1 Introduction

Every object X in a category induces a contravariant functor into the category of sets that sends an object Y to the set \({\text {Hom}}^{}_{}(Y,X)\). Any functor that is naturally isomorphic to such a functor is called representable. There are a number of results in various settings, called Brown representability, when every ‘reasonable’ functor is representable. The first such result is due to Brown; see [11].

The first Brown representability result for triangulated categories was established by Neeman [19, Theorem 3.1]. The work on hand was motivated by [10, Theorem 1.3] and [21, 4.3].

Theorem

(see Theorem 2.7). Let R be a \(\mathbb {Z}\)-graded graded-commutative noetherian ring and \({\textsf{T}}\) a graded R-linear triangulated category, that is strongly generated, Ext-finite, and idempotent complete. Then a graded R-linear cohomological functor \(\textsf{f} :{\textsf{T}}^{op} \rightarrow {\text {grMod}}(R)\) is graded representable if and only if \(\textsf{f}\) only takes values in \({\text {grmod}}(R)\).

In contrast to the previous works, we characterize the graded representable functors; those are the functors naturally isomorphic to

$$\begin{aligned} \coprod _{d \in \mathbb {Z}} {\text {Hom}}^{}_{{\textsf{T}}}(-,\Sigma ^d X) \end{aligned}$$

for some object X. The result is proved in Section 2. Without the assumption that R is noetherian and \({\textsf{T}}\) Ext-finite, we obtain necessary, through not sufficient, conditions for a functor to be graded representable; see Corollary 2.19.

The study of representability is motivated by the fact, that the characterization of representable functors in a triangulated category \({\textsf{T}}\) yields the existence of a right adjoint functor to a functor \({\textsf{S}} \rightarrow {\textsf{T}}\). For a nice discussion on this, see [20, Introduction]. In Section 3, we show the same holds in the graded setting.

Finally we discuss some examples where Theorem 2.7 yields new insight: When G is a finite group and R a commutative noetherian ring, then \({\text {D}}_{b}({\text {mod}}(RG))\) is Ext-finite as a \({{\,\textrm{H}\,}}^*(G,R)\)-linear category. In the second example, we consider the action of Hochschild cohomology \({\text {HH}}^{*}(R/Q)\) on \({\text {D}}_{b}({\text {mod}}(R))\), when Q is a regular ring and \(R = Q/(\varvec{f})\) a quotient by a regular sequence.

2 Representable functors in the graded setting

Let \({\textsf{T}}\) be a triangulated category with suspension functor \(\Sigma \).

2.1. For objects X and Y in \({\textsf{T}}\), we write

(2.1.1)

When \({\textsf{T}} = {\text {D}}_{}(R)\), the derived category of modules over a ring R, and X and Y are R-modules viewed as objects in \({\text {D}}_{}(R)\) via the natural embedding, then this coincides with the classical Ext-groups.

2.2. Let R be a \(\mathbb {Z}\)-graded graded-commutative ring. This means R decomposes as

$$\begin{aligned} R = \coprod _{d \in \mathbb {Z}} R_d\,, \end{aligned}$$

and the Koszul sign rule holds

$$\begin{aligned} rs = (-1)^{de} sr \quad \text {for } r \in R_d \text { and }s \in R_e\,. \end{aligned}$$

We say \(r \in R_d\) is an homogeneous element of degree d.

2.3. A triangulated category \({\textsf{T}}\) is graded R-linear if

  1. 1.

    for any objects X and Y in \({\textsf{T}}\), the abelian group \({\text {Ext}}^{*}_{{\textsf{T}}}(X,Y)\) is a graded R-module with the grading given by the coproduct in (2.1.1), and

  2. 2.

    composition is R-bilinear.

