatorname{Pr}_{k}^{t+},0}=\theta_{\operatorname{Pr}_{k}^{t},0} I + ∘ θ Pr k t + , 0 = θ Pr k t , 0 , and the right adjoint functors of θ Pr k t , 0 \theta_{\operatorname{Pr}_{k}^{t},0} θ Pr k t , 0 and θ Pr k t + , 0 \theta_{\operatorname{Pr}_{k}^{t+},0} θ Pr k t + , 0 are given informally by ( L , L ) ↦ End ( L ) op (L,L)\mapsto\operatorname{End}(L)^{\operatorname{op}} ( L , L ) ↦ End ( L ) op . From the right completion f f f and this commutative diagram, we conclude that f f f induces an equivalence of endomorphism algebra objects of unit objects.
We prove (2). As in (1), End ( P Pr k t ± ( A ^ ) ) ∈ Alg 1 ( 2 Pr k t ± ) \operatorname{End}(P_{\operatorname{Pr}_{k}^{t\pm}}(\widehat{\mathcal{A}})) \in\operatorname{Alg}_{1}(2\operatorname{Pr}_{k}^{t\pm}) End ( P Pr k t ± ( A )) ∈ Alg 1 ( 2 Pr k t ± ) is canonically equivalent to End ( P Pr k t ± ( A ) ) ⊗ Pr k t ± Pr k t ± \operatorname{End}(P_{\operatorname{Pr}_{k}^{t\pm}}(\mathcal{A}))\otimes_{\operatorname{Pr}_{k}^{t\pm}}\operatorname{Pr}_{k}^{t\pm} End ( P Pr k t ± ( A )) ⊗ Pr k t ± Pr k t ± , where ( − ) ⊗ Pr k t + Pr k t ± (-)\otimes_{\operatorname{Pr}_{k}^{t+}}\operatorname{Pr}_{k}^{t\pm} ( − ) ⊗ Pr k t + Pr k t ± indicates the symmetric monoidal base change functor 2 Pr k t + → 2 Pr k t ± 2\operatorname{Pr}_{k}^{t+}\to2\operatorname{Pr}_{k}^{t\pm} 2 Pr k t + → 2 Pr k t ± . It gives rise to
Hom A ^ BMod A ^ ( End ( Fun ( X , Pr k t + ) ) , ( A ^ , A ^ ) ) → Hom A ^ BMod A ^ ( End ( Fun ( X , Pr k t ± ) ) , ( A ^ , A ^ ) ) \operatorname{Hom}_{\widehat{\mathcal{A}}\operatorname{BMod}_{\widehat{\mathcal{A}}}}\left(\operatorname{End}(\operatorname{Fun}(X,\operatorname{Pr}_{k}^{t+})),(\widehat{\mathcal{A}},\widehat{\mathcal{A}})\right)\to\operatorname{Hom}_{\widehat{\mathcal{A}}\operatorname{BMod}_{\widehat{\mathcal{A}}}}\left(\operatorname{End}(\operatorname{Fun}(X,\operatorname{Pr}_{k}^{t\pm})),(\widehat{\mathcal{A}},\widehat{\mathcal{A}})\right) Hom A BMod A ( End ( Fun ( X , Pr k t + )) , ( A , A ) ) → Hom A BMod A ( End ( Fun ( X , Pr k t ± )) , ( A , A ) )
in Alg 2 ( Pr k t + ) \operatorname{Alg}_{2}(\operatorname{Pr}_{k}^{t+}) Alg 2 ( Pr k t + ) . From the proof of Corollary 6.8, it can be identified with the canonical morphism
f : lim [ n ] ∈ Δ Hom Fun ( X × X , Pr k t + ) ( A ^ ⊗ 2 n + 1 , A ^ ) → lim [ n ] ∈ Δ Hom Fun ( X × X , Pr k t ± ) ( A ^ ⊗ 2 n + 1 , A ^ ) f:\lim_{[n]\in\Delta}\operatorname{Hom}_{\operatorname{Fun}(X\times X,\operatorname{Pr}_{k}^{t+})}(\widehat{\mathcal{A}}^{\otimes2n+1},\widehat{\mathcal{A}})\to\lim_{[n]\in\Delta}\operatorname{Hom}_{\operatorname{Fun}(X\times X,\operatorname{Pr}_{k}^{t\pm})}(\widehat{\mathcal{A}}^{\otimes2n+1},\widehat{\mathcal{A}}) f : [ n ] ∈ Δ lim Hom Fun ( X × X , Pr k t + ) ( A ⊗ 2 n + 1 , A ) → [ n ] ∈ Δ lim Hom Fun ( X × X , Pr k t ± ) ( A ⊗ 2 n + 1 , A )
in Pr k t + \operatorname{Pr}_{k}^{t+} Pr k t + . Here A ^ ⊗ 2 n + 1 \widehat{\mathcal{A}}^{\otimes2n+1} A ⊗ 2 n + 1 is the tensor product in Fun ( X × X , Pr k t + ) \operatorname{Fun}(X\times X,\operatorname{Pr}_{k}^{t+}) Fun ( X × X , Pr k t + ) , while A ^ ⊗ ^ 2 n + 1 \widehat{\mathcal{A}}^{\widehat{\otimes}2n+1} A ⊗ 2 n + 1 is the tensor product in Fun ( X × X , Pr k t ± ) \operatorname{Fun}(X\times X,\operatorname{Pr}_{k}^{t\pm}) Fun ( X × X , Pr k t ± ) . It is enough to prove that f f f is an equivalence. Each morphism
Hom Fun ( X × X , Pr k t + ) ( A ^ ⊗ 2 n + 1 , A ^ ) → Hom Fun ( X × X , Pr k t ± ) ( A ^ ⊗ ^ 2 n + 1 , A ^ ) \operatorname{Hom}_{\operatorname{Fun}(X\times X,\operatorname{Pr}_{k}^{t+})}(\widehat{\mathcal{A}}^{\otimes2n+1},\widehat{\mathcal{A}})\to\operatorname{Hom}_{\operatorname{Fun}(X\times X,\operatorname{Pr}_{k}^{t\pm})}(\widehat{\mathcal{A}}^{\widehat{\otimes}2n+1},\widehat{\mathcal{A}}) Hom Fun ( X × X , Pr k t + ) ( A ⊗ 2 n + 1 , A ) → Hom Fun ( X × X , Pr k t ± ) ( A ⊗ 2 n + 1 , A )
is an equivalence since it is induced by the left completion, and A ^ ⊗ 2 n + 1 \widehat{\mathcal{A}}^{\otimes2n+1} A ⊗ 2 n + 1 is a left completion of A ^ ⊗ 2 n + 1 \widehat{\mathcal{A}}^{\otimes2n+1} A ⊗ 2 n + 1 . Thus, f f f is an equivalence. □ \square □
Categorified sheaves
Let X \mathcal{X} X be a perfect stack over k k k in the sense of [4 ] (see loc. cit. ). Typical examples include quasi-compact and separated derived schemes over k k k . We consider the “categorified sheaves” on X \mathcal{X} X , that is, sheaves of categories. We briefly describe them.
For simplicity, we suppose that X \mathcal{X} X is a quasi-compact and separated derived scheme for a moment. Let QC ( X ) \operatorname{QC}(\mathcal{X}) QC ( X ) be the k k k -linear symmetric monoidal ∞ \infty ∞ -category of quasi-coherent complexes on X \mathcal{X} X (cf. [25 VIII] , [4 ] ). If we write AffOp X \operatorname{AffOp}_{\mathcal{X}} AffOp X for the poset of affine open subschemes, then QC ( X ) = lim Spec R ( U ) ∈ AffOp X Mod R ( U ) \operatorname{QC}(\mathcal{X}) = \lim_{\operatorname{Spec} R(U)\in\operatorname{AffOp}_{\mathcal{X}}} \operatorname{Mod}_{R(U)} QC ( X ) = lim Spec R ( U ) ∈ AffOp X Mod R ( U ) where i ∗ : Mod R ( U ) → Mod R ( V ) i^*:\operatorname{Mod}_{R(U)}\to\operatorname{Mod}_{R(V)} i ∗ : Mod R ( U ) → Mod R ( V ) is base change functor along R ( U ) → R ( V ) R(U)\to R(V) R ( U ) → R ( V ) for each i : V → U i:V\to U i : V → U . By definition, if X = Spec R \mathcal{X}=\operatorname{Spec} R X = Spec R , then QC ( X ) = Mod R \operatorname{QC}(\mathcal{X})=\operatorname{Mod}_R QC ( X ) = Mod R . We consider the functor
2 QC : AffOp X o p → 2 Pr k {}^2\operatorname{QC}:\operatorname{AffOp}_{\mathcal{X}}^{\mathrm{op}}\to{}^2\operatorname{Pr}_k 2 QC : AffOp X op → 2 Pr k
which carries U = Spec R ( U ) U=\operatorname{Spec} R(U) U = Spec R ( U ) to Mod Mod R ( U ) ( Pr k ) \operatorname{Mod}_{\operatorname{Mod}_{R(U)}}(\operatorname{Pr}_k) Mod Mod R ( U ) ( Pr k ) and sends a morphism i : V → U i:V\to U i : V → U to Mod Mod R ( U ) ( Pr k ) → Mod Mod R ( V ) ( Pr k ) \operatorname{Mod}_{\operatorname{Mod}_{R(U)}}(\operatorname{Pr}_k)\to\operatorname{Mod}_{\operatorname{Mod}_{R(V)}}(\operatorname{Pr}_k) Mod Mod R ( U ) ( Pr k ) → Mod Mod R ( V ) ( Pr k ) given by D ↦ D ⊗ Mod R ( U ) Mod R ( V ) \mathcal{D}\mapsto\mathcal{D}\otimes_{\operatorname{Mod}_{R(U)}}\operatorname{Mod}_{R(V)} D ↦ D ⊗ Mod R ( U ) Mod R ( V ) (i.e., the base change along i ∗ : Mod R ( U ) → Mod R ( V ) i^*:\operatorname{Mod}_{R(U)}\to\operatorname{Mod}_{R(V)} i ∗ : Mod R ( U ) → Mod R ( V ) ). We define the ∞ \infty ∞ -category of categorified sheaves on X \mathcal{X} X as
QC 2 -cat ( X ) : = lim U ∈ AffOp X 2 QC ( U ) = lim U ∈ AffOp X Mod Mod R ( U ) ( Pr k ) . \operatorname{QC}_{2\text{-cat}}(\mathcal{X}) := \lim_{U\in\operatorname{AffOp}_{\mathcal{X}}}{}^2\operatorname{QC}(U)=\lim_{U\in\operatorname{AffOp}_{\mathcal{X}}}\operatorname{Mod}_{\operatorname{Mod}_{R(U)}}(\operatorname{Pr}_k). QC 2 -cat ( X ) := U ∈ AffOp X lim 2 QC ( U ) = U ∈ AffOp X lim Mod Mod R ( U ) ( Pr k ) .
Informally, an object F ∈ QC 2 -cat ( X ) \mathcal{F}\in\operatorname{QC}_{2\text{-cat}}(\mathcal{X}) F ∈ QC 2 -cat ( X ) is an assignment Γ ( U , F ) ∈ Mod Mod R ( U ) ( Pr k ) \Gamma(U,\mathcal{F})\in\operatorname{Mod}_{\operatorname{Mod}_{R(U)}}(\operatorname{Pr}_k) Γ ( U , F ) ∈ Mod Mod R ( U ) ( Pr k ) to each U ∈ AffOp X U\in\operatorname{AffOp}_{\mathcal{X}} U ∈ AffOp X and an equivalence
Γ ( U , F ) ⊗ Mod R ( U ) Mod R ( V ) ≃ Γ ( V , F ) \Gamma(U,\mathcal{F})\otimes_{\operatorname{Mod}_{R(U)}}\operatorname{Mod}_{R(V)}\simeq\Gamma(V,\mathcal{F}) Γ ( U , F ) ⊗ Mod R ( U ) Mod R ( V ) ≃ Γ ( V , F )
to each i : V → U i:V\to U i : V → U , endowed with homotopy coherent compatibility data.
Each Mod Mod R ( U ) ( Pr k ) \operatorname{Mod}_{\operatorname{Mod}_{R(U)}}(\operatorname{Pr}_k) Mod Mod R ( U ) ( Pr k ) admits a base change functor Mod QC ( X ) ( Pr k ) → Mod Mod R ( U ) ( Pr k ) \operatorname{Mod}_{\operatorname{QC}(\mathcal{X})}(\operatorname{Pr}_k)\to\operatorname{Mod}_{\operatorname{Mod}_{R(U)}}(\operatorname{Pr}_k) Mod QC ( X ) ( Pr k ) → Mod Mod R ( U ) ( Pr k ) along the pullback functor QC ( X ) → Mod R ( U ) \operatorname{QC}(\mathcal{X})\to\operatorname{Mod}_{R(U)} QC ( X ) → Mod R ( U ) . They determine a functor
Mod QC ( X ) ( Pr k ) → lim U ∈ AffOp X Mod Mod R ( U ) ( Pr k ) = QC 2 -cat ( X ) . \operatorname{Mod}_{\operatorname{QC}(\mathcal{X})}(\operatorname{Pr}_k)\to\lim_{U\in\operatorname{AffOp}_{\mathcal{X}}}\operatorname{Mod}_{\operatorname{Mod}_{R(U)}}(\operatorname{Pr}_k)=\operatorname{QC}_{2\text{-cat}}(\mathcal{X}). Mod QC ( X ) ( Pr k ) → U ∈ AffOp X lim Mod Mod R ( U ) ( Pr k ) = QC 2 -cat ( X ) .
By a 1-affineness of X \mathcal{X} X (cf. [10 Theorem 2.1.1] ), this functor is an equivalence. Thus, we may identify Mod QC ( X ) ( Pr k ) \operatorname{Mod}_{\operatorname{QC}(\mathcal{X})}(\operatorname{Pr}_k) Mod QC ( X ) ( Pr k ) with QC 2 -cat ( X ) \operatorname{QC}_{2\text{-cat}}(\mathcal{X}) QC 2 -cat ( X ) .
The stable ∞ \infty ∞ -category QC ( X ) \operatorname{QC}(\mathcal{X}) QC ( X ) has the “standard” accessible left and right complete t t t -structure (see [25 VIII, Proposition 2.3.16, Proposition 2.3.18] ). An object M ∈ QC ( X ) M\in\operatorname{QC}(\mathcal{X}) M ∈ QC ( X ) is connective on this t t t -structure if and only if for any U ∈ AffOp X U\in\operatorname{AffOp}_{\mathcal{X}} U ∈ AffOp X the restriction M ∣ U ∈ Mod R ( U ) M|_U\in\operatorname{Mod}_{R(U)} M ∣ U ∈ Mod R ( U ) is connective. We regard QC ( X ) \operatorname{QC}(\mathcal{X}) QC ( X ) as an object of Alg 1 ( Pr k t + ) \operatorname{Alg}_1(\operatorname{Pr}_k^{t+}) Alg 1 ( Pr k t + ) that comes from CAlg ( Pr k t + ) \operatorname{CAlg}(\operatorname{Pr}_k^{t+}) CAlg ( Pr k t + ) . One can consider the t t t -structured categorified sheaves. We define
QC 2 -cat , t ( X ) : = lim U ∈ AffOp X Mod Mod R ( U ) ( Pr k t ) . \operatorname{QC}_{2\text{-cat},t}(\mathcal{X}):=\lim_{U\in\operatorname{AffOp}_{\mathcal{X}}}\operatorname{Mod}_{\operatorname{Mod}_{R(U)}}(\operatorname{Pr}_k^t). QC 2 -cat , t ( X ) := U ∈ AffOp X lim Mod Mod R ( U ) ( Pr k t ) .
Then, by the above 1-affineness of X \mathcal{X} X , one sees that a canonical equivalence Mod QC ( X ) ( Pr k t ) ≃ QC 2 -cat , t ( X ) \operatorname{Mod}_{\operatorname{QC}(\mathcal{X})}(\operatorname{Pr}_k^t)\simeq\operatorname{QC}_{2\text{-cat},t}(\mathcal{X}) Mod QC ( X ) ( Pr k t ) ≃ QC 2 -cat , t ( X ) (with some elaboration we also have a version of Pr k t + \operatorname{Pr}_k^{t+} Pr k t + ). Therefore, we will adopt Mod QC ( X ) ( Pr k t ) \operatorname{Mod}_{\operatorname{QC}(\mathcal{X})}(\operatorname{Pr}_k^t) Mod QC ( X ) ( Pr k t ) as the t t t -structured enhancement of Mod QC ( X ) ( Pr k ) \operatorname{Mod}_{\operatorname{QC}(\mathcal{X})}(\operatorname{Pr}_k) Mod QC ( X ) ( Pr k ) .
By [25 VIII, Proposition 2.3.18] , the t t t -structure on QC ( X ) \operatorname{QC}(\mathcal{X}) QC ( X ) is both left and right complete. We think of the monoidal ∞ \infty ∞ -category QC ( X ) \operatorname{QC}(\mathcal{X}) QC ( X ) as a Pr k t ± \operatorname{Pr}_k^{t\pm} Pr k t ± -enriched category having a single object, which we usually denote by B QC ( X ) B\operatorname{QC}(\mathcal{X}) B QC ( X ) . In this case, P ( B QC ( X ) ) = Mod QC ( X ) ( Pr k t ± ) \operatorname{P}(B\operatorname{QC}(\mathcal{X}))=\operatorname{Mod}_{\operatorname{QC}(\mathcal{X})}(\operatorname{Pr}_k^{t\pm}) P ( B QC ( X )) = Mod QC ( X ) ( Pr k t ± ) . Note that Mod QC ( X ) ( Pr k t ± ) \operatorname{Mod}_{\operatorname{QC}(\mathcal{X})}(\operatorname{Pr}_k^{t\pm}) Mod QC ( X ) ( Pr k t ± ) belongs to 2 Pr k t ± {}^2\operatorname{Pr}_k^{t\pm} 2 Pr k t ± . We adopt this as the complete t t t -structured enhancement of Mod QC ( X ) ( Pr k ) \operatorname{Mod}_{\operatorname{QC}(\mathcal{X})}(\operatorname{Pr}_k) Mod QC ( X ) ( Pr k ) . Note that the underlying stable ( ∞ , 2 ) (\infty,2) ( ∞ , 2 ) -category Mod QC ( X ) ( Pr k t ± ) ⊗ Pr k t ± Pr k \operatorname{Mod}_{\operatorname{QC}(\mathcal{X})}(\operatorname{Pr}_k^{t\pm})\otimes_{\operatorname{Pr}_k^{t\pm}}\operatorname{Pr}_k Mod QC ( X ) ( Pr k t ± ) ⊗ Pr k t ± Pr k is equivalent to QC 2 -cat ( X ) \operatorname{QC}_{2\text{-cat}}(\mathcal{X}) QC 2 -cat ( X ) .