This data is equivalent to a ring homomorphism \(R \rightarrow {{\,\textrm{Z}\,}}({\textsf{T}})\), where

is the graded center of \({\textsf{T}}\). More precisely, a ring homomorphism \(\varphi :R \rightarrow {{\,\text {Z}\,}}({\textsf {T}})\) yields an R-action on \({\text {Ext}}^{*}_{{\textsf{T}}}(X,Y)\) via

$$\begin{aligned} \begin{aligned} {} r \cdot - :{\text{ Ext }}^{*}_{{\textsf {T}}}(X,Y)&\rightarrow {\text{ Ext }}^{*}_{{\textsf {T}}}(X,\Sigma ^d Y) = {\text{ Ext }}^{*}_{{\textsf {T}}}(X,Y)[d] \,, \\{} {} f&\mapsto (\Sigma ^d f) \circ \varphi (r)_X = \varphi (r)_Y \circ f \,, \end{aligned} \end{aligned}$$

for any homogeneous element \(r \in R\). Conversely, any homogeneous element \(r \in R\) yields a natural transformation \(\eta \) given by

for any \(X \in {\textsf{T}}\). It is straightforward to check that these identifications are well-defined and mutually inverse. The graded center has been studied in a number of works; for example [8, 13].

2.4. We denote by \({\text {grMod}}(R)\) the category of graded R-modules, and by \({\text {grmod}}(R)\) its full subcategory of finitely generated R-modules. The nth shift M[n] of a graded R-module M is given by \((M[n])_d = M_{n+d}\).

The suspension functor of a graded R-linear category \({\textsf{T}}\) in the first component of \({\text {Ext}}^{*}_{{\textsf{T}}}(-,-)\) corresponds to the negative shift in \({\text {grMod}}(R)\):

$$\begin{aligned} {\text {Ext}}^{*}_{{\textsf{T}}}(\Sigma ^n X,Y) \cong {\text {Ext}}^{*}_{{\textsf{T}}}(X,Y)[-n]\,. \end{aligned}$$

2.5. A functor \(\textsf{f} :{\textsf{T}}^{op} \rightarrow {\text {grMod}}(R)\) is graded R-linear if

  1. 1.

    the induced map \({\text {Ext}}^{*}_{{\textsf{T}}}(X,Y) \rightarrow {\text {Ext}}^{*}_{R}(\textsf{f}(Y),\textsf{f}(X))\) is a map of graded R-modules, and

  2. 2.

    the suspension becomes the negative shift under \(\textsf{f}\), that is

    $$\begin{aligned} \textsf{f}(\Sigma ^n X) = \textsf{f}(X)[-n]\,. \end{aligned}$$

The functor \(\textsf{f}\) is cohomological if \(\textsf{f}\) applied to any exact triangle yields a long exact sequence of graded R-modules.

Without explicitly stating, we always assume that a natural transformation between graded R-linear functors respects this structure.

Definition 2.6

A functor \({\textsf{T}}^{op} \rightarrow {\text {grMod}}(R)\) is graded representable if it is naturally isomorphic to

for some object X in \({\textsf{T}}\).

When \({\textsf{T}}\) is graded R-linear, then any graded representable functor is graded R-linear.

A graded R-linear functor \(\textsf{f} :{\textsf{T}}^{op} \rightarrow {\text {grMod}}(R)\) is graded representable if and only if \(\textsf{f}_d :{\textsf{T}}^{op} \rightarrow {\text {Mod}}(R_0)\) is representable for an(y) arbitrary integer d. The functors \(\textsf{f}_d\) are the degree d part of \(\textsf{f}\), that is . Since \(\textsf{f}\) is graded R-linear, the degree d part \(\textsf{f}_d\) for an integer d encodes all the information of \(\textsf{f}\), that is

Theorem 2.7

Let R be a \(\mathbb {Z}\)-graded graded-commutative noetherian ring and \({\textsf{T}}\) a graded R-linear triangulated category, that is strongly generated, Ext-finite, and idempotent complete. Then a graded R-linear cohomological functor \(\textsf{f} :{\textsf{T}}^{op} \rightarrow {\text {grMod}}(R)\) is graded representable if and only if \(\textsf{f}\) is locally finite.

Before we give a proof, we recall some definitions and properties:

2.8. A \(\mathbb {Z}\)-graded ring R is noetherian if and only if \(R_0\) is noetherian and R is finitely generated as an \(R_0\)-algebra; see for example [15, Corollaire (2.1.5)] or [12, Theorem 1.5.5]. In particular, such a ring is bounded below.