Proposition 6.10. Let X \mathcal{X} X be a perfect stack over k k k . Let H ∙ ( Mod QC ( X ) ( Pr k t ± ) / k ) \mathcal{H}^{\bullet}(\operatorname{Mod}_{\operatorname{QC}(\mathcal{X})}(\operatorname{Pr}_k^{t\pm})/k) H ∙ ( Mod QC ( X ) ( Pr k t ± ) / k ) be the Hochschild cohomology of Mod QC ( X ) ( Pr k t ± ) \operatorname{Mod}_{\operatorname{QC}(\mathcal{X})}(\operatorname{Pr}_k^{t\pm}) Mod QC ( X ) ( Pr k t ± ) (cf. Proposition 6.6). Then the underlying complex of Hochschild cohomology H ∙ ( Mod QC ( X ) ( Pr k t ± ) / k ) \mathcal{H}^{\bullet}(\operatorname{Mod}_{\operatorname{QC}(\mathcal{X})}(\operatorname{Pr}_k^{t\pm})/k) H ∙ ( Mod QC ( X ) ( Pr k t ± ) / k ) is equivalent to the Hom \operatorname{Hom} Hom complex Hom QC ( X ) ( O Δ , O Δ ) ≃ Mod k \operatorname{Hom}_{\operatorname{QC}(\mathcal{X})}(\mathcal{O}_{\Delta},\mathcal{O}_{\Delta})\simeq\operatorname{Mod}_k Hom QC ( X ) ( O Δ , O Δ ) ≃ Mod k where X \mathcal{X} X is the loop space of X \mathcal{X} X , defined as the fiber product X × X × k X X \mathcal{X}\times_{\mathcal{X}\times_k\mathcal{X}}\mathcal{X} X × X × k X X . The object O Δ ∈ QC ( X ) \mathcal{O}_{\Delta}\in\operatorname{QC}(\mathcal{X}) O Δ ∈ QC ( X ) is the pushforward of the structure sheaf O X ∈ QC ( X ) \mathcal{O}_{\mathcal{X}}\in\operatorname{QC}(\mathcal{X}) O X ∈ QC ( X ) along the diagonal morphism X → X × k X = L X \mathcal{X}\to\mathcal{X}\times_k\mathcal{X}=\mathcal{L}\mathcal{X} X → X × k X = LX .
Suppose that k k k is of characteristic zero, and X \mathcal{X} X is an ordinary scheme separated and of finite type. There exists an equivalence
H ∙ ( Mod QC ( X ) ( Pr k t ± ) / k ) ≃ Hom QC ( L X ) ( O Δ , O Δ ) ≃ Γ ( X , Sym O X ∙ ( T X / k [ − 2 ] ) ) \mathcal{H}^{\bullet}(\operatorname{Mod}_{\operatorname{QC}(\mathcal{X})}(\operatorname{Pr}_k^{t\pm})/k)\simeq\operatorname{Hom}_{\operatorname{QC}(\mathcal{L}\mathcal{X})}(\mathcal{O}_{\Delta},\mathcal{O}_{\Delta})\simeq\Gamma(\mathcal{X},\operatorname{Sym}^{\bullet}_{\mathcal{O}_{\mathcal{X}}}(T_{\mathcal{X}/k}[-2])) H ∙ ( Mod QC ( X ) ( Pr k t ± ) / k ) ≃ Hom QC ( LX ) ( O Δ , O Δ ) ≃ Γ ( X , Sym O X ∙ ( T X / k [ − 2 ]))
in Mod k \operatorname{Mod}_k Mod k where T X / k \mathbb{T}_{\mathcal{X}/k} T X / k is the tangent complex of X \mathcal{X} X , defined as the (derived) dual H o m QC ( X ) ( L X / k , O X ) \mathcal{H}om_{\operatorname{QC}(\mathcal{X})}(\mathbb{L}_{\mathcal{X}/k},\mathcal{O}_{\mathcal{X}}) H o m QC ( X ) ( L X / k , O X ) of the cotangent complex L X / k \mathbb{L}_{\mathcal{X}/k} L X / k . Here Γ ( X , − ) \Gamma(\mathcal{X},-) Γ ( X , − ) means the (derived) global sections, and Sym O X ∙ ( − ) \operatorname{Sym}^{\bullet}_{\mathcal{O}_{\mathcal{X}}}(-) Sym O X ∙ ( − ) means a free commutative algebra in QC ( X ) \operatorname{QC}(\mathcal{X}) QC ( X ) , that is, the symmetric algebra.
Proof. By Proposition 6.9, there exist equivalences
H H ∙ ( Mod QC ( X ) ( Pr k t ± ) / k ) ≃ H H ∙ ( Mod QC ( X ) ( Pr k t + ) / k ) ≃ H H ∙ ( Mod QC ( X ) ( Pr k ) / k ) . \mathcal{HH}^{\bullet}(\operatorname{Mod}_{\operatorname{QC}(\mathcal{X})}(\operatorname{Pr}_{k}^{\mathrm{t}\pm})/k) \simeq\mathcal{HH}^{\bullet}(\operatorname{Mod}_{\operatorname{QC}(\mathcal{X})}(\operatorname{Pr}_{k}^{\mathrm{t}+})/k) \simeq\mathcal{HH}^{\bullet}(\operatorname{Mod}_{\operatorname{QC}(\mathcal{X})}(\operatorname{Pr}_{k})/k). HH ∙ ( Mod QC ( X ) ( Pr k t ± ) / k ) ≃ HH ∙ ( Mod QC ( X ) ( Pr k t + ) / k ) ≃ HH ∙ ( Mod QC ( X ) ( Pr k ) / k ) .
We may and will replace H H ∙ ( Mod QC ( X ) ( Pr k t ± ) / k ) \mathcal{HH}^{\bullet}(\operatorname{Mod}_{\operatorname{QC}(\mathcal{X})}(\operatorname{Pr}_{k}^{\mathrm{t}\pm})/k) HH ∙ ( Mod QC ( X ) ( Pr k t ± ) / k ) with H H ∙ ( Mod QC ( X ) ( Pr k ) / k ) \mathcal{HH}^{\bullet}(\operatorname{Mod}_{\operatorname{QC}(\mathcal{X})}(\operatorname{Pr}_{k})/k) HH ∙ ( Mod QC ( X ) ( Pr k ) / k ) in the statement. As observed in Section 6.2, there exist canonical equivalences
H o m Pr k ( Mod QC ( X ) ( Pr k ) , Mod QC ( X ) ( Pr k ) ) ≃ Mod QC ( X ) ( Pr k ) ⊗ Pr k Mod QC ( X ) ( Pr k ) ≃ Mod QC ( X ) ⊗ QC ( X ) ( Pr k ) . \begin{aligned}
\mathcal{H}om_{\operatorname{Pr}_{k}}(\operatorname{Mod}_{\operatorname{QC}(\mathcal{X})}(\operatorname{Pr}_{k}),\operatorname{Mod}_{\operatorname{QC}(\mathcal{X})}(\operatorname{Pr}_{k})) \simeq\operatorname{Mod}_{\operatorname{QC}(\mathcal{X})}(\operatorname{Pr}_{k}) \otimes_{\operatorname{Pr}_{k}} \operatorname{Mod}_{\operatorname{QC}(\mathcal{X})}(\operatorname{Pr}_{k}) \\
&\simeq\operatorname{Mod}_{\operatorname{QC}(\mathcal{X})\otimes\operatorname{QC}(\mathcal{X})}(\operatorname{Pr}_{k}).
\end{aligned} H o m Pr k ( Mod QC ( X ) ( Pr k ) , Mod QC ( X ) ( Pr k )) ≃ Mod QC ( X ) ( Pr k ) ⊗ Pr k Mod QC ( X ) ( Pr k ) ≃ Mod QC ( X ) ⊗ QC ( X ) ( Pr k ) .
Since there exists an equivalence p 1 ⊠ p 2 : QC ( X ) ⊗ QC ( X ) → QC ( X × k X ) p_{1}\boxtimes p_{2}:\operatorname{QC}(\mathcal{X})\otimes\operatorname{QC}(\mathcal{X})\to\operatorname{QC}(\mathcal{X}\times_{k}\mathcal{X}) p 1 ⊠ p 2 : QC ( X ) ⊗ QC ( X ) → QC ( X × k X ) induced by the pullbacks of projections p 1 , p 2 : X × k X ⇉ X p_{1},p_{2}:\mathcal{X}\times_{k}\mathcal{X}\rightrightarrows\mathcal{X} p 1 , p 2 : X × k X ⇉ X (see [4 Theorem 4.7] ), we have the equivalence
Mod QC ( X ) ⊗ QC ( X ) ( Pr k ) ≃ Mod QC ( X × k X ) ( Pr k ) . \operatorname{Mod}_{\operatorname{QC}(\mathcal{X})\otimes\operatorname{QC}(\mathcal{X})}(\operatorname{Pr}_{k}) \simeq\operatorname{Mod}_{\operatorname{QC}(\mathcal{X}\times_{k}\mathcal{X})}(\operatorname{Pr}_{k}). Mod QC ( X ) ⊗ QC ( X ) ( Pr k ) ≃ Mod QC ( X × k X ) ( Pr k ) .
The identity functor Mod QC ( X ) ( Pr k ) → Mod QC ( X ) ( Pr k ) \operatorname{Mod}_{\operatorname{QC}(\mathcal{X})}(\operatorname{Pr}_{k})\to\operatorname{Mod}_{\operatorname{QC}(\mathcal{X})}(\operatorname{Pr}_{k}) Mod QC ( X ) ( Pr k ) → Mod QC ( X ) ( Pr k ) corresponds to QC ( X ) \operatorname{QC}(\mathcal{X}) QC ( X ) endowed with the diagonal QC ( X ) \operatorname{QC}(\mathcal{X}) QC ( X ) -QC ( X ) \operatorname{QC}(\mathcal{X}) QC ( X ) -bimodule structure. This bimodule is a QC ( X × k X ) \operatorname{QC}(\mathcal{X}\times_{k}\mathcal{X}) QC ( X × k X ) -module QC ( X ) \operatorname{QC}(\mathcal{X}) QC ( X ) induced by the symmetric monoidal pullback functor Δ ∗ : QC ( X × k X ) → QC ( X ) \Delta^{*}:\operatorname{QC}(\mathcal{X}\times_{k}\mathcal{X})\to\operatorname{QC}(\mathcal{X}) Δ ∗ : QC ( X × k X ) → QC ( X ) along the diagonal morphism Δ : X → X × k X \Delta:\mathcal{X}\to\mathcal{X}\times_{k}\mathcal{X} Δ : X → X × k X . Moreover, QC ( X ) \operatorname{QC}(\mathcal{X}) QC ( X ) is dualizable in Pr k \operatorname{Pr}_{k} Pr k whose dual is given by QC ( X ) \operatorname{QC}(\mathcal{X}) QC ( X ) itself. Indeed, according to [4 Corollary 4.8] QC ( X ) \operatorname{QC}(\mathcal{X}) QC ( X ) is self-dual in Mod QC ( X × k X ) ( Pr k ) \operatorname{Mod}_{\operatorname{QC}(\mathcal{X}\times_{k}\mathcal{X})}(\operatorname{Pr}_{k}) Mod QC ( X × k X ) ( Pr k ) . Therefore, we have
H o m Mod QC ( X × k X ) ( Pr k ) ( QC ( X ) , QC ( X ) ) ≃ QC ( X ) ⊗ QC ( X × k X ) QC ( X ) ≃ QC ( X × X × k X X ) \mathcal{H}om_{\operatorname{Mod}_{\operatorname{QC}(\mathcal{X}\times_{k}\mathcal{X})}(\operatorname{Pr}_{k})}(\operatorname{QC}(\mathcal{X}),\operatorname{QC}(\mathcal{X})) \simeq\operatorname{QC}(\mathcal{X})\otimes_{\operatorname{QC}(\mathcal{X}\times_{k}\mathcal{X})}\operatorname{QC}(\mathcal{X}) \simeq\operatorname{QC}(\mathcal{X}\mathop{\times}_{\mathcal{X}\times_{k}\mathcal{X}}\mathcal{X}) H o m Mod QC ( X × k X ) ( Pr k ) ( QC ( X ) , QC ( X )) ≃ QC ( X ) ⊗ QC ( X × k X ) QC ( X ) ≃ QC ( X × X × k X X )
in Pr k \operatorname{Pr}_{k} Pr k , see [4 Theorem 4.7] for the latter equivalence. Through the equivalence, the identity functor QC ( X ) → QC ( X ) \operatorname{QC}(\mathcal{X})\to\operatorname{QC}(\mathcal{X}) QC ( X ) → QC ( X ) corresponds to the O Δ ∈ QC ( L X ) \mathcal{O}_{\Delta}\in\operatorname{QC}(\mathcal{L}\mathcal{X}) O Δ ∈ QC ( LX ) , which is the pushforward of the structure sheaf O X ∈ QC ( X ) \mathcal{O}_{\mathcal{X}}\in\operatorname{QC}(\mathcal{X}) O X ∈ QC ( X ) along the diagonal morphism Δ L : X → X × X × k X X = L X \Delta_{\mathcal{L}}:\mathcal{X}\to\mathcal{X}\mathop{\times}_{\mathcal{X}\times_{k}\mathcal{X}}\mathcal{X}=\mathcal{L}\mathcal{X} Δ L : X → X × X × k X X = LX . Consequently, Hochschild cohomology is equivalent to H o m QC ( L X ) ( O Δ , O Δ ) ∈ Mod k \mathcal{H}om_{\operatorname{QC}(\mathcal{L}\mathcal{X})}(\mathcal{O}_{\Delta},\mathcal{O}_{\Delta})\in\operatorname{Mod}_{k} H o m QC ( LX ) ( O Δ , O Δ ) ∈ Mod k (the equivalence can be promoted to the level of Alg 1 ( Mod k ) \operatorname{Alg}_{1}(\operatorname{Mod}_{k}) Alg 1 ( Mod k ) ).
We show the second assertion. Under our assumption, Koszul duality [27 ] and [16 ] imply that there exists an equivalence
H o m QC ( L X ) ( O Δ , O Δ ) ≃ Γ ( X , Sym O X ∙ ( T X / k [ − 2 ] ) ) \mathcal{H}om_{\operatorname{QC}(\mathcal{L}\mathcal{X})}(\mathcal{O}_{\Delta},\mathcal{O}_{\Delta})\simeq\Gamma(\mathcal{X},\operatorname{Sym}^{\bullet}_{\mathcal{O}_{\mathcal{X}}}(\mathbb{T}_{\mathcal{X}/k}[-2])) H o m QC ( LX ) ( O Δ , O Δ ) ≃ Γ ( X , Sym O X ∙ ( T X / k [ − 2 ]))
in Mod k \operatorname{Mod}_{k} Mod k .
Higher module category
Suppose that P P P is any of Pr k \operatorname{Pr}_{k} Pr k or Pr k t + \operatorname{Pr}_{k}^{\mathrm{t}+} Pr k t + or Pr k t ± \operatorname{Pr}_{k}^{\mathrm{t}\pm} Pr k t ± . Let A A A be an object of CAlg ( P ) \operatorname{CAlg}(P) CAlg ( P ) . We will associate to A A A a kind of higher module category. We write 0 Mod A [ P ] = A {}^{0}\operatorname{Mod}_{A}[P]=A 0 Mod A [ P ] = A and 1 Mod A [ P ] = Mod A ( P ) {}^{1}\operatorname{Mod}_{A}[P]=\operatorname{Mod}_{A}(P) 1 Mod A [ P ] = Mod A ( P ) . We note that 0 Mod A ∈ P {}^{0}\operatorname{Mod}_{A}\in P 0 Mod A ∈ P and 1 Mod A ∈ 2 P {}^{1}\operatorname{Mod}_{A}\in{}^{2}P 1 Mod A ∈ 2 P . We define n Mod A [ P ] ∈ n + 1 P {}^{n}\operatorname{Mod}_{A}[P]\in{}^{n+1}P n Mod A [ P ] ∈ n + 1 P as Mod n − 1 Mod A [ P ] ( n P ) \operatorname{Mod}_{{}^{n-1}\operatorname{Mod}_{A}[P]}({}^{n}P) Mod n − 1 Mod A [ P ] ( n P ) . For ease of notation we often write n Mod A {}^{n}\operatorname{Mod}_{A} n Mod A for n Mod A [ P ] {}^{n}\operatorname{Mod}_{A}[P] n Mod A [ P ] if the setting is clear.
Proposition 6.11. If P P P is any of Pr k \operatorname{Pr}_{k} Pr k or Pr k t + \operatorname{Pr}_{k}^{\mathrm{t}+} Pr k t + or Pr k t ± \operatorname{Pr}_{k}^{\mathrm{t}\pm} Pr k t ± , then n Mod A [ P ] ∈ n + 1 P {}^{n}\operatorname{Mod}_{A}[P]\in{}^{n+1}P n Mod A [ P ] ∈ n + 1 P has Hochschild cohomology H H ∙ ( n Mod A / k ) ∈ Alg n + 2 ( Mod k ) \mathcal{HH}^{\bullet}({}^{n}\operatorname{Mod}_{A}/k)\in\operatorname{Alg}_{n+2}(\operatorname{Mod}_{k}) HH ∙ ( n Mod A / k ) ∈ Alg n + 2 ( Mod k ) .
We show this proposition step by step.
Lemma 6.12. Let i i i be a natural number. Let B B B be an object of CAlg ( i P ) \operatorname{CAlg}(iP) CAlg ( i P ) . Let m m m be a nonnegative integer. Then there exists an equivalence Mod B ( i P ) ⊗ S m ≃ Mod B ⊗ S m ( i P ) \operatorname{Mod}_{B}(iP)\otimes S^{m}\simeq\operatorname{Mod}_{B\otimes S^{m}}(iP) Mod B ( i P ) ⊗ S m ≃ Mod B ⊗ S m ( i P ) in CAlg ( i P ) \operatorname{CAlg}(iP) CAlg ( i P ) . Here Mod B ( i P ) ⊗ S m \operatorname{Mod}_{B}(iP)\otimes S^{m} Mod B ( i P ) ⊗ S m denotes the tensor with S m S^{m} S m in CAlg ( ( i + 1 ) P ) \operatorname{CAlg}((i+1)P) CAlg (( i + 1 ) P ) , and B ⊗ S m B\otimes S^{m} B ⊗ S m denotes the tensor with the sphere S m S^{m} S m in CAlg ( i P ) \operatorname{CAlg}(iP) CAlg ( i P ) .
Proof. The proof proceeds by induction on m m m . If m = 0 m=0 m = 0 , there are canonical equivalences
Mod B ( i P ) ⊗ S 0 ≃ Mod B ( i P ) ⊗ Mod B ( i P ) ≃ Mod B ⊗ B ( i P ) \operatorname{Mod}_{B}(iP)\otimes S^{0}\simeq\operatorname{Mod}_{B}(iP)\otimes\operatorname{Mod}_{B}(iP)\simeq\operatorname{Mod}_{B\otimes B}(iP) Mod B ( i P ) ⊗ S 0 ≃ Mod B ( i P ) ⊗ Mod B ( i P ) ≃ Mod B ⊗ B ( i P )
where the middle is the coproduct of Mod B ( i P ) \operatorname{Mod}_{B}(iP) Mod B ( i P ) in CAlg ( ( i + 1 ) P ) \operatorname{CAlg}((i+1)P) CAlg (( i + 1 ) P ) as well as the underlying object in ( i + 1 ) P (i+1)P ( i + 1 ) P is the tensor product. The right equivalence follows from [24 4.8.5.16] : The right map is defined by the base changes along 1 ⊗ B → B ⊗ B ← B ⊗ 1 1\otimes B\to B\otimes B\leftarrow B\otimes1 1 ⊗ B → B ⊗ B ← B ⊗ 1 . (1 is the unit algebra of i P iP i P ). Thus, the case of m = 0 m=0 m = 0 holds. Next we assume that Mod B ( i P ) ⊗ S m ≃ Mod B ⊗ S m ( i P ) \operatorname{Mod}_{B}(iP)\otimes S^{m}\simeq\operatorname{Mod}_{B\otimes S^{m}}(iP) Mod B ( i P ) ⊗ S m ≃ Mod B ⊗ S m ( i P ) . More precisely, there is a canonically defined functor Mod B ( i P ) ⊗ S m → Mod B ⊗ S m ( i P ) \operatorname{Mod}_{\mathcal{B}}(i\mathcal{P}) \otimes S^{m} \to\operatorname{Mod}_{\mathcal{B}\otimes S^{m}}(i\mathcal{P}) Mod B ( i P ) ⊗ S m → Mod B ⊗ S m ( i P ) , which is assumed to be an equivalence: If we have a canonically defined equivalence Mod B ( i P ) ⊗ S m − 1 ≃ Mod B ⊗ S m − 1 ( i P ) \operatorname{Mod}_{\mathcal{B}}(i\mathcal{P}) \otimes S^{m-1} \simeq\operatorname{Mod}_{\mathcal{B}\otimes S^{m-1}}(i\mathcal{P}) Mod B ( i P ) ⊗ S m − 1 ≃ Mod B ⊗ S m − 1 ( i P ) , then it gives rise to
Mod B ( i P ) ⊗ S m ≃ Mod B ( i P ) ⊗ Mod B ( i P ) ⊗ S m − 1 Mod B ( i P ) ≃ Mod B ( i P ) ⊗ Mod B ⊗ S m − 1 ( i P ) Mod B ( i P ) → Mod B ⊗ S m ( i P ) \begin{aligned}
\operatorname{Mod}_{\mathcal{B}}(i\mathcal{P}) \otimes S^{m} \simeq\operatorname{Mod}_{\mathcal{B}}(i\mathcal{P}) \otimes_{\operatorname{Mod}_{\mathcal{B}}(i\mathcal{P})\otimes S^{m-1}} \operatorname{Mod}_{\mathcal{B}}(i\mathcal{P}) \simeq\operatorname{Mod}_{\mathcal{B}}(i\mathcal{P}) \otimes_{\operatorname{Mod}_{\mathcal{B}\otimes S^{m-1}}(i\mathcal{P})} \operatorname{Mod}_{\mathcal{B}}(i\mathcal{P}) \\
&\to\operatorname{Mod}_{\mathcal{B}\otimes S^{m}}(i\mathcal{P})
\end{aligned} Mod B ( i P ) ⊗ S m ≃ Mod B ( i P ) ⊗ Mod B ( i P ) ⊗ S m − 1 Mod B ( i P ) ≃ Mod B ( i P ) ⊗ Mod B ⊗ S m − 1 ( i P ) Mod B ( i P ) → Mod B ⊗ S m ( i P )
where the last map is induced by the canonical morphism from the pushout of B ← B ⊗ S m − 1 → B \mathcal{B} \leftarrow\mathcal{B} \otimes S^{m-1} \to\mathcal{B} B ← B ⊗ S m − 1 → B to B ⊗ S m \mathcal{B} \otimes S^{m} B ⊗ S m . We will prove that the canonically defined morphism
Mod B ( i P ) ⊗ S m + 1 ≃ Mod B ( i P ) ⊗ Mod B ( i P ) ⊗ S m Mod B ( i P ) ≃ Mod B ( i P ) ⊗ Mod B ⊗ S m ( i P ) Mod B ( i P ) \operatorname{Mod}_{\mathcal{B}}(i\mathcal{P}) \otimes S^{m+1} \simeq\operatorname{Mod}_{\mathcal{B}}(i\mathcal{P}) \otimes_{\operatorname{Mod}_{\mathcal{B}}(i\mathcal{P})\otimes S^{m}} \operatorname{Mod}_{\mathcal{B}}(i\mathcal{P}) \simeq\operatorname{Mod}_{\mathcal{B}}(i\mathcal{P}) \otimes_{\operatorname{Mod}_{\mathcal{B}\otimes S^{m}}(i\mathcal{P})} \operatorname{Mod}_{\mathcal{B}}(i\mathcal{P}) Mod B ( i P ) ⊗ S m + 1 ≃ Mod B ( i P ) ⊗ Mod B ( i P ) ⊗ S m Mod B ( i P ) ≃ Mod B ( i P ) ⊗ Mod B ⊗ S m ( i P ) Mod B ( i P )
→ f Mod B ⊗ S m + 1 ( i P ) \xrightarrow{f} \operatorname{Mod}_{\mathcal{B}\otimes S^{m+1}}(i\mathcal{P}) f Mod B ⊗ S m + 1 ( i P )
is an equivalence. To this end, we will show that the arrow f f f is an equivalence. We regard B \mathcal{B} B as a diagonal B ⊗ S m \mathcal{B} \otimes S^{m} B ⊗ S m -module through B ⊗ S m → B ⊗ ∗ ≃ B \mathcal{B} \otimes S^{m} \to\mathcal{B} \otimes* \simeq\mathcal{B} B ⊗ S m → B ⊗ ∗ ≃ B induced by the contraction S m → ∗ S^{m} \to* S m → ∗ . We then see that f f f is decomposed as
Mod B ( i P ) ⊗ Mod B ⊗ S m ( i P ) Mod B ( i P ) ≃ Mod B ( Mod B ⊗ S m ( i P ) ) ⊗ Mod B ⊗ S m ( i P ) Mod B ( Mod B ⊗ S m ( i P ) ) ≃ Mod B ⊗ B ⊗ S m B ( Mod B ⊗ S m ( i P ) ) ≃ Mod B ⊗ B ⊗ S m B ( i P ) ≃ Mod B ⊗ ( ∗ ⊔ S m ∗ ) ( i P ) ≃ Mod B ⊗ S m + 1 ( i P ) \begin{aligned}
\operatorname{Mod}_{\mathcal{B}}(i\mathcal{P}) \otimes_{\operatorname{Mod}_{\mathcal{B}\otimes S^{m}}(i\mathcal{P})} \operatorname{Mod}_{\mathcal{B}}(i\mathcal{P})
&\simeq\operatorname{Mod}_{\mathcal{B}}\bigl(\operatorname{Mod}_{\mathcal{B}\otimes S^{m}}(i\mathcal{P})\bigr) \otimes_{\operatorname{Mod}_{\mathcal{B}\otimes S^{m}}(i\mathcal{P})} \operatorname{Mod}_{\mathcal{B}}\bigl(\operatorname{Mod}_{\mathcal{B}\otimes S^{m}}(i\mathcal{P})\bigr) \\
&\simeq\operatorname{Mod}_{\mathcal{B}\otimes\mathcal{B}\otimes S^{m}\mathcal{B}}\bigl(\operatorname{Mod}_{\mathcal{B}\otimes S^{m}}(i\mathcal{P})\bigr) \\
&\simeq\operatorname{Mod}_{\mathcal{B}\otimes\mathcal{B}\otimes S^{m}\mathcal{B}}(i\mathcal{P}) \\
&\simeq\operatorname{Mod}_{\mathcal{B}\otimes(*\sqcup_{S^{m}}*)}(i\mathcal{P}) \\
&\simeq\operatorname{Mod}_{\mathcal{B}\otimes S^{m+1}}(i\mathcal{P})
\end{aligned} Mod B ( i P ) ⊗ Mod B ⊗ S m ( i P ) Mod B ( i P ) ≃ Mod B ( Mod B ⊗ S m ( i P ) ) ⊗ Mod B ⊗ S m ( i P ) Mod B ( Mod B ⊗ S m ( i P ) ) ≃ Mod B ⊗ B ⊗ S m B ( Mod B ⊗ S m ( i P ) ) ≃ Mod B ⊗ B ⊗ S m B ( i P ) ≃ Mod B ⊗ ( ∗ ⊔ S m ∗ ) ( i P ) ≃ Mod B ⊗ S m + 1 ( i P )
where the first map and the third map are induced by the forgetful functor Mod B ⊗ S m ( i P ) → i P \operatorname{Mod}_{\mathcal{B}\otimes S^{m}}(i\mathcal{P}) \to i\mathcal{P} Mod B ⊗ S m ( i P ) → i P (see [24 7.1.3.1] ), and the second equivalence follows from the symmetric monoidal structure on Mod B ⊗ S m ( i P ) \operatorname{Mod}_{\mathcal{B}\otimes S^{m}}(i\mathcal{P}) Mod B ⊗ S m ( i P ) (see [24 4.8.5.16] ). The fourth equivalence follows from the fact that the pushout B ← B ⊗ S m → B \mathcal{B} \leftarrow\mathcal{B} \otimes S^{m} \to\mathcal{B} B ← B ⊗ S m → B in CAlg ( i P ) \operatorname{CAlg}(i\mathcal{P}) CAlg ( i P ) is naturally equivalent to the bar construction B ⊗ B ⊗ S m B \mathcal{B} \otimes_{\mathcal{B}\otimes S^{m}} \mathcal{B} B ⊗ B ⊗ S m B in i P i\mathcal{P} i P . Thus, f f f is an equivalence. This completes the proof. □ \square □
There exists an endomorphism algebra End ( n Mod A [ P ] ) ∈ Alg 1 ( ( n + 1 ) P ) \operatorname{End}(n\operatorname{Mod}_{\mathcal{A}}[\mathcal{P}]) \in\operatorname{Alg}_{1}((n+1)\mathcal{P}) End ( n Mod A [ P ]) ∈ Alg 1 (( n + 1 ) P ) whose underlying object Hom ( n Mod A [ P ] , n Mod A [ P ] ) \operatorname{Hom}(n\operatorname{Mod}_{\mathcal{A}}[\mathcal{P}],n\operatorname{Mod}_{\mathcal{A}}[\mathcal{P}]) Hom ( n Mod A [ P ] , n Mod A [ P ]) is equivalent to Mod ( n − 1 ) Mod A ( n P ) ⊗ Mod ( n − 1 ) Mod A ( n P ) \operatorname{Mod}_{(n-1)\operatorname{Mod}_{\mathcal{A}}}(n\mathcal{P}) \otimes\operatorname{Mod}_{(n-1)\operatorname{Mod}_{\mathcal{A}}}(n\mathcal{P}) Mod ( n − 1 ) Mod A ( n P ) ⊗ Mod ( n − 1 ) Mod A ( n P ) since A \mathcal{A} A is a commutative algebra object, and n Mod A [ P ] = Mod ( n − 1 ) Mod A [ P ] ( n P ) n\operatorname{Mod}_{\mathcal{A}}[\mathcal{P}] = \operatorname{Mod}_{(n-1)\operatorname{Mod}_{\mathcal{A}}[\mathcal{P}]}(n\mathcal{P}) n Mod A [ P ] = Mod ( n − 1 ) Mod A [ P ] ( n P ) is self-dual in ( n + 1 ) P (n+1)\mathcal{P} ( n + 1 ) P . Here ⊗ \otimes ⊗ means the tensor product in ( n + 1 ) P (n+1)\mathcal{P} ( n + 1 ) P . According to [24 4.8.5.16] , there exists a canonical equivalence
Mod ( n − 1 ) Mod A ( n P ) ⊗ Mod ( n − 1 ) Mod A ( n P ) ≃ Mod ( n − 1 ) Mod A ⊗ ( n − 1 ) Mod A ( n P ) \operatorname{Mod}_{(n-1)\operatorname{Mod}_{\mathcal{A}}}(n\mathcal{P}) \otimes\operatorname{Mod}_{(n-1)\operatorname{Mod}_{\mathcal{A}}}(n\mathcal{P}) \simeq\operatorname{Mod}_{(n-1)\operatorname{Mod}_{\mathcal{A}}\otimes(n-1)\operatorname{Mod}_{\mathcal{A}}}(n\mathcal{P}) Mod ( n − 1 ) Mod A ( n P ) ⊗ Mod ( n − 1 ) Mod A ( n P ) ≃ Mod ( n − 1 ) Mod A ⊗ ( n − 1 ) Mod A ( n P )
in ( n + 1 ) P (n+1)\mathcal{P} ( n + 1 ) P .
Lemma 6.13. Let ( n − 1 ) Mod A [ P ] ∈ Mod ( n − 1 ) Mod A ⊗ ( n − 1 ) Mod A ( n P ) (n-1)\operatorname{Mod}_{\mathcal{A}}[\mathcal{P}] \in\operatorname{Mod}_{(n-1)\operatorname{Mod}_{\mathcal{A}}\otimes(n-1)\operatorname{Mod}_{\mathcal{A}}}(n\mathcal{P}) ( n − 1 ) Mod A [ P ] ∈ Mod ( n − 1 ) Mod A ⊗ ( n − 1 ) Mod A ( n P ) be the ( n − 1 ) Mod A ⊗ ( n − 1 ) Mod A (n-1)\operatorname{Mod}_{\mathcal{A}}\otimes(n-1)\operatorname{Mod}_{\mathcal{A}} ( n − 1 ) Mod A ⊗ ( n − 1 ) Mod A -module induced by the multiplication map ( n − 1 ) Mod A ⊗ ( n − 1 ) Mod A → ( n − 1 ) Mod A (n-1)\operatorname{Mod}_{\mathcal{A}}\otimes(n-1)\operatorname{Mod}_{\mathcal{A}} \to(n-1)\operatorname{Mod}_{\mathcal{A}} ( n − 1 ) Mod A ⊗ ( n − 1 ) Mod A → ( n − 1 ) Mod A . This corresponds to the identity functor in End ( n Mod A ) \operatorname{End}(n\operatorname{Mod}_{\mathcal{A}}) End ( n Mod A ) . Let
K n : = Hom ( n − 1 ) Mod A ⊗ ( n − 1 ) Mod A ( ( n − 1 ) Mod A , ( n − 1 ) Mod A ) ∈ n P K_n := \operatorname{Hom}_{(n-1)\operatorname{Mod}_{\mathcal{A}}\otimes(n-1)\operatorname{Mod}_{\mathcal{A}}}\bigl((n-1)\operatorname{Mod}_{\mathcal{A}},(n-1)\operatorname{Mod}_{\mathcal{A}}\bigr) \in n\mathcal{P} K n := Hom ( n − 1 ) Mod A ⊗ ( n − 1 ) Mod A ( ( n − 1 ) Mod A , ( n − 1 ) Mod A ) ∈ n P be the Hom object from ( n − 1 ) Mod A ∈ Mod ( n − 1 ) Mod A ⊗ ( n − 1 ) Mod A ( n P ) (n-1)\operatorname{Mod}_{\mathcal{A}} \in\operatorname{Mod}_{(n-1)\operatorname{Mod}_{\mathcal{A}}\otimes(n-1)\operatorname{Mod}_{\mathcal{A}}}(n\mathcal{P}) ( n − 1 ) Mod A ∈ Mod ( n − 1 ) Mod A ⊗ ( n − 1 ) Mod A ( n P ) to itself, where the subscript ( n − 1 ) Mod A ⊗ ( n − 1 ) Mod A (n-1)\operatorname{Mod}_{\mathcal{A}}\otimes(n-1)\operatorname{Mod}_{\mathcal{A}} ( n − 1 ) Mod A ⊗ ( n − 1 ) Mod A indicates the ( n − 1 ) Mod A ⊗ ( n − 1 ) Mod A (n-1)\operatorname{Mod}_{\mathcal{A}}\otimes(n-1)\operatorname{Mod}_{\mathcal{A}} ( n − 1 ) Mod A ⊗ ( n − 1 ) Mod A -linear. Then there exists a canonical equivalence
K n ≃ Mod ( n − 2 ) Mod A ⊗ S 1 ( ( n − 1 ) P ) K_n \simeq\operatorname{Mod}_{(n-2)\operatorname{Mod}_{\mathcal{A}}\otimes S^{1}}((n-1)\mathcal{P}) K n ≃ Mod ( n − 2 ) Mod A ⊗ S 1 (( n − 1 ) P ) in n P n\mathcal{P} n P . Here ( n − 2 ) Mod A ⊗ S 1 (n-2)\operatorname{Mod}_{\mathcal{A}}\otimes S^{1} ( n − 2 ) Mod A ⊗ S 1 means the tensor with the sphere S 1 ∈ S S^{1} \in\mathcal{S} S 1 ∈ S in CAlg ( ( n − 1 ) P ) \operatorname{CAlg}((n-1)\mathcal{P}) CAlg (( n − 1 ) P ) .
Proof. We note that
( n − 1 ) Mod A = Mod ( n − 2 ) Mod A ( ( n − 1 ) P ) ≃ Mod ( n − 2 ) Mod A ( Mod ( n − 2 ) Mod A ⊗ ( n − 2 ) Mod A ( ( n − 1 ) P ) ) . (n-1)\operatorname{Mod}_{\mathcal{A}} = \operatorname{Mod}_{(n-2)\operatorname{Mod}_{\mathcal{A}}}((n-1)\mathcal{P}) \simeq\operatorname{Mod}_{(n-2)\operatorname{Mod}_{\mathcal{A}}}\bigl(\operatorname{Mod}_{(n-2)\operatorname{Mod}_{\mathcal{A}}\otimes(n-2)\operatorname{Mod}_{\mathcal{A}}}((n-1)\mathcal{P})\bigr). ( n − 1 ) Mod A = Mod ( n − 2 ) Mod A (( n − 1 ) P ) ≃ Mod ( n − 2 ) Mod A ( Mod ( n − 2 ) Mod A ⊗ ( n − 2 ) Mod A (( n − 1 ) P ) ) .