2.9. A graded R-linear triangulated category \({\textsf{T}}\) is Ext-finite if for all \(X, Y \in {\textsf{T}}\), the graded R-module \({\text {Ext}}^{*}_{{\textsf{T}}}(X,Y)\) is finitely generated.

A triangulated category \({\textsf{T}}\) is idempotent complete if for every object X in \({\textsf{T}}\) and every idempotent \(e \in {{\,\textrm{End}\,}}_{\textsf{T}}(X)\), that is \(e^2 = e\), there exists an object Y and maps

$$\begin{aligned} i :Y \rightarrow X \quad \text {and}\quad p :X \rightarrow Y \end{aligned}$$

such that \(p \circ i = {{\,\textrm{id}\,}}_Y\) and \(i \circ p = e\).

2.10. A subcategory \({\textsf{S}} \subseteq {\textsf{T}}\) is thick if it is triangulated and closed under retracts. Since the intersection of thick subcategories is thick, there exists a smallest thick subcategory of \({\textsf{T}}\) containing an object G, which we denote by \({{\,\textrm{thick}\,}}(G)\). We say G finitely builds an object X in \({\textsf{T}}\) when \(X \in {{\,\textrm{thick}\,}}(G)\).

There is an exhaustive filtration of \({{\,\textrm{thick}\,}}_R(G)\): Let \({{\,\textrm{thick}\,}}^1(G)\) be the smallest full subcategory containing G that is closed under finite coproducts, retracts, and suspension. Then

These are full subcategories and form an exhaustive filtration of \({{\,\textrm{thick}\,}}(G)\); cf. [3, 10]. In particular, if X lies in \({{\,\textrm{thick}\,}}(G)\), then there exists an integer n such that \(X \in {{\,\textrm{thick}\,}}^n(G)\).

A triangulated category \({\textsf{T}}\) is strongly generated if there exists an object G in \({\textsf{T}}\) and a non-negative integer n such that \({\textsf{T}} = {{\,\textrm{thick}\,}}^n(G)\). The object X is a strong generator of \({\textsf{T}}\); cf.  [21].

In the remainder of this section, we give a proof of Theorem 2.7. We fix a \(\mathbb {Z}\)-graded graded-commutative ring R, a graded R-linear triangulated category \({\textsf{T}}\), and a graded R-linear cohomological functor \(\textsf{f} :{\textsf{T}}^{op} \rightarrow {\text {grMod}}(R)\).

Lemma 2.11

(Graded version of Yoneda’s lemma). For any \(X \in {\textsf{T}}\), the map

$$\begin{aligned} {{\,\textrm{Nat}\,}}(\textsf{g}_X,\textsf{f}) \rightarrow \textsf{f}_0(X) \quad \text {given by}\quad \eta \mapsto \eta (X)({{\,\textrm{id}\,}}_X) \end{aligned}$$

is an isomorphism of abelian groups.

Proof

For \(u \in \textsf{f}_0(X)\), we define a natural transformation

where \(Y \in {\textsf{T}}\) and \(f \in {\text {Ext}}^{*}_{{\textsf{T}}}(Y,X)\). Since u is a degree zero element, the map \(\eta _u(Y)\) is homogeneous. It is straightforward to verify that this is the inverse of the map in the claim and both are maps of abelian groups. \(\square \)

In particular, any morphism \(f :X \rightarrow Y\) corresponds to a natural transformation

$$\begin{aligned} f_* :\textsf{g}_X \rightarrow \textsf{g}_Y \end{aligned}$$

given by post-composition.

2.12. Adapting the definitions in [21, Section 4] a graded R-linear functor \(\textsf{f} :{\textsf{T}}^{op} \rightarrow {\text {grMod}}(R)\) is

  • locally finitely generated if for every X in \({\textsf{T}}\), there exists Y in \({\textsf{T}}\) and a natural transformation \(\zeta :\textsf{g}_Y \rightarrow \textsf{f}\) such that \(\zeta (X)\) is surjective,

  • locally finitely presented if it is locally finitely generated and the kernel of any natural transformation \(\textsf{g}_Y \rightarrow \textsf{f}\) is locally finitely generated, and

  • locally finite if \(\textsf{f}\) only takes values in \({\text {grmod}}(R)\).