The right equivalence is obtained by regarding ( n − 2 ) Mod A (n-2)\operatorname{Mod}_{\mathcal{A}} ( n − 2 ) Mod A as the diagonal ( n − 2 ) Mod A ⊗ ( n − 2 ) Mod A (n-2)\operatorname{Mod}_{\mathcal{A}}\otimes(n-2)\operatorname{Mod}_{\mathcal{A}} ( n − 2 ) Mod A ⊗ ( n − 2 ) Mod A -module (cf. [24 7.1.3.1] ). We put
( n − 1 ) Mod A ⊗ 2 : = ( n − 1 ) Mod A ⊗ ( n − 1 ) Mod A ≃ Mod ( n − 2 ) Mod A ⊗ ( n − 2 ) Mod A ( ( n − 1 ) P ) . (n-1)\operatorname{Mod}_{\mathcal{A}}^{\otimes2} := (n-1)\operatorname{Mod}_{\mathcal{A}} \otimes(n-1)\operatorname{Mod}_{\mathcal{A}} \simeq\operatorname{Mod}_{(n-2)\operatorname{Mod}_{\mathcal{A}}\otimes(n-2)\operatorname{Mod}_{\mathcal{A}}}((n-1)\mathcal{P}). ( n − 1 ) Mod A ⊗ 2 := ( n − 1 ) Mod A ⊗ ( n − 1 ) Mod A ≃ Mod ( n − 2 ) Mod A ⊗ ( n − 2 ) Mod A (( n − 1 ) P ) .
Here Lemma 6.12 implies the right equivalence. By the canonical equivalence Mod ( n − 2 ) Mod A ( ( n − 1 ) Mod A ⊗ 2 ) ≃ Mod ( n − 2 ) Mod A ( ( n − 1 ) P ) \operatorname{Mod}_{(n-2)\operatorname{Mod}_{\mathcal{A}}}((n-1)\operatorname{Mod}_{\mathcal{A}}^{\otimes2}) \simeq\operatorname{Mod}_{(n-2)\operatorname{Mod}_{\mathcal{A}}}((n-1)\mathcal{P}) Mod ( n − 2 ) Mod A (( n − 1 ) Mod A ⊗ 2 ) ≃ Mod ( n − 2 ) Mod A (( n − 1 ) P ) (again [24 7.1.3.1] ) and the self-duality of $\operatorname{Mod}_{(n-2)\operatorname{Mod}_{\mathcal{A}}}((n1 ) Mod A ⊗ 2 ) 1)\operatorname{Mod}_{A}^{\otimes2}) 1 ) Mod A ⊗ 2 ) , there exist equivalences
K n = Hom ( n − 1 ) Mod A ⊗ 2 ( Mod ( n − 2 ) Mod A ( ( n − 1 ) P ) , Mod ( n − 2 ) Mod A ( ( n − 1 ) P ) ) ≃ Hom ( n − 1 ) Mod A ⊗ 2 ( Mod ( n − 2 ) Mod A ( ( n − 1 ) Mod A ⊗ 2 ) , Mod ( n − 2 ) Mod A ( ( n − 1 ) Mod A ⊗ 2 ) ) ≃ Mod ( n − 2 ) Mod A ( ( n − 1 ) Mod A ⊗ 2 ) ⊗ ( n − 1 ) Mod A ⊗ 2 Mod ( n − 2 ) Mod A ( ( n − 1 ) Mod A ⊗ 2 ) ≃ Mod ( n − 2 ) Mod A ⊗ ( n − 2 ) Mod A ( ( n − 1 ) Mod A ⊗ 2 ) ≃ Mod ( n − 2 ) Mod A ⊗ ( n − 2 ) Mod A ⊗ ( n − 2 ) Mod A ( n − 2 ) Mod A ( ( n − 1 ) P ) . \begin{aligned}
K_n=\operatorname{Hom}_{(n-1)\operatorname{Mod}_{A}^{\otimes2}}\left(\operatorname{Mod}_{(n-2)\operatorname{Mod}_{A}}((n-1)\mathcal{P}),\operatorname{Mod}_{(n-2)\operatorname{Mod}_{A}}((n-1)\mathcal{P})\right) \\
&\simeq\operatorname{Hom}_{(n-1)\operatorname{Mod}_{A}^{\otimes2}}\left(\operatorname{Mod}_{(n-2)\operatorname{Mod}_{A}}\left((n-1)\operatorname{Mod}_{A}^{\otimes2}\right),\operatorname{Mod}_{(n-2)\operatorname{Mod}_{A}}\left((n-1)\operatorname{Mod}_{A}^{\otimes2}\right)\right) \\
&\simeq\operatorname{Mod}_{(n-2)\operatorname{Mod}_{A}}\left((n-1)\operatorname{Mod}_{A}^{\otimes2}\right)\otimes_{(n-1)\operatorname{Mod}_{A}^{\otimes2}}\operatorname{Mod}_{(n-2)\operatorname{Mod}_{A}}\left((n-1)\operatorname{Mod}_{A}^{\otimes2}\right) \\
&\simeq\operatorname{Mod}_{(n-2)\operatorname{Mod}_{A}\otimes_{(n-2)\operatorname{Mod}_{A}}\left((n-1)\operatorname{Mod}_{A}^{\otimes2}\right)} \\
&\simeq\operatorname{Mod}_{(n-2)\operatorname{Mod}_{A}\otimes_{(n-2)\operatorname{Mod}_{A}\otimes_{(n-2)\operatorname{Mod}_{A}}(n-2)\operatorname{Mod}_{A}}\left((n-1)\mathcal{P}\right).}
\end{aligned} K n = Hom ( n − 1 ) Mod A ⊗ 2 ( Mod ( n − 2 ) Mod A (( n − 1 ) P ) , Mod ( n − 2 ) Mod A (( n − 1 ) P ) ) ≃ Hom ( n − 1 ) Mod A ⊗ 2 ( Mod ( n − 2 ) Mod A ( ( n − 1 ) Mod A ⊗ 2 ) , Mod ( n − 2 ) Mod A ( ( n − 1 ) Mod A ⊗ 2 ) ) ≃ Mod ( n − 2 ) Mod A ( ( n − 1 ) Mod A ⊗ 2 ) ⊗ ( n − 1 ) Mod A ⊗ 2 Mod ( n − 2 ) Mod A ( ( n − 1 ) Mod A ⊗ 2 ) ≃ Mod ( n − 2 ) Mod A ⊗ ( n − 2 ) Mod A ( ( n − 1 ) Mod A ⊗ 2 ) ≃ Mod ( n − 2 ) Mod A ⊗ ( n − 2 ) Mod A ⊗ ( n − 2 ) Mod A ( n − 2 ) Mod A ( ( n − 1 ) P ) .
We note that there exists an equivalence
( n − 2 ) Mod A ⊗ ( n − 2 ) Mod A ( n − 2 ) Mod A ⊗ ( n − 2 ) Mod A ( n − 2 ) Mod A ≃ ( n − 2 ) Mod A ⊗ S 1 (n-2)\operatorname{Mod}_{A}\otimes_{(n-2)\operatorname{Mod}_{A}}(n-2)\operatorname{Mod}_{A}\otimes_{(n-2)\operatorname{Mod}_{A}}(n-2)\operatorname{Mod}_{A}\simeq(n-2)\operatorname{Mod}_{A}\otimes S^{1} ( n − 2 ) Mod A ⊗ ( n − 2 ) Mod A ( n − 2 ) Mod A ⊗ ( n − 2 ) Mod A ( n − 2 ) Mod A ≃ ( n − 2 ) Mod A ⊗ S 1
in CAlg ( ( n − 1 ) P ) \operatorname{CAlg}((n-1)\mathcal{P}) CAlg (( n − 1 ) P ) , where the right-hand side means the tensor with the circle S 1 ∈ S S^{1}\in\mathcal{S} S 1 ∈ S in CAlg ( ( n − 1 ) P ) \operatorname{CAlg}((n-1)\mathcal{P}) CAlg (( n − 1 ) P ) . Thus, we have an equivalence
K n ≃ Mod ( n − 2 ) Mod A ⊗ S 1 ( ( n − 1 ) P ) K_n\simeq\operatorname{Mod}_{(n-2)\operatorname{Mod}_{A}\otimes S^{1}}((n-1)\mathcal{P}) K n ≃ Mod ( n − 2 ) Mod A ⊗ S 1 (( n − 1 ) P )
in n P n\mathcal{P} n P . □
We take the next step. Let
K n − 1 = Hom ( n − 2 ) Mod A ⊗ S 1 ( ( n − 2 ) Mod A , ( n − 2 ) Mod A ) ∈ ( n − 1 ) P K_{n-1}=\operatorname{Hom}_{(n-2)\operatorname{Mod}_{A}\otimes S^{1}}\left((n-2)\operatorname{Mod}_{A},(n-2)\operatorname{Mod}_{A}\right)\in(n-1)\mathcal{P} K n − 1 = Hom ( n − 2 ) Mod A ⊗ S 1 ( ( n − 2 ) Mod A , ( n − 2 ) Mod A ) ∈ ( n − 1 ) P
be the Hom object from ( n − 2 ) Mod A ∈ Mod ( n − 2 ) Mod A ⊗ S 1 ( ( n − 1 ) P ) (n-2)\operatorname{Mod}_{A}\in\operatorname{Mod}_{(n-2)\operatorname{Mod}_{A}\otimes S^{1}}((n-1)\mathcal{P}) ( n − 2 ) Mod A ∈ Mod ( n − 2 ) Mod A ⊗ S 1 (( n − 1 ) P ) to itself, where the subscript ( n − 2 ) Mod A ⊗ S 1 (n-2)\operatorname{Mod}_{A}\otimes S^{1} ( n − 2 ) Mod A ⊗ S 1 indicates the ( n − 2 ) Mod A ⊗ S 1 (n-2)\operatorname{Mod}_{A}\otimes S^{1} ( n − 2 ) Mod A ⊗ S 1 -linear. Using Lemma 6.12 , the same argument shows that there exists an equivalence
K n − 1 ≃ Mod ( n − 3 ) Mod A ⊗ ( n − 3 ) Mod A ⊗ S 1 ( n − 3 ) Mod A ( ( n − 2 ) P ) K_{n-1}\simeq\operatorname{Mod}_{(n-3)\operatorname{Mod}_{A}\otimes_{(n-3)\operatorname{Mod}_{A}\otimes S^{1}}(n-3)\operatorname{Mod}_{A}}((n-2)\mathcal{P}) K n − 1 ≃ Mod ( n − 3 ) Mod A ⊗ ( n − 3 ) Mod A ⊗ S 1 ( n − 3 ) Mod A (( n − 2 ) P )
in ( n − 1 ) P (n-1)\mathcal{P} ( n − 1 ) P . Here the contraction map S 1 → ∗ S^{1}\to* S 1 → ∗ to the contractible space ∗ * ∗ induces a morphism ( n − 3 ) Mod A ⊗ S 1 → ( n − 3 ) Mod A (n-3)\operatorname{Mod}_{A}\otimes S^{1}\to(n-3)\operatorname{Mod}_{A} ( n − 3 ) Mod A ⊗ S 1 → ( n − 3 ) Mod A in CAlg ( ( n − 1 ) P ) \operatorname{CAlg}((n-1)\mathcal{P}) CAlg (( n − 1 ) P ) , which exhibits ( n − 3 ) Mod A (n-3)\operatorname{Mod}_{A} ( n − 3 ) Mod A as a ( n − 3 ) Mod A ⊗ S 1 (n-3)\operatorname{Mod}_{A}\otimes S^{1} ( n − 3 ) Mod A ⊗ S 1 -module. Since the pushout ∗ ⊔ S 1 ∗ *\sqcup_{S^{1}}* ∗ ⊔ S 1 ∗ is S 2 S^{2} S 2 , we have ( n − 3 ) Mod A ⊗ ( n − 3 ) Mod A ⊗ S 1 ( n − 3 ) Mod A ≃ ( n − 3 ) Mod A ⊗ S 2 (n-3)\operatorname{Mod}_{A}\otimes_{(n-3)\operatorname{Mod}_{A}\otimes S^{1}}(n-3)\operatorname{Mod}_{A}\simeq(n-3)\operatorname{Mod}_{A}\otimes S^{2} ( n − 3 ) Mod A ⊗ ( n − 3 ) Mod A ⊗ S 1 ( n − 3 ) Mod A ≃ ( n − 3 ) Mod A ⊗ S 2 . Consequently, we have the following.
Lemma 6.14. There exists a canonical equivalence K n − 1 ≃ Mod ( n − 3 ) Mod A [ P ] ⊗ S 2 ( ( n − 2 ) P ) K_{n-1}\simeq\operatorname{Mod}_{(n-3)\operatorname{Mod}_{A}[P]\otimes S^{2}}((n-2)\mathcal{P}) K n − 1 ≃ Mod ( n − 3 ) Mod A [ P ] ⊗ S 2 (( n − 2 ) P ) in ( n − 1 ) P (n-1)\mathcal{P} ( n − 1 ) P .
Repeating this procedure, we have
Lemma 6.15. If we put K i + 1 = Hom i Mod A ⊗ S n − i − 1 ( i Mod A , i Mod A ) ∈ Alg n − i + 1 ( i P ) K_{i+1}=\operatorname{Hom}_{i\operatorname{Mod}_{A}\otimes S^{n-i-1}}(i\operatorname{Mod}_{A},i\operatorname{Mod}_{A})\in\operatorname{Alg}_{n-i+1}(i\mathcal{P}) K i + 1 = Hom i Mod A ⊗ S n − i − 1 ( i Mod A , i Mod A ) ∈ Alg n − i + 1 ( i P ) , there exists a canonical equivalence
K i + 1 ≃ Mod ( i − 1 ) Mod A ⊗ S n − i ( i P ) . K_{i+1}\simeq\operatorname{Mod}_{(i-1)\operatorname{Mod}_{A}\otimes S^{n-i}}(i\mathcal{P}). K i + 1 ≃ Mod ( i − 1 ) Mod A ⊗ S n − i ( i P ) .
We arrive at
K 2 = Hom Mod A ⊗ S n − 2 ( Mod A ( P ) , Mod A ( P ) ) ≃ Mod A ( P ) ⊗ Mod A ⊗ S n − 2 Mod A ( P ) ≃ Mod A ⊗ S n − 1 \begin{aligned}
K_2=\operatorname{Hom}_{\operatorname{Mod}_{A}\otimes S^{n-2}}\left(\operatorname{Mod}_{A}(\mathcal{P}),\operatorname{Mod}_{A}(\mathcal{P})\right)\simeq\operatorname{Mod}_{A}(\mathcal{P})\otimes_{\operatorname{Mod}_{A}\otimes S^{n-2}}\operatorname{Mod}_{A}(\mathcal{P}) \\
&\simeq\operatorname{Mod}_{A}\otimes S^{n-1}
\end{aligned} K 2 = Hom Mod A ⊗ S n − 2 ( Mod A ( P ) , Mod A ( P ) ) ≃ Mod A ( P ) ⊗ Mod A ⊗ S n − 2 Mod A ( P ) ≃ Mod A ⊗ S n − 1
Using Lemma 6.12 , we see K 2 ≃ Mod A ⊗ S n − 1 K_2\simeq\operatorname{Mod}_{A\otimes S^{n-1}} K 2 ≃ Mod A ⊗ S n − 1 .
Next, we let A Δ A_{\Delta} A Δ denote the A ⊗ S n − 1 A\otimes S^{n-1} A ⊗ S n − 1 -module A A A whose A ⊗ S n − 1 A\otimes S^{n-1} A ⊗ S n − 1 -module structure is determined by the symmetric monoidal functor A ⊗ S n − 1 → A A\otimes S^{n-1}\to A A ⊗ S n − 1 → A induced by S n − 1 → ∗ S^{n-1}\to* S n − 1 → ∗ . This diagonal module A Δ A_{\Delta} A Δ corresponds to the identity functor. Next, we will show
Lemma 6.16. Suppose that P \mathcal{P} P is either Pr k t + \operatorname{Pr}_{k}^{t+} Pr k t + or Pr k t ± \operatorname{Pr}_{k}^{t\pm} Pr k t ± . Then there exists a Hom object Hom A ⊗ S n − 1 ( A Δ , A Δ ) \operatorname{Hom}_{A\otimes S^{n-1}}(A_{\Delta},A_{\Delta}) Hom A ⊗ S n − 1 ( A Δ , A Δ ) in Pr k t + \operatorname{Pr}_{k}^{t+} Pr k t + (the subscript again indicates A ⊗ S n − 1 A\otimes S^{n-1} A ⊗ S n − 1 -linear). In particular, there exists an endomorphism algebra End A ⊗ S n − 1 ( A Δ ) \operatorname{End}_{A\otimes S^{n-1}}(A_{\Delta}) End A ⊗ S n − 1 ( A Δ ) in Alg n + 1 ( Pr k t + ) \operatorname{Alg}_{n+1}(\operatorname{Pr}_{k}^{t+}) Alg n + 1 ( Pr k t + ) .
Suppose that P \mathcal{P} P is Pr k \operatorname{Pr}_{k} Pr k . Then there exists a Hom object Hom A ⊗ S n − 1 ( A Δ , A Δ ) \operatorname{Hom}_{A\otimes S^{n-1}}(A_{\Delta},A_{\Delta}) Hom A ⊗ S n − 1 ( A Δ , A Δ ) in Pr k \operatorname{Pr}_{k} Pr k , and there exists an endomorphism algebra End A ⊗ S n − 1 ( A Δ ) \operatorname{End}_{A\otimes S^{n-1}}(A_{\Delta}) End A ⊗ S n − 1 ( A Δ ) in Alg n + 1 ( Pr k ) \operatorname{Alg}_{n+1}(\operatorname{Pr}_{k}) Alg n + 1 ( Pr k ) .
Proof. As in the case of t t t -structured ( ∞ , 2 ) (\infty,2) ( ∞ , 2 ) -categories (see § 6.2 ), from the free-forget adjunction of modules
Pr k t + ⇄ ( − ) ⊗ ( A ⊗ S n − 1 ) f o r g e t Mod A ⊗ S n − 1 ( Pr k t + ) \operatorname{Pr}_{k}^{t+}\underset{\mathrm{forget}}{\overset{(-)\otimes(A\otimes S^{n-1})}{\rightleftarrows}}\operatorname{Mod}_{A\otimes S^{n-1}}(\operatorname{Pr}_{k}^{t+}) Pr k t + forget ⇄ ( − ) ⊗ ( A ⊗ S n − 1 ) Mod A ⊗ S n − 1 ( Pr k t + )
we obtain a simplicial object R ∙ \mathcal{R}_{\bullet} R ∙ in Mod A ⊗ S n − 1 ( Pr k t + ) \operatorname{Mod}_{\mathcal{A}\otimes S^{n-1}}(\operatorname{Pr}_{k}^{t+}) Mod A ⊗ S n − 1 ( Pr k t + ) with an augmentation ϵ : R ∙ → A Δ \epsilon:\mathcal{R}_{\bullet}\to\mathcal{A}_{\Delta} ϵ : R ∙ → A Δ such that each m m m -th term is a free module E ( m ) ⊗ ( A ⊗ S n − 1 ) \mathcal{E}(m)\otimes(\mathcal{A}\otimes S^{n-1}) E ( m ) ⊗ ( A ⊗ S n − 1 ) , and ϵ \epsilon ϵ exhibits A Δ \mathcal{A}_{\Delta} A Δ as a colimit. We then have
Hom A ⊗ S n − 1 ( A Δ , A Δ ) ≃ lim [ m ] ∈ Δ Hom ( E ( m ) , A ) \operatorname{Hom}_{\mathcal{A}\otimes S^{n-1}}(\mathcal{A}_{\Delta},\mathcal{A}_{\Delta}) \simeq\lim_{[m]\in\Delta}\operatorname{Hom}(\mathcal{E}(m),\mathcal{A}) Hom A ⊗ S n − 1 ( A Δ , A Δ ) ≃ [ m ] ∈ Δ lim Hom ( E ( m ) , A )
where each Hom ( E ( m ) , A ) \operatorname{Hom}(\mathcal{E}(m),\mathcal{A}) Hom ( E ( m ) , A ) is a Hom object, defined as an object of Pr k t + \operatorname{Pr}_{k}^{t+} Pr k t + , from E ( m ) \mathcal{E}(m) E ( m ) to A \mathcal{A} A . Since Pr k t + \operatorname{Pr}_{k}^{t+} Pr k t + and Pr k t ± \operatorname{Pr}_{k}^{t\pm} Pr k t ± admit Hom objects defined in Pr k t + \operatorname{Pr}_{k}^{t+} Pr k t + and have small limits, it follows that Hom A ⊗ S n − 1 ( A Δ , A Δ ) \operatorname{Hom}_{\mathcal{A}\otimes S^{n-1}}(\mathcal{A}_{\Delta},\mathcal{A}_{\Delta}) Hom A ⊗ S n − 1 ( A Δ , A Δ ) exists in Pr k t + \operatorname{Pr}_{k}^{t+} Pr k t + .
This argument also works when P = Pr k \mathrm{P}=\operatorname{Pr}_{k} P = Pr k .
Corollary 6.17
Corollary 6.17. If P \mathrm{P} P is either Pr k t + \operatorname{Pr}_{k}^{t+} Pr k t + or Pr k \operatorname{Pr}_{k} Pr k , End ( n Mod A [ P ] ) \operatorname{End}(n\operatorname{Mod}_{\mathcal{A}}[\mathrm{P}]) End ( n Mod A [ P ]) has a good endomorphism algebra of a unit.
Proof. Apply Lemma 6.15 and Lemma 6.16.
Proof of Proposition 6.11. We first suppose that P \mathrm{P} P is either Pr k t + \operatorname{Pr}_{k}^{t+} Pr k t + or Pr k t ± \operatorname{Pr}_{k}^{t\pm} Pr k t ± . As in Section 6.2, applying Alg n + 1 ( − ) \operatorname{Alg}_{n+1}(-) Alg n + 1 ( − ) to the right adjoint E Pr k t + : ( Pr k t + ) Mod k / → Alg 1 ( Mod k ) E_{\operatorname{Pr}_{k}^{t+}}:(\operatorname{Pr}_{k}^{t+})_{\operatorname{Mod}_{k}/}\to\operatorname{Alg}_{1}(\operatorname{Mod}_{k}) E Pr k t + : ( Pr k t + ) Mod k / → Alg 1 ( Mod k ) , we have Alg n + 1 ( E Pr k t + ) : Alg n + 1 ( Pr k t + ) → Alg n + 2 ( Mod k ) \operatorname{Alg}_{n+1}(E_{\operatorname{Pr}_{k}^{t+}}):\operatorname{Alg}_{n+1}(\operatorname{Pr}_{k}^{t+})\to\operatorname{Alg}_{n+2}(\operatorname{Mod}_{k}) Alg n + 1 ( E Pr k t + ) : Alg n + 1 ( Pr k t + ) → Alg n + 2 ( Mod k ) . The Hochschild cohomology H H ∙ ( n Mod A [ P ] / k ) ∈ Alg n + 2 ( Mod k ) \mathcal{HH}^{\bullet}(n\operatorname{Mod}_{\mathcal{A}}[\mathrm{P}]/k)\in\operatorname{Alg}_{n+2}(\operatorname{Mod}_{k}) HH ∙ ( n Mod A [ P ] / k ) ∈ Alg n + 2 ( Mod k ) is constructed as the image of End A ⊗ S n − 1 ( A Δ ) \operatorname{End}_{\mathcal{A}\otimes S^{n-1}}(\mathcal{A}_{\Delta}) End A ⊗ S n − 1 ( A Δ ) under Alg n + 2 ( E Pr k t + ) \operatorname{Alg}_{n+2}(E_{\operatorname{Pr}_{k}^{t+}}) Alg n + 2 ( E Pr k t + ) . By construction, this has the required property (when P = Pr k t ± \mathrm{P}=\operatorname{Pr}_{k}^{t\pm} P = Pr k t ± , note that Pr k t ± \operatorname{Pr}_{k}^{t\pm} Pr k t ± is a localization of Pr k t + \operatorname{Pr}_{k}^{t+} Pr k t + ).
Next, suppose that P = Pr k \mathrm{P}=\operatorname{Pr}_{k} P = Pr k . Using the adjoint pair θ 0 Pr k , 0 : Alg 1 ( Mod k ) ⇄ ( Pr k ) Mod k / : E \theta_{0\operatorname{Pr}_{k},0}:\operatorname{Alg}_{1}(\operatorname{Mod}_{k})\rightleftarrows(\operatorname{Pr}_{k})_{\operatorname{Mod}_{k}/}:E θ 0 Pr k , 0 : Alg 1 ( Mod k ) ⇄ ( Pr k ) Mod k / : E instead of ( θ 0 Pr k t + , 0 , E Pr k t + ) (\theta_{0\operatorname{Pr}_{k}^{t+},0},E_{\operatorname{Pr}_{k}^{t+}}) ( θ 0 Pr k t + , 0 , E Pr k t + ) , one can construct Hochschild cohomology in the same way as above.
Proposition 6.18
Proposition 6.18. The following holds.
For A ∈ CAlg ( Pr k t + ) \mathcal{A}\in\operatorname{CAlg}(\operatorname{Pr}_{k}^{t+}) A ∈ CAlg ( Pr k t + ) , we let A \mathcal{A} A denote the image of A \mathcal{A} A under the forgetful functor CAlg ( Pr k t + ) → CAlg ( Pr k ) \operatorname{CAlg}(\operatorname{Pr}_{k}^{t+})\to\operatorname{CAlg}(\operatorname{Pr}_{k}) CAlg ( Pr k t + ) → CAlg ( Pr k ) . We obtain n Mod A [ Pr k ] ∈ ( n + 1 ) Pr k n\operatorname{Mod}_{\mathcal{A}}[\operatorname{Pr}_{k}]\in(n+1)\operatorname{Pr}_{k} n Mod A [ Pr k ] ∈ ( n + 1 ) Pr k . We note that n Mod A [ Pr k ] n\operatorname{Mod}_{\mathcal{A}}[\operatorname{Pr}_{k}] n Mod A [ Pr k ] is canonically equivalent to n Mod A [ Pr k t + ] ⊗ n Pr k t + n Pr k n\operatorname{Mod}_{\mathcal{A}}[\operatorname{Pr}_{k}^{t+}]\otimes_{n\operatorname{Pr}_{k}^{t+}}n\operatorname{Pr}_{k} n Mod A [ Pr k t + ] ⊗ n Pr k t + n Pr k , that is, the base change along ( n + 1 ) Pr k t + → ( n + 1 ) Pr k (n+1)\operatorname{Pr}_{k}^{t+}\to(n+1)\operatorname{Pr}_{k} ( n + 1 ) Pr k t + → ( n + 1 ) Pr k . Then there exists a canonical equivalence H H ∙ ( n Mod A [ Pr k t + ] / k ) ≃ H H ∙ ( n Mod A [ Pr k ] / k ) \mathcal{HH}^{\bullet}(n\operatorname{Mod}_{\mathcal{A}}[\operatorname{Pr}_{k}^{t+}]/k)\simeq\mathcal{HH}^{\bullet}(n\operatorname{Mod}_{\mathcal{A}}[\operatorname{Pr}_{k}]/k) HH ∙ ( n Mod A [ Pr k t + ] / k ) ≃ HH ∙ ( n Mod A [ Pr k ] / k ) in Alg n + 2 ( Mod k ) \operatorname{Alg}_{n+2}(\operatorname{Mod}_{k}) Alg n + 2 ( Mod k ) .
Let A \mathcal{A} A be an object of CAlg ( Pr k t + ) \operatorname{CAlg}(\operatorname{Pr}_{k}^{t+}) CAlg ( Pr k t + ) . Suppose that A = A ^ \mathcal{A}=\widehat{\mathcal{A}} A = A belongs to Pr k t ± \operatorname{Pr}_{k}^{t\pm} Pr k t ± (when we regard A \mathcal{A} A as an object of Pr k t ± \operatorname{Pr}_{k}^{t\pm} Pr k t ± , we write A ^ \widehat{\mathcal{A}} A for A \mathcal{A} A ). Consider n Mod A [ Pr k t + ] ∈ ( n + 1 ) Pr k t + n\operatorname{Mod}_{\mathcal{A}}[\operatorname{Pr}_{k}^{t+}]\in(n+1)\operatorname{Pr}_{k}^{t+} n Mod A [ Pr k t + ] ∈ ( n + 1 ) Pr k t + and n Mod A ^ [ Pr k t ± ] ∈ ( n + 1 ) Pr k t ± n\operatorname{Mod}_{\widehat{\mathcal{A}}}[\operatorname{Pr}_{k}^{t\pm}]\in(n+1)\operatorname{Pr}_{k}^{t\pm} n Mod A [ Pr k t ± ] ∈ ( n + 1 ) Pr k t ± . We then consider Hochschild cohomology H H ∙ ( n Mod A [ Pr k t + ] / k ) \mathcal{HH}^{\bullet}(n\operatorname{Mod}_{\mathcal{A}}[\operatorname{Pr}_{k}^{t+}]/k) HH ∙ ( n Mod A [ Pr k t + ] / k ) , obtained from n Mod A [ Pr k t + ] ∈ ( n + 1 ) Pr k t + n\operatorname{Mod}_{\mathcal{A}}[\operatorname{Pr}_{k}^{t+}]\in(n+1)\operatorname{Pr}_{k}^{t+} n Mod A [ Pr k t + ] ∈ ( n + 1 ) Pr k t + , and Hochschild cohomology H H ∙ ( n Mod A ^ [ Pr k t ± ] / k ) \mathcal{HH}^{\bullet}(n\operatorname{Mod}_{\widehat{\mathcal{A}}}[\operatorname{Pr}_{k}^{t\pm}]/k) HH ∙ ( n Mod A [ Pr k t ± ] / k ) obtained from n Mod A ^ [ Pr k t ± ] ∈ ( n + 1 ) Pr k t ± n\operatorname{Mod}_{\widehat{\mathcal{A}}}[\operatorname{Pr}_{k}^{t\pm}]\in(n+1)\operatorname{Pr}_{k}^{t\pm} n Mod A [ Pr k t ± ] ∈ ( n + 1 ) Pr k t ± . Then there exists a canonical equivalence H H ∙ ( n Mod A [ Pr k t + ] / k ) ≃ H H ∙ ( n Mod A ^ [ Pr k t ± ] / k ) \mathcal{HH}^{\bullet}(n\operatorname{Mod}_{\mathcal{A}}[\operatorname{Pr}_{k}^{t+}]/k)\simeq\mathcal{HH}^{\bullet}(n\operatorname{Mod}_{\widehat{\mathcal{A}}}[\operatorname{Pr}_{k}^{t\pm}]/k) HH ∙ ( n Mod A [ Pr k t + ] / k ) ≃ HH ∙ ( n Mod A [ Pr k t ± ] / k ) in Alg n + 2 ( Mod k ) \operatorname{Alg}_{n+2}(\operatorname{Mod}_{k}) Alg n + 2 ( Mod k ) .
Proof. The proof is similar to Proposition 6.9. We prove (1). Note first that End ( n Mod A [ Pr k t + ] ) ⊗ n Pr k t + n Pr k ≃ End ( n Mod A [ Pr k ] ) \operatorname{End}(n\operatorname{Mod}_{\mathcal{A}}[\operatorname{Pr}_{k}^{t+}])\otimes_{n\operatorname{Pr}_{k}^{t+}}n\operatorname{Pr}_{k}\simeq\operatorname{End}(n\operatorname{Mod}_{\mathcal{A}}[\operatorname{Pr}_{k}]) End ( n Mod A [ Pr k t + ]) ⊗ n Pr k t + n Pr k ≃ End ( n Mod A [ Pr k ]) . The algebraization functor n\operatorname{Pr}_{k}^{t+}_{\mathcal{HH}^{\bullet}(n\operatorname{Mod}_{\mathcal{A}}[\operatorname{Pr}_{k}^{t+}]/k)}\to\operatorname{End}(n\operatorname{Mod}_{\mathcal{A}}[\operatorname{Pr}_{k}^{t+}]) (in Alg 1 ( ( n + 1 ) Pr k t + ) \operatorname{Alg}_{1}((n+1)\operatorname{Pr}_{k}^{t+}) Alg 1 (( n + 1 ) Pr k t + ) ) and the algebraization functor n Pr k H H ∙ ( n Mod A [ Pr k ] / k ) → End ( n Mod A [ Pr k ] ) n\operatorname{Pr}_{k}^{\mathcal{HH}^{\bullet}(n\operatorname{Mod}_{\mathcal{A}}[\operatorname{Pr}_{k}]/k)}\to\operatorname{End}(n\operatorname{Mod}_{\mathcal{A}}[\operatorname{Pr}_{k}]) n Pr k HH ∙ ( n Mod A [ Pr k ] / k ) → End ( n Mod A [ Pr k ]) (in Alg 1 ( ( n + 1 ) Pr k ) ) \operatorname{Alg}_{1}((n+1)\operatorname{Pr}_{k})) Alg 1 (( n + 1 ) Pr k )) induce
n\operatorname{Pr}_{k}^{\mathcal{HH}^{\bullet}(n\operatorname{Mod}_{\mathcal{A}}[\operatorname{Pr}_{k}^{t+}]/k)}\simeq n\operatorname{Pr}_{k}^{t+}_{\mathcal{HH}^{\bullet}(n\operatorname{Mod}_{\mathcal{A}}[\operatorname{Pr}_{k}^{t+}]/k)}\otimes_{n\operatorname{Pr}_{k}^{t+}}n\operatorname{Pr}_{k}\longrightarrow n\operatorname{Pr}_{k}^{\mathcal{HH}^{\bullet}(n\operatorname{Mod}_{\mathcal{A}}[\operatorname{Pr}_{k}]/k)}
⟶ End ( n Mod A [ Pr k ] ) . \longrightarrow\operatorname{End}(n\operatorname{Mod}_{\mathcal{A}}[\operatorname{Pr}_{k}]). ⟶ End ( n Mod A [ Pr k ]) .
Then the composite n Pr k H H ∙ ( n Mod A [ Pr k t + ] / k ) → n Pr k H H ∙ ( n Mod A [ Pr k ] / k ) n\operatorname{Pr}_{k}^{\mathcal{HH}^{\bullet}(n\operatorname{Mod}_{\mathcal{A}}[\operatorname{Pr}_{k}^{t+}]/k)}\to n\operatorname{Pr}_{k}^{\mathcal{HH}^{\bullet}(n\operatorname{Mod}_{\mathcal{A}}[\operatorname{Pr}_{k}]/k)} n Pr k HH ∙ ( n Mod A [ Pr k t + ] / k ) → n Pr k HH ∙ ( n Mod A [ Pr k ] / k ) gives rise to a canonical morphism H H ∙ ( n Mod A [ Pr k t + ] / k ) → H H ∙ ( n Mod A [ Pr k ] / k ) \mathcal{HH}^{\bullet}(n\operatorname{Mod}_{\mathcal{A}}[\operatorname{Pr}_{k}^{t+}]/k)\to\mathcal{HH}^{\bullet}(n\operatorname{Mod}_{\mathcal{A}}[\operatorname{Pr}_{k}]/k) HH ∙ ( n Mod A [ Pr k t + ] / k ) → HH ∙ ( n Mod A [ Pr k ] / k ) in Alg n + 2 ( Mod k ) \operatorname{Alg}_{n+2}(\operatorname{Mod}_{k}) Alg n + 2 ( Mod k ) . As in Proposition 6.9, we are reduced to showing that
π : Hom A ⊗ S n − 1 ( A Δ , A Δ ) → Hom A ⊗ S n − 1 ( A Δ , A Δ ) \pi:\operatorname{Hom}_{\mathcal{A}\otimes S^{n-1}}(\mathcal{A}_{\Delta},\mathcal{A}_{\Delta})\to\operatorname{Hom}_{\mathcal{A}\otimes S^{n-1}}(\mathcal{A}_{\Delta},\mathcal{A}_{\Delta}) π : Hom A ⊗ S n − 1 ( A Δ , A Δ ) → Hom A ⊗ S n − 1 ( A Δ , A Δ )
induces an equivalence of endomorphism algebras of the identity functors. The right-hand side denotes the Hom object in Pr k \operatorname{Pr}_{k} Pr k , which is defined in the same way as the left-hand side (the subscript A ⊗ S n − 1 ∈ CAlg ( Pr k ) \mathcal{A}\otimes S^{n-1}\in\operatorname{CAlg}(\operatorname{Pr}_{k}) A ⊗ S n − 1 ∈ CAlg ( Pr k ) indicates the A ⊗ S n − 1 \mathcal{A}\otimes S^{n-1} A ⊗ S n − 1 -linear). By the proof of Lemma 6.16, π \pi π can be identified with
f : lim [ m ] ∈ Δ Hom ( E ( m ) , A ) → lim [ m ] ∈ Δ Hom ( E ( m ) , A ) (3) f:\lim_{[m]\in\Delta}\operatorname{Hom}(\mathcal{E}(m),\mathcal{A})\to\lim_{[m]\in\Delta}\operatorname{Hom}(\mathcal{E}(m),\mathcal{A}) \tag*{(3)} f : [ m ] ∈ Δ lim Hom ( E ( m ) , A ) → [ m ] ∈ Δ lim Hom ( E ( m ) , A ) (3)
where each E ( m ) \mathcal{E}(m) E ( m ) is the image of E ( m ) \mathcal{E}(m) E ( m ) in Pr k \operatorname{Pr}_{k} Pr k (the left-hand side is the Hom object in Pr k t + \operatorname{Pr}_{k}^{t+} Pr k t + discarding the t t t -structure, and the right-hand is the Hom object in Pr k \operatorname{Pr}_{k} Pr k ). By the same argument as in Proposition 6.9, we see that f f f induces an equivalence of endomorphism algebras of the identity functors.