When \(\textsf{f} :{\textsf{T}}^{op} \rightarrow {\text {grMod}}(R)\) is locally finitely generated or locally finitely presented, then \(\textsf{f}_d :{\textsf{T}}^{op} \rightarrow {\text {Mod}}(R_0)\) is locally finitely generated or locally finitely presented in the sense of [21, Section 4], respectively. The same need not hold for locally finite, for examples, see Section 3.

If \({\textsf{T}}\) is Ext-finite, then any graded representable functor is locally finite. Without the assumption that \({\textsf{T}}\) is Ext-finite, we can make the following statement:

Lemma 2.13

Any graded representable functor is locally finitely presented.

Proof

It is clear that a graded representable functor is locally finitely generated. Let \(\textsf{g}_X\) be a graded representable functor, and \(\textsf{g}_Y \rightarrow \textsf{g}_X\) a natural transformation. By Yoneda’s lemma 2.11, this corresponds to a morphism \(Y \rightarrow X\). If we complete this to an exact triangle \(Z \rightarrow Y \rightarrow X \rightarrow \Sigma Z\), the sequence

$$\begin{aligned} \textsf{g}_Z \rightarrow \textsf{g}_Y \rightarrow \textsf{g}_X \end{aligned}$$

is exact on \({\textsf{T}}\). In particular, the kernel of \(\textsf{g}_Y \rightarrow \textsf{g}_X\) is locally finitely generated. \(\square \)

Lemma 2.14

If \(\textsf{f}\) is locally finite, then \(\textsf{f}\) is locally finitely generated.

Proof

Let X be an object in \({\textsf{T}}\). Then the R-module \(\textsf{f}(X)\) is finitely generated, and we can choose a finite set of homogeneous generators \(x_1, \dots , x_n\) of \(\textsf{f}(X)\) in degrees \(d_1, \dots , d_n\). Set

For every generator \(x_j\), we obtain canonical maps

$$\begin{aligned} \Sigma ^{d_j} X \xrightarrow {i_j} Y \xrightarrow {p_j} \Sigma ^{d_j} X \end{aligned}$$

whose composition is the identity map on \(\Sigma ^{d_j} X\). Let \(y \in \textsf{f}(Y)\) be the canonical element, for which

$$\begin{aligned} x_j = \textsf{f}(i_j)(y) \quad \text {for } 1 \le j \le n\,. \end{aligned}$$

Because of the suspensions introduced in the definition of Y, the element y is homogeneous of degree 0. By Yoneda’s lemma 2.11, the element y corresponds to the natural transformation \(\zeta :\textsf{g}_Y \rightarrow \textsf{f}\) with \(\zeta (Y)({{\,\textrm{id}\,}}_Y) = y\). Then \(\zeta (y)(i_j) = x_j\), and \(\zeta (X)\) is surjective. That is \(\textsf{f}\) is locally finitely generated. \(\square \)

In general a locally finite functor need not be locally finitely presented. This requires further assumptions on R and \({\textsf{T}}\):

Lemma 2.15

If R is noetherian and \({\textsf{T}}\) Ext-finite, then a locally finite functor \(\textsf{f}:{\textsf{T}}^{op} \rightarrow {\text {grMod}}(R)\) is locally finitely presented.