The proof of (2) is similar to that of Proposition 6.9 (2). The same argument applies to this setting.
Higher categorified sheaves
Let X \mathcal{X} X be a perfect stack over k k k . We will consider a higher version of § 6.3 . We take A \mathcal{A} A to be QC ( X ) ∈ CAlg ( Pr k t ± ) \operatorname{QC}(\mathcal{X}) \in\operatorname{CAlg}(\operatorname{Pr}_{k}^{t\pm}) QC ( X ) ∈ CAlg ( Pr k t ± ) (cf. § 6.3 ). We consider n Mod QC ( X ) [ Pr k t ± ] ∈ ( n + 1 ) Pr k t ± {}_{n}\operatorname{Mod}_{\operatorname{QC}(\mathcal{X})}[\operatorname{Pr}_{k}^{t\pm}] \in(n+1)\operatorname{Pr}_{k}^{t\pm} n Mod QC ( X ) [ Pr k t ± ] ∈ ( n + 1 ) Pr k t ± and its Hochschild cohomology (see Proposition 6.11). The underlying stable ( ∞ , n + 1 ) (\infty,n+1) ( ∞ , n + 1 ) -category of n Mod QC ( X ) [ Pr k t ± ] {}_{n}\operatorname{Mod}_{\operatorname{QC}(\mathcal{X})}[\operatorname{Pr}_{k}^{t\pm}] n Mod QC ( X ) [ Pr k t ± ] is n Mod QC ( X ) [ Pr k ] ∈ ( n + 1 ) Pr k {}_{n}\operatorname{Mod}_{\operatorname{QC}(\mathcal{X})}[\operatorname{Pr}_{k}] \in(n+1)\operatorname{Pr}_{k} n Mod QC ( X ) [ Pr k ] ∈ ( n + 1 ) Pr k , which we take as a version of higher categorified sheaves on X \mathcal{X} X . Here we will not discuss the affineness problem, which relates it to the sheaf Aff / X op → ( n + 1 ) Pr k \operatorname{Aff}^{\operatorname{op}}_{/\mathcal{X}} \to(n+1)\operatorname{Pr}_{k} Aff / X op → ( n + 1 ) Pr k given by U ↦ n Pr R ( U ) U \mapsto{}_{n}\operatorname{Pr}_{R(U)} U ↦ n Pr R ( U ) .
The Hochschild cohomology H H ∙ ( n Mod QC ( X ) [ Pr k t ± ] / k ) ∈ Alg n + 2 ( Mod k ) \mathcal{HH}^{\bullet}({}_{n}\operatorname{Mod}_{\operatorname{QC}(\mathcal{X})}[\operatorname{Pr}_{k}^{t\pm}]/k) \in\operatorname{Alg}_{n+2}(\operatorname{Mod}_{k}) HH ∙ ( n Mod QC ( X ) [ Pr k t ± ] / k ) ∈ Alg n + 2 ( Mod k ) is defined to be the image of End QC ( X ) ⊗ S n − 1 ( QC ( X ) ) \operatorname{End}_{\operatorname{QC}(\mathcal{X})^{\otimes S^{n-1}}}(\operatorname{QC}(\mathcal{X})) End QC ( X ) ⊗ S n − 1 ( QC ( X )) under Alg n + 2 ( E Pr k t + + ) \operatorname{Alg}_{n+2}(E^{+}_{\operatorname{Pr}_{k}^{t+}}) Alg n + 2 ( E Pr k t + + ) (cf. the proof of Proposition 6.11). According to Proposition 6.18, H H ∙ ( n Mod QC ( X ) [ Pr k t ± ] / k ) \mathcal{HH}^{\bullet}({}_{n}\operatorname{Mod}_{\operatorname{QC}(\mathcal{X})}[\operatorname{Pr}_{k}^{t\pm}]/k) HH ∙ ( n Mod QC ( X ) [ Pr k t ± ] / k ) is equivalent to Hochschild cohomology H H ∙ ( n Mod QC ( X ) [ Pr k ] / k ) \mathcal{HH}^{\bullet}({}_{n}\operatorname{Mod}_{\operatorname{QC}(\mathcal{X})}[\operatorname{Pr}_{k}]/k) HH ∙ ( n Mod QC ( X ) [ Pr k ] / k ) of n Mod QC ( X ) [ Pr k ] ∈ ( n + 1 ) Pr k {}_{n}\operatorname{Mod}_{\operatorname{QC}(\mathcal{X})}[\operatorname{Pr}_{k}] \in(n+1)\operatorname{Pr}_{k} n Mod QC ( X ) [ Pr k ] ∈ ( n + 1 ) Pr k when we regard QC ( X ) \operatorname{QC}(\mathcal{X}) QC ( X ) as an object of CAlg ( Pr k ) \operatorname{CAlg}(\operatorname{Pr}_{k}) CAlg ( Pr k ) .
Let m m m be a nonnegative integer. Let Map ( S m , X ) \operatorname{Map}(S^{m},\mathcal{X}) Map ( S m , X ) be the functor CAlg k conn → S \operatorname{CAlg}_{k}^{\operatorname{conn}} \to\mathcal{S} CAlg k conn → S , defined by R ↦ Map ( S m , X ( R ) ) R \mapsto\operatorname{Map}(S^{m},\mathcal{X}(R)) R ↦ Map ( S m , X ( R )) , where X ( R ) \mathcal{X}(R) X ( R ) is the space of R R R -valued points, and the sphere S m S^{m} S m is regarded as an object of S \mathcal{S} S . Here CAlg k conn \operatorname{CAlg}_{k}^{\operatorname{conn}} CAlg k conn is the full subcategory of CAlg k \operatorname{CAlg}_{k} CAlg k , spanned by connective spectra, and S \mathcal{S} S is the ∞ \infty ∞ -category of small spaces. This functor is represented by a perfect stack over k k k , so that we think of Map ( S m , X ) \operatorname{Map}(S^{m},\mathcal{X}) Map ( S m , X ) as a derived stack. Indeed, this stack is constructed as an iterated fiber product. When m = 1 m=1 m = 1 , Map ( S m , X ) \operatorname{Map}(S^{m},\mathcal{X}) Map ( S m , X ) is the loop space L X L\mathcal{X} L X . This derived stack Map ( S m , X ) \operatorname{Map}(S^{m},\mathcal{X}) Map ( S m , X ) admits a closed immersion ι m : X → Map ( S m , X ) \iota_m:\mathcal{X}\to\operatorname{Map}(S^{m},\mathcal{X}) ι m : X → Map ( S m , X ) , given by X ( R ) → Map ( S m , X ( R ) ) \mathcal{X}(R)\to\operatorname{Map}(S^{m},\mathcal{X}(R)) X ( R ) → Map ( S m , X ( R )) determined by the assignment x ↦ c x x\mapsto c_x x ↦ c x , where c x c_x c x is the constant map taking the value x x x .
Proposition 6.19. (1) Let O Δ \mathcal{O}_{\Delta} O Δ in QC ( Map ( S n , X ) ) \operatorname{QC}(\operatorname{Map}(S^{n},\mathcal{X})) QC ( Map ( S n , X )) be the pushforward of the structure sheaf O X \mathcal{O}_{\mathcal{X}} O X of X \mathcal{X} X along ι n \iota_n ι n . Let Hom QC ( Map ( S n , X ) ) ( O Δ , O Δ ) \operatorname{Hom}_{\operatorname{QC}(\operatorname{Map}(S^{n},\mathcal{X}))}(\mathcal{O}_{\Delta},\mathcal{O}_{\Delta}) Hom QC ( Map ( S n , X )) ( O Δ , O Δ ) in Mod k \operatorname{Mod}_{k} Mod k denote the Hom complex. Then there exists an equivalence
H H ∙ ( n Mod QC ( X ) [ Pr k t ± ] / k ) ≃ H H ∙ ( n Mod QC ( X ) [ Pr k ] / k ) ≃ Hom QC ( Map ( S n , X ) ) ( O Δ , O Δ ) \mathcal{HH}^{\bullet}({}_{n}\operatorname{Mod}_{\operatorname{QC}(\mathcal{X})}[\operatorname{Pr}_{k}^{t\pm}]/k) \simeq\mathcal{HH}^{\bullet}({}_{n}\operatorname{Mod}_{\operatorname{QC}(\mathcal{X})}[\operatorname{Pr}_{k}]/k) \simeq\operatorname{Hom}_{\operatorname{QC}(\operatorname{Map}(S^{n},\mathcal{X}))}(\mathcal{O}_{\Delta},\mathcal{O}_{\Delta}) HH ∙ ( n Mod QC ( X ) [ Pr k t ± ] / k ) ≃ HH ∙ ( n Mod QC ( X ) [ Pr k ] / k ) ≃ Hom QC ( Map ( S n , X )) ( O Δ , O Δ ) in Mod k \operatorname{Mod}_{k} Mod k .
(2) Suppose that k k k is of characteristic zero, and X \mathcal{X} X is an ordinary scheme separated and of finite type. There exists an equivalence
H H ∙ ( n Mod QC ( X ) [ Pr k t ± ] / k ) ≃ Γ ( X , Sym O X ( T X / k [ − n − 1 ] ) ) \mathcal{HH}^{\bullet}({}_{n}\operatorname{Mod}_{\operatorname{QC}(\mathcal{X})}[\operatorname{Pr}_{k}^{t\pm}]/k) \simeq\Gamma(\mathcal{X},\operatorname{Sym}_{\mathcal{O}_{\mathcal{X}}}(T_{\mathcal{X}/k}[-n-1])) HH ∙ ( n Mod QC ( X ) [ Pr k t ± ] / k ) ≃ Γ ( X , Sym O X ( T X / k [ − n − 1 ])) in Mod k \operatorname{Mod}_{k} Mod k , induced by Koszul duality (cf. the proof).
This proposition is a direct generalization of the first part of Proposition 6.10 . The proof is also a rather straightforward generalization.
Proof. We have already proved H H ∙ ( n Mod QC ( X ) [ Pr k t ± ] / k ) ≃ H H ∙ ( n Mod QC ( X ) [ Pr k ] / k ) \mathcal{HH}^{\bullet}({}_{n}\operatorname{Mod}_{\operatorname{QC}(\mathcal{X})}[\operatorname{Pr}_{k}^{t\pm}]/k) \simeq\mathcal{HH}^{\bullet}({}_{n}\operatorname{Mod}_{\operatorname{QC}(\mathcal{X})}[\operatorname{Pr}_{k}]/k) HH ∙ ( n Mod QC ( X ) [ Pr k t ± ] / k ) ≃ HH ∙ ( n Mod QC ( X ) [ Pr k ] / k ) . According to [4 Theorem 4.7, Corollary 4.12] , there exists an equivalence
QC ( X ) ⊗ S n − 1 ≃ QC ( Map ( S n − 1 , X ) ) \operatorname{QC}(\mathcal{X})\otimes S^{n-1} \simeq\operatorname{QC}(\operatorname{Map}(S^{n-1},\mathcal{X})) QC ( X ) ⊗ S n − 1 ≃ QC ( Map ( S n − 1 , X ))
in CAlg ( Pr k ) \operatorname{CAlg}(\operatorname{Pr}_{k}) CAlg ( Pr k ) . Now we consider the underlying object of End QC ( X ) ⊗ S n − 1 ( QC ( X ) ) \operatorname{End}_{\operatorname{QC}(\mathcal{X})^{\otimes S^{n-1}}}(\operatorname{QC}(\mathcal{X})) End QC ( X ) ⊗ S n − 1 ( QC ( X )) . This is equivalent to Fun QC ( Map ( S n − 1 , X ) ) ( QC ( X ) , QC ( X ) ) ∈ Pr k \operatorname{Fun}_{\operatorname{QC}(\operatorname{Map}(S^{n-1},\mathcal{X}))}(\operatorname{QC}(\mathcal{X}),\operatorname{QC}(\mathcal{X})) \in\operatorname{Pr}_{k} Fun QC ( Map ( S n − 1 , X )) ( QC ( X ) , QC ( X )) ∈ Pr k , which denotes the Hom object from QC ( X ) \operatorname{QC}(\mathcal{X}) QC ( X ) to itself. Here QC ( X ) \operatorname{QC}(\mathcal{X}) QC ( X ) is the QC ( Map ( S n − 1 , X ) ) \operatorname{QC}(\operatorname{Map}(S^{n-1},\mathcal{X})) QC ( Map ( S n − 1 , X )) -module induced by the pullback functor ι n − 1 ∗ : QC ( Map ( S n − 1 , X ) ) → QC ( X ) \iota_{n-1}^{*}:\operatorname{QC}(\operatorname{Map}(S^{n-1},\mathcal{X}))\to\operatorname{QC}(\mathcal{X}) ι n − 1 ∗ : QC ( Map ( S n − 1 , X )) → QC ( X ) along the map X → Map ( S n − 1 , X ) \mathcal{X}\to\operatorname{Map}(S^{n-1},\mathcal{X}) X → Map ( S n − 1 , X ) . Since QC ( X ) \operatorname{QC}(\mathcal{X}) QC ( X ) is self-dual as an object of Mod QC ( Map ( S n − 1 , X ) ) ( Pr k ) \operatorname{Mod}_{\operatorname{QC}(\operatorname{Map}(S^{n-1},\mathcal{X}))}(\operatorname{Pr}_{k}) Mod QC ( Map ( S n − 1 , X )) ( Pr k ) (see [4 Corollary 4.8] ), we have
Fun QC ( Map ( S n − 1 , X ) ) ( QC ( X ) , QC ( X ) ) ≃ QC ( X ) ⊗ QC ( Map ( S n − 1 , X ) ) QC ( X ) . \operatorname{Fun}_{\operatorname{QC}(\operatorname{Map}(S^{n-1},\mathcal{X}))}(\operatorname{QC}(\mathcal{X}),\operatorname{QC}(\mathcal{X})) \simeq\operatorname{QC}(\mathcal{X})\otimes_{\operatorname{QC}(\operatorname{Map}(S^{n-1},\mathcal{X}))}\operatorname{QC}(\mathcal{X}). Fun QC ( Map ( S n − 1 , X )) ( QC ( X ) , QC ( X )) ≃ QC ( X ) ⊗ QC ( Map ( S n − 1 , X )) QC ( X ) .
Moreover, by [4 Theorem 4.7] there exist equivalences
QC ( X ) ⊗ QC ( Map ( S n − 1 , X ) ) QC ( X ) ≃ QC ( X × Map ( S n − 1 , X ) X ) ≃ QC ( Map ( S n , X ) ) \operatorname{QC}(\mathcal{X})\otimes_{\operatorname{QC}(\operatorname{Map}(S^{n-1},\mathcal{X}))}\operatorname{QC}(\mathcal{X}) \simeq\operatorname{QC}(\mathcal{X}\mathop{\times}_{\operatorname{Map}(S^{n-1},\mathcal{X})}\mathcal{X}) \simeq\operatorname{QC}(\operatorname{Map}(S^{n},\mathcal{X})) QC ( X ) ⊗ QC ( Map ( S n − 1 , X )) QC ( X ) ≃ QC ( X × Map ( S n − 1 , X ) X ) ≃ QC ( Map ( S n , X ))
in Pr k \operatorname{Pr}_{k} Pr k . Through these equivalences, O Δ \mathcal{O}_{\Delta} O Δ corresponds to the identity functor QC ( X ) → QC ( X ) \operatorname{QC}(\mathcal{X})\to\operatorname{QC}(\mathcal{X}) QC ( X ) → QC ( X ) . Then the underlying complex of the Hochschild cohomology is equivalent to Hom QC ( Map ( S n , X ) ) ( O Δ , O Δ ) \operatorname{Hom}_{\operatorname{QC}(\operatorname{Map}(S^{n},\mathcal{X}))}(\mathcal{O}_{\Delta},\mathcal{O}_{\Delta}) Hom QC ( Map ( S n , X )) ( O Δ , O Δ ) . Since k k k is of characteristic zero, the affinization of S n S^n S n is the trivial square zero extension Spec k ⊕ k ϵ n \operatorname{Spec} k \oplus k\epsilon_n Spec k ⊕ k ϵ n of Spec k \operatorname{Spec} k Spec k by k ⋅ ϵ n k\cdot\epsilon_n k ⋅ ϵ n where ϵ n \epsilon_n ϵ n is a “formal” element of homological degree n n n . Using this, we have an equivalence of derived schemes Map ( S n , X ) ≃ Spec Sym O X ∙ ( L X [ n ] ) \operatorname{Map}(S^n,\mathcal{X}) \simeq\operatorname{Spec}\operatorname{Sym}^{\bullet}_{\mathcal{O}_{\mathcal{X}}}(\mathbb{L}_{\mathcal{X}}[n]) Map ( S n , X ) ≃ Spec Sym O X ∙ ( L X [ n ]) . Then by Koszul duality [16 Proof of Theorem 3.8, Theorem 4.3, Corollary 4.4] , there exists an equivalence H o m QC ( Spec Sym O X ∙ ( L X [ n ] ) ) ( O Δ , O Δ ) ≃ Sym O X ∙ ( T X / k [ − n − 1 ] ) \mathcal{H}om_{\operatorname{QC}(\operatorname{Spec}\operatorname{Sym}^{\bullet}_{\mathcal{O}_{\mathcal{X}}}(\mathbb{L}_{\mathcal{X}}[n]))}(\mathcal{O}_{\Delta},\mathcal{O}_{\Delta}) \simeq\operatorname{Sym}^{\bullet}_{\mathcal{O}_{\mathcal{X}}}(\mathbb{T}_{\mathcal{X}/k}[-n-1]) H o m QC ( Spec Sym O X ∙ ( L X [ n ])) ( O Δ , O Δ ) ≃ Sym O X ∙ ( T X / k [ − n − 1 ]) in QC ( X ) \operatorname{QC}(\mathcal{X}) QC ( X ) . Taking global sections, our assertion follows. □ \square □
Remark 6.20. It is natural to ask if the equivalence of Proposition 6.19 (2) is promoted to an equivalence of dg Lie algebras such that the underlying dg Lie algebra H H ∙ ( n Mod QC ( X ) [ Pr k ± ] / k ) [ n + 1 ] \mathcal{HH}^{\bullet}(n\operatorname{Mod}_{\operatorname{QC}(\mathcal{X})}[\operatorname{Pr}^{\pm}_{k}]/k)[n+1] HH ∙ ( n Mod QC ( X ) [ Pr k ± ] / k ) [ n + 1 ] is obtained from E n + 2 \mathrm{E}_{n+2} E n + 2 -algebra, which describes the deformation theory of n Mod QC ( X ) [ Pr k ± ] n\operatorname{Mod}_{\operatorname{QC}(\mathcal{X})}[\operatorname{Pr}^{\pm}_{k}] n Mod QC ( X ) [ Pr k ± ] on commutative Artin algebras (see § 7 ). The right-hand side T = Γ ( X , Sym X ∙ ( T X / k [ − n − 1 ] ) ) [ n + 1 ] T=\Gamma(\mathcal{X},\operatorname{Sym}^{\bullet}_{\mathcal{X}}(\mathbb{T}_{\mathcal{X}/k}[-n-1]))[n+1] T = Γ ( X , Sym X ∙ ( T X / k [ − n − 1 ])) [ n + 1 ] also has a kind of Schouten bracket. The work of Toën [32 Section 5] gives an equivalence between the E n + 1 \mathrm{E}_{n+1} E n + 1 -Hochschild cohomology H H E n + 1 ( X / k ) \mathcal{HH}^{\mathrm{E}_{n+1}}(\mathcal{X}/k) HH E n + 1 ( X / k ) of QC ( X ) \operatorname{QC}(\mathcal{X}) QC ( X ) and T T T as dg Lie algebras, where the bracket on T T T is defined using the homology operad of the E n + 1 \mathrm{E}_{n+1} E n + 1 -operad. This construction also provides a link to the theory of shifted Poisson structures. We note that the E n + 2 \mathrm{E}_{n+2} E n + 2 -algebra structure on H H E n + 1 ( X / k ) \mathcal{HH}^{\mathrm{E}_{n+1}}(\mathcal{X}/k) HH E n + 1 ( X / k ) in [32 ] is defined by constructing an E n + 1 \mathrm{E}_{n+1} E n + 1 -operad action on S n S^n S n at the level of cobordisms (see Theorem 3.1 of loc. cit. for details). Thus, the definition of the E n + 2 \mathrm{E}_{n+2} E n + 2 -algebra H H E n + 1 ( X / k ) \mathcal{HH}^{\mathrm{E}_{n+1}}(\mathcal{X}/k) HH E n + 1 ( X / k ) differs from the definition of Hochschild cohomology by means of universality (i.e., centers or algebraization/iterated endomorphisms), though these will likely coincide: the key step is a comparison of E n + 1 \mathrm{E}_{n+1} E n + 1 -monoidal structures on QC ( Map ( S n , X ) ) \operatorname{QC}(\operatorname{Map}(S^n,\mathcal{X})) QC ( Map ( S n , X )) between that in [32 ] and ours (or one constructed as the center), which is beyond the scope of this paper. We also remark that if one uses [32 ] , then we have an equivalence in (2), at the level of Mod k \operatorname{Mod}_k Mod k for a perfect stack X \mathcal{X} X , which is a priori different from the above one.