Proof

By Lemma 2.14, the functor \(\textsf{f}\) is locally finitely generated. Let \(\textsf{g}_Y \rightarrow \textsf{f}\) be a natural transformation. We set

Since \({\textsf{T}}\) is Ext-finite, the R-module \(\textsf{g}_Y(X)\) is finitely generated. By assumption on \(\textsf{f}\), so is \(\textsf{f}(X)\). Since R is noetherian, the kernel \(\textsf{f}'(X)\) is also finitely generated. Thus \(\textsf{f}'\) is a locally finite functor and by Lemma 2.14 it is locally finitely generated. In particular, \(\textsf{f}\) is locally finitely presented. \(\square \)

2.16. Let \((\textsf{f}_i,\eta _i)_{i > 0}\) be a direct system of cohomological functors \(\textsf{f}_i :{\textsf{T}} \rightarrow {\textsf{A}}\) where \({\textsf{A}}\) an abelian category and natural transformations \(\eta _i :\textsf{f}_i \rightarrow \textsf{f}_{i+1}\). Following [21, 4.2.2], a direct system \((\textsf{f}_i,\eta _i)_{i > 0}\) is almost constant on a subcategory \({\textsf{S}}\) of \({\textsf{T}}\) if for every \(X \in {\textsf{S}}\), the sequence

$$\begin{aligned} 0 \rightarrow \ker (\eta _i(X)) \rightarrow \textsf{f}_i(X) \rightarrow {{\,\textrm{colim}\,}}_j \textsf{f}_j(X) \rightarrow 0 \end{aligned}$$

is exact for all positive integers i.

A direct system \((X_i,f_i)_{i>0}\) of objects \(X_i\) and morphisms \(f_i :X_i \rightarrow X_{i+1}\) in \({\textsf{T}}\) is almost constant on \({\textsf{S}}\) if the induced direct system of functors \((\textsf{g}_{X_i},(f_i)_*)_{i>0}\) is almost constant on \({\textsf{S}}\).

For almost constant direct systems, the following hold; see [21, Proposition 4.13].

Facts 2.17

Let \({\textsf{S}} \subseteq {\textsf{T}}\) be a subcategory closed under suspension, and \((\textsf{f}_i,\eta _i)_{i>0}\) a direct system that is almost constant on \({\textsf{S}}\). Then

  1. 1.

    \((\textsf{f}_{ni+r})_{i \geqslant 0}\) is almost constant on \({{\,\textrm{thick}\,}}^n({\textsf{S}})\) for any \(r > 0\), and

  2. 2.

    \(\textsf{f}_{n+1} \rightarrow {{\,\textrm{colim}\,}}\textsf{f}_i\) is split surjective on \({{\,\textrm{thick}\,}}^n({\textsf{S}})\).

If the functors \(\textsf{f}_i\) are graded R-linear, the assumption that \({\textsf{S}}\) is closed under suspension is redundant.

Proposition 2.18

Let \({\textsf{T}}\) be a strongly generated, graded R-linear triangulated category and \(\textsf{f} :{\textsf{T}}^{op} \rightarrow {\text {grMod}}(R)\) a cohomological graded R-linear functor. Then \(\textsf{f}\) is locally finitely presented if and only if \(\textsf{f}\) is a retract of a graded representable functor.

Proof

We assume \(\textsf{f}\) is locally finitely presented. Let \(G \in {\textsf{T}}\) be a strong generator of \({\textsf{T}}\) with \({{\,\textrm{thick}\,}}^d(G) = {\textsf{T}}\). Then there exist \(A_1 \in {\textsf{T}}\) and a natural transformation \(\zeta _1 :\textsf{g}_{A_1} \rightarrow \textsf{f}\) such that \(\zeta _1(G)\) is surjective. Inductively we construct a direct system

$$\begin{aligned} \textsf{g}_{A_1} \rightarrow \textsf{g}_{A_2} \rightarrow \cdots \end{aligned}$$

with compatible natural transformations \(\zeta _i :\textsf{g}_{A_i} \rightarrow \textsf{f}\): Assume we have constructed \(A_i\) and \(\zeta _i\) for \(i \le n\). Since \(\textsf{f}\) is locally finitely presented, there exists

$$\begin{aligned} \textsf{g}_B \rightarrow \ker (\textsf{g}_{A_n} \rightarrow \textsf{f}) \end{aligned}$$

that is surjective on G. This induces a natural transformation \(\textsf{g}_B \rightarrow \textsf{g}_{A_n}\), which by the graded version of Yoneda’s lemma 2.11 corresponds to a morphism \(f :B \rightarrow A_n\). We complete this morphism to an exact triangle

$$\begin{aligned} B \rightarrow A_n \rightarrow A_{n+1} \rightarrow \Sigma B \end{aligned}$$

and apply \(\textsf{f}_0\), the degree 0 part of \(\textsf{f}\). By the graded version of Yoneda’s lemma 2.11, we obtain the exact sequence

$$\begin{aligned} {{\,\textrm{Nat}\,}}(\textsf{g}_B,\textsf{f}) \leftarrow {{\,\textrm{Nat}\,}}(\textsf{g}_{A_n},\textsf{f}) \leftarrow {{\,\textrm{Nat}\,}}(\textsf{g}_{A_{n+1}},\textsf{f})\,. \end{aligned}$$

Thus by construction of B, there exists a natural transformation \(\zeta _{n+1}\) whose image is \(\zeta _n\).

By this construction, we have

$$\begin{aligned} \ker (\textsf{g}_{A_n}(G) \rightarrow \textsf{f}(G)) = \ker (\textsf{g}_{A_n}(G) \rightarrow \textsf{g}_{A_{n+1}}(G))\,. \end{aligned}$$

Using this and that \(\zeta _1(G)\) is surjective, it is straightforward to verify that the direct system is almost constant on G. Then the induced natural transformation \({{\,\textrm{colim}\,}}_i \textsf{g}_{A_i} \rightarrow \textsf{f}\) is a natural isomorphism. By Facts 2.17, the natural transformation

$$\begin{aligned} \textsf{g}_{A_{d+1}} \rightarrow {{\,\textrm{colim}\,}}_i \textsf{g}_{A_i} \xrightarrow {\sim } \textsf{f} \end{aligned}$$

is split surjective, and thus \(\textsf{f}\) is a retract of \(\textsf{g}_{A_{d+1}}\) on \({\textsf{T}}\).

For the converse direction, we assume \(\textsf{f}\) is the retract of \(\textsf{g}_X\) for some object X. Then we have a canonical projection and a canonical injection

$$\begin{aligned} \textsf{g}_X \rightarrow \textsf{f} \quad \text {and}\quad \textsf{f} \rightarrow \textsf{g}_X\,, \end{aligned}$$

respectively. The canonical projection is surjective on \({\textsf{T}}\), the canonical injection is injective. In particular, the canonical projection yields that \(\textsf{f}\) is locally finitely generated. Given a natural transformation \(\textsf{g}_Y \rightarrow \textsf{f}\), its kernel coincides with the kernel of the composition \(\textsf{g}_Y \rightarrow \textsf{f} \rightarrow \textsf{g}_X\). By Lemma 2.13, any representable functor is locally finitely presented, and thus is \(\textsf{f}\). \(\square \)

Corollary 2.19

If \({\textsf{T}}\) is additionally idempotent complete, then every locally finitely presented functor is graded representable.

Proof

Let \(\textsf{f}\) be a locally finitely presented functor. By Proposition 2.18, it is a retract of a graded representable functor \(\textsf{g}_X\). Then the natural transformation

$$\begin{aligned} \textsf{g}_X \rightarrow \textsf{f} \rightarrow \textsf{g}_X \end{aligned}$$

corresponds to an idempotent \(e :X \rightarrow X\). Since \({\textsf{T}}\) is idempotent complete, there exists a retract of Y of X such that e decomposes as the natural inclusion and projection morphism. Then \(\textsf{f} \rightarrow \textsf{g}_X \rightarrow \textsf{g}_Y\) is a natural isomorphism, and \(\textsf{f}\) is graded representable. \(\square \)

Proof of Theorem 2.7

Since \({\textsf{T}}\) is Ext-finite, any graded representable functor is locally finite. For the converse, we assume \(\textsf{f}\) is locally finite. Since R is noetherian and \({\textsf{T}}\) Ext-finite, we can apply Lemma 2.15 to obtain that \(\textsf{f}\) is locally finitely presented. Then \(\textsf{f}\) is graded representable by Corollary 2.19. \(\square \)

3 Applications

Adjoint functors. As explained in [20, Introduction], there is a connection between representable functors and adjoint functors. In our context, we obtain the following:

Let R be a \(\mathbb {Z}\)-graded graded-commutative ring. A functor \(\textsf{f} :{\textsf{S}} \rightarrow {\textsf{T}}\) between R-linear graded triangulated categories is graded R-linear if it is exact and the induced map

$$\begin{aligned} {\text {Ext}}^{*}_{{\textsf{S}}}(X,Y) \rightarrow {\text {Ext}}^{*}_{{\textsf{T}}}(\textsf{f}(X),\textsf{f}(Y)) \end{aligned}$$

is a map of graded R-modules.