Deformation theory
We will give applications to the deformation theory of t t t -structured stable presentable ( ∞ , n ) (\infty,n) ( ∞ , n ) -categories.
Deformation theory
Deformations and Hochschild cohomology
Let R R R be an almost Artin E n + 1 \mathrm{E}_{n+1} E n + 1 -algebra (see Definition 4.17). Let D \mathcal{D} D be an object of n Pr k ± n\operatorname{Pr}^{\pm}_{k} n Pr k ± . Suppose that there exists the endomorphism algebra End ( D ) ∈ Alg 1 ( n Pr k ± ) \operatorname{End}(\mathcal{D}) \in\operatorname{Alg}_1(n\operatorname{Pr}^{\pm}_{k}) End ( D ) ∈ Alg 1 ( n Pr k ± ) .
We define Deform D ( R ) c a t ∈ Cat ^ ∞ \operatorname{Deform}_{\mathcal{D}}(R)^{\mathrm{cat}} \in\widehat{\operatorname{Cat}}_{\infty} Deform D ( R ) cat ∈ Cat ∞ to be the pullback of
{ D } → n Pr k ± β ± ↑ RMod ( n − 1 ) Pr R ± ( n Pr k ± ) → β ± n Pr k ± \begin{CD}
\{\mathcal{D}\} @>>> n\operatorname{Pr}^{\pm}_{k} \\
@. @A\beta^{\pm}AA \\
\operatorname{RMod}_{(n-1)\operatorname{Pr}^{\pm}_{R}}(n\operatorname{Pr}^{\pm}_{k}) @>{\beta^{\pm}}>> n\operatorname{Pr}^{\pm}_{k}
\end{CD} { D } RMod ( n − 1 ) Pr R ± ( n Pr k ± ) β ± n Pr k ± β ± ⏐ ↑ n Pr k ±
where β ± \beta^{\pm} β ± is the base change functor along ( n − 1 ) Pr R ± → ( n − 1 ) Pr k ± (n-1)\operatorname{Pr}^{\pm}_{R}\to(n-1)\operatorname{Pr}^{\pm}_{k} ( n − 1 ) Pr R ± → ( n − 1 ) Pr k ± . We may regard an object of Deform D ( R ) c a t \operatorname{Deform}_{\mathcal{D}}(R)^{\mathrm{cat}} Deform D ( R ) cat as
( D R , ϕ : β ± ( D R ) = D R ⊗ ^ ( n − 1 ) Pr R ± ( n − 1 ) Pr k ± ≃ D ) (\mathcal{D}_{R},\phi:\beta^{\pm}(\mathcal{D}_{R})=\mathcal{D}_{R}\widehat{\otimes}_{(n-1)\operatorname{Pr}^{\pm}_{R}}(n-1)\operatorname{Pr}^{\pm}_{k}\simeq\mathcal{D}) ( D R , ϕ : β ± ( D R ) = D R ⊗ ( n − 1 ) Pr R ± ( n − 1 ) Pr k ± ≃ D )
such that D R \mathcal{D}_{R} D R is a ( n − 1 ) Pr R ± (n-1)\operatorname{Pr}^{\pm}_{R} ( n − 1 ) Pr R ± -module, and ϕ \phi ϕ is an equivalence in n Pr k ± n\operatorname{Pr}^{\pm}_{k} n Pr k ± . We define Deform D ( R ) = ( Deform D ( R ) c a t ) ≃ \operatorname{Deform}_{\mathcal{D}}(R)=(\operatorname{Deform}_{\mathcal{D}}(R)^{\mathrm{cat}})^{\simeq} Deform D ( R ) = ( Deform D ( R ) cat ) ≃ (cf. Section 2.1). We shall refer to Deform D ( R ) \operatorname{Deform}_{\mathcal{D}}(R) Deform D ( R ) as the space of deformations of D \mathcal{D} D to R R R . We will describe Deform D ( R ) \operatorname{Deform}_{\mathcal{D}}(R) Deform D ( R ) in terms of Hochschild cohomology.
Let us recall
Φ R , n : RMod ( n − 1 ) Pr R ± ( n Pr k ± ) ⟶ LMod ( n − 1 ) Pr D n + 1 ( R ) ± ( n Pr k ± ) \Phi_{R,n}:\operatorname{RMod}_{(n-1)\operatorname{Pr}^{\pm}_{R}}(n\operatorname{Pr}^{\pm}_{k})\longrightarrow\operatorname{LMod}_{(n-1)\operatorname{Pr}^{\pm}_{\operatorname{D}_{n+1}(R)}}(n\operatorname{Pr}^{\pm}_{k}) Φ R , n : RMod ( n − 1 ) Pr R ± ( n Pr k ± ) ⟶ LMod ( n − 1 ) Pr D n + 1 ( R ) ± ( n Pr k ± )
Passing to the base change to Pr k ± \operatorname{Pr}^{\pm}_{k} Pr k ± , we consider the commutative diagram
RMod ( n − 1 ) Pr R ± ( n Pr k ± ) → ≃ Φ ^ R , n LMod ( n − 1 ) Pr D n + 1 ( R ) ± ( n Pr k ± ) β ± ↓ ↓ q ± n Pr k ± → id n Pr k ± \begin{CD}
\operatorname{RMod}_{(n-1)\operatorname{Pr}^{\pm}_{R}}(n\operatorname{Pr}^{\pm}_{k}) @>{\widehat{\Phi}_{R,n}}>{\simeq}> \operatorname{LMod}_{(n-1)\operatorname{Pr}^{\pm}_{\operatorname{D}_{n+1}(R)}}(n\operatorname{Pr}^{\pm}_{k}) \\
@V{\beta^{\pm}}VV @VV{q^{\pm}}V \\
n\operatorname{Pr}^{\pm}_{k} @>{\operatorname{id}}>> n\operatorname{Pr}^{\pm}_{k}
\end{CD} RMod ( n − 1 ) Pr R ± ( n Pr k ± ) β ± ↓ ⏐ n Pr k ± Φ R , n ≃ id LMod ( n − 1 ) Pr D n + 1 ( R ) ± ( n Pr k ± ) ↓ ⏐ q ± n Pr k ±
where q ± q^{\pm} q ± the forgetful functor and β ± \beta^{\pm} β ± is the base change functor along ( n − 1 ) Prt R ± → ( n − 1 ) Prt k ± (n-1)\operatorname{Prt}^{\pm}_{R} \to(n-1)\operatorname{Prt}^{\pm}_{k} ( n − 1 ) Prt R ± → ( n − 1 ) Prt k ± . By Theorem 4.19 , Φ ^ R , n \widehat{\Phi}_{R,n} Φ R , n induces an equivalence. It gives rise to an equivalence
Φ ^ R , n D : RMod ( n − 1 ) Prt R ± ( n Prt k ± ) × β ± , n Prt k ± { D } ≃ LMod ( n − 1 ) Prt D n + 1 ( R ) ± ( n Prt k ± ) × q ± , n Prt k ± { D } . \widehat{\Phi}^{D}_{R,n}\colon\operatorname{RMod}_{(n-1)\operatorname{Prt}^{\pm}_{R}}(n\operatorname{Prt}^{\pm}_{k})\mathop{\times}_{\beta^{\pm},n\operatorname{Prt}^{\pm}_{k}}\{D\}\simeq\operatorname{LMod}_{(n-1)\operatorname{Prt}^{\pm}_{D_{n+1}(R)}}(n\operatorname{Prt}^{\pm}_{k})\mathop{\times}_{q^{\pm},n\operatorname{Prt}^{\pm}_{k}}\{D\}. Φ R , n D : RMod ( n − 1 ) Prt R ± ( n Prt k ± ) × β ± , n Prt k ± { D } ≃ LMod ( n − 1 ) Prt D n + 1 ( R ) ± ( n Prt k ± ) × q ± , n Prt k ± { D } .
We furthermore assume that D D D has Hochschild cohomology H H ∙ ( D / k ) ∈ Alg n + 1 ( Mod k ) \mathcal{HH}^{\bullet}(D/k)\in\operatorname{Alg}_{n+1}(\operatorname{Mod}_{k}) HH ∙ ( D / k ) ∈ Alg n + 1 ( Mod k ) (see Definition 6.1). By the universal property of End ( D ) \operatorname{End}(D) End ( D ) [24 6.1.2.40] and the definition of H H ∙ ( D / k ) \mathcal{HH}^{\bullet}(D/k) HH ∙ ( D / k ) in terms of algebraizations, we have equivalences of spaces
( LMod ( n − 1 ) Prt D n + 1 ( R ) ± ( n Prt k ± ) × q ± , n Prt k ± { D } ) ≃ ≃ Map Alg 1 ( Mod k ) ( ( n − 1 ) Prt D n + 1 ( R ) ± , End ( D ) ) ≃ Map Alg n + 1 ( Mod k ) ( D n + 1 ( R ) , H H ∙ ( D / k ) ) . \begin{aligned}
\left(\operatorname{LMod}_{(n-1)\operatorname{Prt}^{\pm}_{D_{n+1}(R)}}(n\operatorname{Prt}^{\pm}_{k})\mathop{\times}_{q^{\pm},n\operatorname{Prt}^{\pm}_{k}}\{D\}\right)^{\simeq}\simeq\operatorname{Map}_{\operatorname{Alg}_{1}(\operatorname{Mod}_{k})}\left((n-1)\operatorname{Prt}^{\pm}_{D_{n+1}(R)},\operatorname{End}(D)\right) \\
&\simeq\operatorname{Map}_{\operatorname{Alg}_{n+1}(\operatorname{Mod}_{k})}\left(D_{n+1}(R),\mathcal{HH}^{\bullet}(D/k)\right).
\end{aligned} ( LMod ( n − 1 ) Prt D n + 1 ( R ) ± ( n Prt k ± ) × q ± , n Prt k ± { D } ) ≃ ≃ Map Alg 1 ( Mod k ) ( ( n − 1 ) Prt D n + 1 ( R ) ± , End ( D ) ) ≃ Map Alg n + 1 ( Mod k ) ( D n + 1 ( R ) , HH ∙ ( D / k ) ) .
Combined with the equivalence Φ ^ R , n D \widehat{\Phi}^{D}_{R,n} Φ R , n D , we obtain an equivalence of spaces
Deform D ( R ) ≃ Map Alg n + 1 ( Mod k ) ( D n + 1 ( R ) , H H ∙ ( D / k ) ) . \operatorname{Deform}_{D}(R)\simeq\operatorname{Map}_{\operatorname{Alg}_{n+1}(\operatorname{Mod}_{k})}\left(D_{n+1}(R),\mathcal{HH}^{\bullet}(D/k)\right). Deform D ( R ) ≃ Map Alg n + 1 ( Mod k ) ( D n + 1 ( R ) , HH ∙ ( D / k ) ) .
If k ⊕ H H ∙ ( D / k ) → k k\oplus\mathcal{HH}^{\bullet}(D/k)\to k k ⊕ HH ∙ ( D / k ) → k denotes the associated augmented E n + 1 \mathrm{E}_{n+1} E n + 1 -algebra, then
Map Alg n + 1 ( Mod k ) ( D n + 1 ( R ) , H H ∙ ( D / k ) ) ≃ Map Alg n + 1 + ( Mod k ) ( D n + 1 ( R ) , k ⊕ H H ∙ ( D / k ) ) . \operatorname{Map}_{\operatorname{Alg}_{n+1}(\operatorname{Mod}_{k})}\left(D_{n+1}(R),\mathcal{HH}^{\bullet}(D/k)\right)\simeq\operatorname{Map}_{\operatorname{Alg}^{+}_{n+1}(\operatorname{Mod}_{k})}\left(D_{n+1}(R),k\oplus\mathcal{HH}^{\bullet}(D/k)\right). Map Alg n + 1 ( Mod k ) ( D n + 1 ( R ) , HH ∙ ( D / k ) ) ≃ Map Alg n + 1 + ( Mod k ) ( D n + 1 ( R ) , k ⊕ HH ∙ ( D / k ) ) .
Consequently, we see
Theorem 7.1. Suppose that D D D has Hochschild cohomology H H ∙ ( D / k ) \mathcal{HH}^{\bullet}(D/k) HH ∙ ( D / k ) . There exists a canonical equivalence
Deform D ( R ) ≃ Map Alg n + 1 + ( Mod k ) ( D n + 1 ( R ) , k ⊕ H H ∙ ( D / k ) ) . \operatorname{Deform}_{D}(R)\simeq\operatorname{Map}_{\operatorname{Alg}^{+}_{n+1}(\operatorname{Mod}_{k})}\left(D_{n+1}(R),k\oplus\mathcal{HH}^{\bullet}(D/k)\right). Deform D ( R ) ≃ Map Alg n + 1 + ( Mod k ) ( D n + 1 ( R ) , k ⊕ HH ∙ ( D / k ) ) .
For instance, when R = k ⊕ k [ j ] R=k\oplus k[j] R = k ⊕ k [ j ] is the trivial square extension by k [ j ] k[j] k [ j ] (j ≥ 0 j\geq0 j ≥ 0 ), then π 0 ( Deform D ( R ) ) ≃ H H n ( D / k ) = H n + j + 1 ( H H ∙ ( D / k ) ) \pi_{0}(\operatorname{Deform}_{D}(R))\simeq\mathcal{HH}^{n}(D/k)=H^{n+j+1}(\mathcal{HH}^{\bullet}(D/k)) π 0 ( Deform D ( R )) ≃ HH n ( D / k ) = H n + j + 1 ( HH ∙ ( D / k )) since D n + 1 ( R ) D_{n+1}(R) D n + 1 ( R ) is the free E n + 1 \mathrm{E}_{n+1} E n + 1 -algebra generated by k [ − n − j − 1 ] k[-n-j-1] k [ − n − j − 1 ] .
Next, we consider the deformation functor.
Definition 7.2. Let θ k , n + 1 → n ± : Alg n + 1 ( Mod k ≤ 0 ) → Alg 1 ( n Prt k ± ) \theta^{\pm}_{k,n+1\to n}\colon\operatorname{Alg}_{n+1}(\operatorname{Mod}^{\leq0}_{k})\to\operatorname{Alg}_{1}(n\operatorname{Prt}^{\pm}_{k}) θ k , n + 1 → n ± : Alg n + 1 ( Mod k ≤ 0 ) → Alg 1 ( n Prt k ± ) be the functor given by B ↦ n Prt B ± B\mapsto n\operatorname{Prt}^{\pm}_{B} B ↦ n Prt B ± .