Lemma 3.1

Let R be a \(\mathbb {Z}\)-graded graded-commutative ring, and \({\textsf{S}}\), \({\textsf{T}}\) graded R-linear triangulated categories. Suppose \({\textsf{T}}\) is Ext-finite and every cohomological graded R-linear functor \({\textsf{S}}^{op} \rightarrow {\text {grMod}}(R)\), that is locally finite, is graded representable. Then every graded R-linear functor \(\textsf{f} :{\textsf{S}} \rightarrow {\textsf{T}}\) has a right adjoint.

Proof

We adapt the proof of [20, Theorem 8.4.4]. Given \(Y \in {\textsf{T}}\), we define a functor \(\textsf{h} :{\textsf{S}} \rightarrow {\text {grMod}}(R)\) by

This is a graded R-linear functor. Since \({\textsf{T}}\) is Ext-finite, this functor is locally finite. So by assumption, \({\textsf{h}}\) is graded representable, that is there exists an object \(\textsf{f}'(Y) \in {\textsf{S}}\) such that

$$\begin{aligned} {\text {Ext}}^{*}_{{\textsf{T}}}(\textsf{f}(-),Y) \cong {\text {Ext}}^{*}_{{\textsf{S}}}(-,\textsf{f}'(Y))\,. \end{aligned}$$

It remains to verify that \(\textsf{f}'\) is a functor and this isomorphism is natural in both components. Let \(f :Y \rightarrow Z\) be a morphism in \({\textsf{T}}\). Then the induced map

$$\begin{aligned} {\text {Ext}}^{*}_{{\textsf{S}}}(-,\textsf{f}'(Y)) \rightarrow {\text {Ext}}^{*}_{{\textsf{S}}}(-,\textsf{f}'(Z)) \end{aligned}$$

corresponds to a morphism \(\textsf{f}'(Y) \rightarrow \textsf{f}'(Z)\) by Yoneda’s lemma 2.11. Thus \(\textsf{f}'\) is a functor. The above isomorphism is natural by construction. So \(\textsf{f}'\) is a right adjoint of \(\textsf{f}\).

Corollary 3.2

Let R be a \(\mathbb {Z}\)-graded graded-commutative noetherian ring and \({\textsf{S}}\), \({\textsf{T}}\) Ext-finite graded R-linear triangulated categories. Suppose \({\textsf{S}}\) is strongly generated and idempotent complete. Then every graded R-linear functor \(\textsf{f} :{\textsf{S}} \rightarrow {\textsf{T}}\) has a right adjoint. \(\square \)

Derived category. Let R be a commutative noetherian ring and A an R-algebra that is finitely generated as an R-module. Then A is noetherian; see for example [18, Theorem 3.7]. The bounded derived category of finitely generated modules over A, denoted by \({\text {D}}_{b}({\text {mod}}(A))\), has a canonical structure as an R-linear category, and the R-module \({\text {Hom}}^{}_{{\text {D}}_{b}({\text {mod}}(A))}(X,Y) = {\text {Ext}}^{0}_{R}(X,Y)\) is finitely generated for any XY. In general, the category \({\text {D}}_{b}({\text {mod}}(A))\) need not be Ext-finite as an R-linear category. By [6, Corollary 2.10], the category \({\text {D}}_{b}({\text {mod}}(A))\) is idempotent complete.