We note that θ k , n + 1 → n ± \theta^{\pm}_{k,n+1\to n} θ k , n + 1 → n ± is fully faithful. We here write θ ± = θ k , n + 1 → n ± \theta^{\pm}=\theta^{\pm}_{k,n+1\to n} θ ± = θ k , n + 1 → n ± . We set
RLin ∧ ( k ) : = RMod ( n Prt k ± ) × Alg 1 ( n Prt k ± ) , θ ± Alg n + 1 ( Mod k ≤ 0 ) . \operatorname{RLin}^{\wedge}(k):=\operatorname{RMod}(n\operatorname{Prt}^{\pm}_{k})\mathop{\times}_{\operatorname{Alg}_{1}(n\operatorname{Prt}^{\pm}_{k}),\theta^{\pm}}\operatorname{Alg}_{n+1}(\operatorname{Mod}^{\leq0}_{k}). RLin ∧ ( k ) := RMod ( n Prt k ± ) × Alg 1 ( n Prt k ± ) , θ ± Alg n + 1 ( Mod k ≤ 0 ) .
where RMod ( n Prt k ± ) → Alg 1 ( n Prt k ± ) \operatorname{RMod}(n\operatorname{Prt}^{\pm}_{k})\to\operatorname{Alg}_{1}(n\operatorname{Prt}^{\pm}_{k}) RMod ( n Prt k ± ) → Alg 1 ( n Prt k ± ) is the forgetful functor (cf. § 2.1 ). Let RLin ∧ ( k ) † \operatorname{RLin}^{\wedge}(k)^{\dagger} RLin ∧ ( k ) † be the subcategory spanned by coCartesian morphisms over Alg n + 1 ( Mod k ≤ 0 ) \operatorname{Alg}_{n+1}(\operatorname{Mod}^{\leq0}_{k}) Alg n + 1 ( Mod k ≤ 0 ) . Let RLin ∧ ( k ) / ( D , k ) † \operatorname{RLin}^{\wedge}(k)^{\dagger}_{/(D,k)} RLin ∧ ( k ) / ( D , k ) † be the overcategory where ( D , k ) (D,k) ( D , k ) means D ∈ n Prt k ± D\in n\operatorname{Prt}^{\pm}_{k} D ∈ n Prt k ± lying over k ∈ Alg n + 1 ( Mod k ≤ 0 ) k\in\operatorname{Alg}_{n+1}(\operatorname{Mod}^{\leq0}_{k}) k ∈ Alg n + 1 ( Mod k ≤ 0 ) . We then have a left fibration r : RLin ∧ ( k ) / ( D , k ) † → Alg n + 1 + ( Mod k ≤ 0 ) r\colon\operatorname{RLin}^{\wedge}(k)^{\dagger}_{/(D,k)}\to\operatorname{Alg}^{+}_{n+1}(\operatorname{Mod}^{\leq0}_{k}) r : RLin ∧ ( k ) / ( D , k ) † → Alg n + 1 + ( Mod k ≤ 0 ) . Given R ∈ Alg n + 1 + ( Mod k ≤ 0 ) R\in\operatorname{Alg}^{+}_{n+1}(\operatorname{Mod}^{\leq0}_{k}) R ∈ Alg n + 1 + ( Mod k ≤ 0 ) we refer to an object of the fiber r − 1 ( R ) r^{-1}(R) r − 1 ( R ) over R R R as a complete deformation of D D D to R R R .
Recall that S \mathcal{S} S is the ∞ \infty ∞ -category of small spaces (∞ \infty ∞ -groupoids), and S ^ \widehat{\mathcal{S}} S is the ∞ \infty ∞ -category of V-small spaces.
We define
Deform ‾ D : Alg n + 1 + ( Mod k ≤ 0 ) → S ^ \overline{\operatorname{Deform}}_{D}\colon\operatorname{Alg}^{+}_{n+1}(\operatorname{Mod}^{\leq0}_{k})\to\widehat{\mathcal{S}} Deform D : Alg n + 1 + ( Mod k ≤ 0 ) → S
to be the functor which corresponds to the left fibration r r r . This is given informally by A ↦ Deform D ( A ) A\mapsto\operatorname{Deform}_{D}(A) A ↦ Deform D ( A ) . Let Deform D : Art n + 1 → S ^ \operatorname{Deform}_{D}\colon\operatorname{Ar t}_{n+1}\to\widehat{\mathcal{S}} Deform D : Art n + 1 → S be the restriction of Deform ‾ D \overline{\operatorname{Deform}}_{D} Deform D to Art n + 1 \operatorname{Ar t}_{n+1} Art n + 1 .
Next, we recall formal E m \mathrm{E}_{m} E m -moduli problems from [25 Section 4] . Let m m m be a natural number or ∞ \infty ∞ . A functor F : Art n → S F\colon\operatorname{Art}_{n}\to\mathcal{S} F : Art n → S from the ∞ \infty ∞ -category of Artin E m \mathrm{E}_{m} E m -algebras is a formal E m \mathrm{E}_{m} E m -moduli problem if
R → R 0 ↓ ↓ R 1 → R 01 \begin{CD}
R @>>> R_{0}\\
@VVV @VVV\\
R_{1} @>>> R_{01}
\end{CD} R ↓ ⏐ R 1 R 0 ↓ ⏐ R 01
in Art m \operatorname{Art}_{m} Art m such that H 0 ( R 0 ) → H 0 ( R 01 ) H^{0}(R_{0})\to H^{0}(R_{01}) H 0 ( R 0 ) → H 0 ( R 01 ) and H 0 ( R 1 ) → H 0 ( R 01 ) H^{0}(R_{1})\to H^{0}(R_{01}) H 0 ( R 1 ) → H 0 ( R 01 ) are surjective, the induced map
F ( R ) → F ( R 0 ) × F ( R 01 ) F ( R 1 ) F(R)\to F(R_{0})\mathop{\times}_{F(R_{01})}F(R_{1}) F ( R ) → F ( R 0 ) × F ( R 01 ) F ( R 1 )
is an equivalence. Let Stack ^ m ∗ \widehat{\operatorname{Stack}}_{m}^{*} Stack m ∗ be the full subcategory of Fun ( Art m , S ) \operatorname{Fun}(\operatorname{Art}_{m},\mathcal{S}) Fun ( Art m , S ) , which consists of formal E m \mathbb{E}_{m} E m -moduli problems. When m m m is a natural number, thanks to a theorem of Lurie [25 X, Theorem 4.0.8] there exists an equivalence of ∞ \infty ∞ -categories
Stack ^ m ∗ ≃ Alg m + ( Mod k ) . \widehat{\operatorname{Stack}}_{m}^{*} \simeq\operatorname{Alg}_{m}^{+}(\operatorname{Mod}_{k}). Stack m ∗ ≃ Alg m + ( Mod k ) .
Through the equivalence, an augmented algebra B ∈ Alg m + ( Mod k ) B \in\operatorname{Alg}_{m}^{+}(\operatorname{Mod}_{k}) B ∈ Alg m + ( Mod k ) corresponds to the formal E m \mathbb{E}_{m} E m -moduli problem F B ( m ) : Art m → S \mathcal{F}_{B}^{(m)} : \operatorname{Art}_{m} \to\mathcal{S} F B ( m ) : Art m → S defined informally by R ↦ Map Alg m + ( Mod k ) ( D m ( R ) , B ) R \mapsto\operatorname{Map}_{\operatorname{Alg}_{m}^{+}(\operatorname{Mod}_{k})}(\mathbb{D}_{m}(R), B) R ↦ Map Alg m + ( Mod k ) ( D m ( R ) , B ) .
Theorem 7.3. The functor Deform D \operatorname{Deform}_{\mathcal{D}} Deform D is a formal E n + 1 \mathbb{E}_{n+1} E n + 1 -moduli problem F k ⊕ H H ∙ ( D / k ) ( n + 1 ) \mathcal{F}_{k\oplus\mathcal{HH}^{\bullet}(\mathcal{D}/k)}^{(n+1)} F k ⊕ HH ∙ ( D / k ) ( n + 1 ) , that is represented by the augmented Hochschild cohomology k ⊕ H H ∙ ( D / k ) → k k\oplus\mathcal{HH}^{\bullet}(\mathcal{D}/k)\to k k ⊕ HH ∙ ( D / k ) → k through Stack ^ n + 1 ∗ ≃ Alg n + 1 + ( Mod k ) \widehat{\operatorname{Stack}}_{n+1}^{*}\simeq\operatorname{Alg}_{n+1}^{+}(\operatorname{Mod}_{k}) Stack n + 1 ∗ ≃ Alg n + 1 + ( Mod k ) .
Proof. It is enough to construct an equivalence Deform D ≃ F k ⊕ H H ∙ ( D / k ) ( n + 1 ) \operatorname{Deform}_{\mathcal{D}}\simeq\mathcal{F}_{k\oplus\mathcal{HH}^{\bullet}(\mathcal{D}/k)}^{(n+1)} Deform D ≃ F k ⊕ HH ∙ ( D / k ) ( n + 1 ) . It will suffice to construct an equivalence Deform D ( R ) ≃ Map Alg n + 1 + ( Mod k ) ( D n + 1 ( R ) , k ⊕ H H ∙ ( D / k ) ) \operatorname{Deform}_{\mathcal{D}}(R)\simeq\operatorname{Map}_{\operatorname{Alg}_{n+1}^{+}(\operatorname{Mod}_{k})}(\mathbb{D}_{n+1}(R),k\oplus\mathcal{HH}^{\bullet}(\mathcal{D}/k)) Deform D ( R ) ≃ Map Alg n + 1 + ( Mod k ) ( D n + 1 ( R ) , k ⊕ HH ∙ ( D / k )) in a functorial way with respect to R ∈ Art n + 1 R\in\operatorname{Art}_{n+1} R ∈ Art n + 1 . This can be done by applying the procedure in [25 X, Construction 5.3.18] . □ \square □
Obstruction theory
We consider obstructions for deformations. Let p : R ′ → R p:R'\to R p : R ′ → R be a morphism in Art n + 1 \operatorname{Art}_{n+1} Art n + 1 , which fits in the pullback square
R ′ → p R ↓ ↓ u k → k ⊕ V [ i ] \begin{CD}
R' @>p>> R \\
@VVV @VV u V \\
k @>>> k\oplus V[i]
\end{CD} R ′ ↓ ⏐ k p R ↓ ⏐ u k ⊕ V [ i ]
in Art n + 1 \operatorname{Art}_{n+1} Art n + 1 , where V V V is a finite dimensional vector space, i i i is a positive integer, and k ⊕ V [ i ] k\oplus V[i] k ⊕ V [ i ] is the trivial square zero extension by V [ i ] V[i] V [ i ] . We shall refer to p p p as an elementary extension of R R R by V [ i − 1 ] V[i-1] V [ i − 1 ] , associated to u u u . Every map R 1 → R 2 R_{1}\to R_{2} R 1 → R 2 in Art n + 1 \operatorname{Art}_{n+1} Art n + 1 is the composition of a sequence of elementary extensions.
Theorem 7.4. Let D R \mathcal{D}_{R} D R be a deformation of D \mathcal{D} D to R R R , that is an object of Deform D ( R ) \operatorname{Deform}_{\mathcal{D}}(R) Deform D ( R ) . For the above square diagram, there exists a canonically defined element Obs ( D R , p ) ∈ H H i + n + 1 ( D / k ) ⊗ k V \operatorname{Obs}(\mathcal{D}_{R},p)\in\mathcal{HH}^{i+n+1}(\mathcal{D}/k)\otimes_{k}V Obs ( D R , p ) ∈ HH i + n + 1 ( D / k ) ⊗ k V such that Obs ( D R , p ) = 0 \operatorname{Obs}(\mathcal{D}_{R},p)=0 Obs ( D R , p ) = 0 if and only if there exists D R ′ ∈ π 0 ( Deform D ( R ′ ) ) \mathcal{D}_{R'}\in\pi_{0}(\operatorname{Deform}_{\mathcal{D}}(R')) D R ′ ∈ π 0 ( Deform D ( R ′ )) lying over D R \mathcal{D}_{R} D R along the map p p p . Here H H i + n + 1 ( D / k ) \mathcal{HH}^{i+n+1}(\mathcal{D}/k) HH i + n + 1 ( D / k ) is H i + n + 1 ( H H ∙ ( D / k ) ) H^{i+n+1}(\mathcal{HH}^{\bullet}(\mathcal{D}/k)) H i + n + 1 ( HH ∙ ( D / k )) .
Proof. We note from Theorem 7.3 that there exists the canonical equivalence
α : Deform D ( R ′ ) ≃ Deform D ( R ) × Deform D ( k ⊕ V [ i ] ) Deform D ( k ) . \alpha:\operatorname{Deform}_{\mathcal{D}}(R')\simeq\operatorname{Deform}_{\mathcal{D}}(R)\mathop{\times}_{\operatorname{Deform}_{\mathcal{D}}(k\oplus V[i])}\operatorname{Deform}_{\mathcal{D}}(k). α : Deform D ( R ′ ) ≃ Deform D ( R ) × Deform D ( k ⊕ V [ i ]) Deform D ( k ) .
where Deform D ( k ) \operatorname{Deform}_{\mathcal{D}}(k) Deform D ( k ) is a contractible space. Since D n + 1 ( k ⊕ V [ i ] ) \mathbb{D}_{n+1}(k\oplus V[i]) D n + 1 ( k ⊕ V [ i ]) is a free E n + 1 \mathbb{E}_{n+1} E n + 1 -algebra generated by V ∨ [ − i − n − 1 ] V^{\vee}[-i-n-1] V ∨ [ − i − n − 1 ] [25 X, Proposition 4.5.6] , there exists an equivalence ϕ : π 0 ( Deform D ( k ⊕ V [ i ] ) ) ≃ H H i + n + 1 ( D / k ) ⊗ k V \phi:\pi_{0}(\operatorname{Deform}_{\mathcal{D}}(k\oplus V[i]))\simeq\mathcal{HH}^{i+n+1}(\mathcal{D}/k)\otimes_{k}V ϕ : π 0 ( Deform D ( k ⊕ V [ i ])) ≃ HH i + n + 1 ( D / k ) ⊗ k V . Here, V ∨ V^{\vee} V ∨ is the dual vector space. Moreover, by construction, 0 ∈ H H i + n + 1 ( D / k ) ⊗ k V 0\in\mathcal{HH}^{i+n+1}(\mathcal{D}/k)\otimes_{k}V 0 ∈ HH i + n + 1 ( D / k ) ⊗ k V corresponds to the trivial deformation, that is, the image of Deform D ( k ) → Deform D ( k ⊕ V [ i ] ) \operatorname{Deform}_{\mathcal{D}}(k)\to\operatorname{Deform}_{\mathcal{D}}(k\oplus V[i]) Deform D ( k ) → Deform D ( k ⊕ V [ i ]) . We define Obs ( D R , p ) \operatorname{Obs}(\mathcal{D}_{R},p) Obs ( D R , p ) to be the image of D R ∈ π 0 ( Deform D ( R ) ) \mathcal{D}_{R}\in\pi_{0}(\operatorname{Deform}_{\mathcal{D}}(R)) D R ∈ π 0 ( Deform D ( R )) in H H i + n + 1 ( D / k ) ⊗ k V \mathcal{HH}^{i+n+1}(\mathcal{D}/k)\otimes_{k}V HH i + n + 1 ( D / k ) ⊗ k V via ϕ \phi ϕ . By the equivalence α \alpha α , we see that there exists some D R ′ ∈ Deform D ( R ′ ) \mathcal{D}_{R'}\in\operatorname{Deform}_{\mathcal{D}}(R') D R ′ ∈ Deform D ( R ′ ) lying over D R \mathcal{D}_{R} D R exactly when the image of D R \mathcal{D}_{R} D R in Deform D ( k ⊕ V [ i ] ) \operatorname{Deform}_{\mathcal{D}}(k\oplus V[i]) Deform D ( k ⊕ V [ i ]) is equivalent to the trivial deformation. This condition corresponds to the condition Obs ( D R , p ) = 0 \operatorname{Obs}(\mathcal{D}_{R},p)=0 Obs ( D R , p ) = 0 . □ \square □
Deformation of categories of Higher categorified sheaves
We consider n Mod QC ( X ) [ Pr k t ± ] ∈ ( n + 1 ) Pr k t ± n\operatorname{Mod}_{\operatorname{QC}(\mathcal{X})}[\operatorname{Pr}_{k}^{t\pm}]\in(n+1)\operatorname{Pr}_{k}^{t\pm} n Mod QC ( X ) [ Pr k t ± ] ∈ ( n + 1 ) Pr k t ± , which we thought of as the ( ∞ , n + 1 ) (\infty,n+1) ( ∞ , n + 1 ) -category of t t t -structured (higher) categorified sheaves. Note that the underlying stable ( ∞ , n + 1 ) (\infty,n+1) ( ∞ , n + 1 ) -category is Mod QC ( X ) [ Pr k ] \operatorname{Mod}_{\operatorname{QC}(\mathcal{X})}[\operatorname{Pr}_{k}] Mod QC ( X ) [ Pr k ] .
Theorem 7.5. Let X \mathcal{X} X be a perfect stack over k k k . Consider the deformation functor Deform n Mod QC ( X ) [ Pr k t ± ] \operatorname{Deform}_{n\operatorname{Mod}_{\operatorname{QC}(\mathcal{X})}[\operatorname{Pr}_{k}^{t\pm}]} Deform n Mod QC ( X ) [ Pr k t ± ] . Then the isomorphism classes of deformations of n Mod QC ( X ) [ Pr k t ± ] n\operatorname{Mod}_{\operatorname{QC}(\mathcal{X})}[\operatorname{Pr}_{k}^{t\pm}] n Mod QC ( X ) [ Pr k t ± ] to the trivial square zero extension k ⊕ k [ i ] k\oplus k[i] k ⊕ k [ i ] (i ≥ 0 i\geq0 i ≥ 0 ) (i.e. π 0 ( Deform n Mod QC ( X ) [ Pr k t ± ] ( k ⊕ k [ i ] ) ) \pi_{0}(\operatorname{Deform}_{n\operatorname{Mod}_{\operatorname{QC}(\mathcal{X})}[\operatorname{Pr}_{k}^{t\pm}]}(k\oplus k[i])) π 0 ( Deform n Mod QC ( X ) [ Pr k t ± ] ( k ⊕ k [ i ])) is isomorphic to H i + n + 2 ( H o m QC ( X ) ( O X , O X ) ) H^{i+n+2}(\mathcal{H}om_{\operatorname{QC}(\mathcal{X})}(\mathcal{O}_{\mathcal{X}},\mathcal{O}_{\mathcal{X}})) H i + n + 2 ( H o m QC ( X ) ( O X , O X )) .
We further assume that X \mathcal{X} X is a scheme separated of finite type over k k k of characteristic zero. Then there is an isomorphism
π 0 ( Deform n Mod QC ( X ) [ Pr k t ± ] ( k ⊕ k [ i ] ) ) ≃ H i + n + 2 ( Γ ( X , Sym O X ( T X / k [ − n − 1 ] ) ) ) . \pi_{0}(\operatorname{Deform}_{n\operatorname{Mod}_{\operatorname{QC}(\mathcal{X})}[\operatorname{Pr}_{k}^{t\pm}]}(k\oplus k[i]))\simeq H^{i+n+2}(\Gamma(\mathcal{X},\operatorname{Sym}_{\mathcal{O}_{\mathcal{X}}}(\mathbb{T}_{\mathcal{X}/k}[-n-1]))). π 0 ( Deform n Mod QC ( X ) [ Pr k t ± ] ( k ⊕ k [ i ])) ≃ H i + n + 2 ( Γ ( X , Sym O X ( T X / k [ − n − 1 ]))) . Proof. This assertion follows from Proposition 6.10, Proposition 6.19 and Theorem 7.1 together with the fact that D n + 2 ( k ⊕ k [ i ] ) \mathbb{D}_{n+2}(k\oplus k[i]) D n + 2 ( k ⊕ k [ i ]) is a free E n + 2 \mathbb{E}_{n+2} E n + 2 -algebra generated by a shifted vector space k [ − i − n − 2 ] k[-i-n-2] k [ − i − n − 2 ] . □ \square □