3.3. In general the question whether \({\text {D}}_{b}({\text {mod}}(A))\) is strongly generated is rather difficult. When A is artinian, then \({\text {D}}_{b}({\text {mod}}(A))\) is strongly generated by [21, Proposition 7.37]. When \(A = R\) is a commutative notherian ring, then \({\text {D}}_{b}({\text {mod}}(R))\) is strongly generated when R is either essentially of finite type over a field or over an equicharacteristic excellent local ring; see [1, Main Theorem] and [16, Corollary 7.2].

In the following, we discuss two examples in which \({\text {D}}_{b}({\text {mod}}(A))\) is Ext-finite for some cohomology ring connected to A.

Finite croup over a commutative ring. We consider \(A = RG\), the group algebra of a finite group G.

3.4. The group cohomology of the group algebra RG with coefficients in an RG-complex M is

When \(M = R\), this is a \(\mathbb {Z}\)-graded graded-commutative ring, and every \({{\,\textrm{H}\,}}^*(G,M)\) is a graded \({{\,\textrm{H}\,}}^*(G,R)\)-module. In particular, for RG-complexes X, Y, the identification

$$\begin{aligned} {\text {Ext}}^{*}_{{\text {D}}_{b}({\text {mod}}(RG))}(X,Y) = {\text {Ext}}^{*}_{RG}(X,Y) \cong {{\,\textrm{H}\,}}^*(G,{\text {Hom}}^{}_{R}(X,Y)) \end{aligned}$$

holds and the cohomology ring \({{\,\textrm{H}\,}}^*(G,R)\) acts on any Ext-module; see for example [9, Proposition 3.1.8]. So the bounded derived category of finitely generated RG-modules \({\text {D}}_{b}({\text {mod}}(RG))\) is graded \({{\,\textrm{H}\,}}^*(G,R)\)-linear.

3.5. By [14, 22], the group cohomology ring \({{\,\textrm{H}\,}}^*(G,R)\) is noetherian, and \({{\,\textrm{H}\,}}^*(G,M)\) is finitely generated over \({{\,\textrm{H}\,}}^*(G,R)\) for every finitely generated RG-module M. In particular, the derived category \({\text {D}}_{b}({\text {mod}}(RG))\) is Ext-finite as a graded \({{\,\textrm{H}\,}}^*(G,R)\)-linear triangulated category.

Corollary 3.6

Let R be a commutative notherian ring and G a finite group. If \({\text {D}}_{b}({\text {mod}}(RG))\) is strongly generated, then a graded \({{\,\textrm{H}\,}}^*(G,R)\)-linear functor

$$\begin{aligned} \textsf{f} :{\text {D}}_{b}({\text {mod}}(RG)) \rightarrow {\text {grMod}}({{\,\textrm{H}\,}}^*(G,R)) \end{aligned}$$

is graded representable if and only if \(\textsf{f}\) is locally finite. \(\square \)

Regular ring modulo a regular sequence. We consider \(R = A\) a commutative noetherian ring.

3.7. The category \({\text {D}}_{b}({\text {mod}}(R))\) is Ext-finite over R if and only if the Ext-modules \({\text {Ext}}^{*}_{R}(X,Y)\) are bounded for all X and Y in \({\text {D}}_{b}({\text {mod}}(R))\). That is precisely when R is regular: When R is regular, the Ext-modules are bounded by definition. For the converse, for every X in \({\text {D}}_{b}({\text {mod}}(R))\), the Ext-module \({\text {Ext}}^{*}_{R}(X,R/\mathfrak {p})\) is bounded and \(X_\mathfrak {p}\) has finite projective dimension for any prime ideal \(\mathfrak {p}\) of R. Then X has finite projective dimension; see [7, Lemma 4.5] for modules, and [5, Theorem 4.1] and [17, Theorem 3.6] for complexes.

When R is regular, the bounded derived category \({\text {D}}_{b}({\text {mod}}(R))\) is strongly generated if and only if R is a strong generator. The later holds precisely when R has finite global dimension, that is R has finite Krull dimension. Then Rouquier’s representability theorem [21, Corollary 4.18] applies.

3.8. Suppose \(R = Q/(\varvec{f})\) is the quotient of a regular ring Q by a regular sequence \(\varvec{f} = f_1, \ldots , f_c\). Then there exist cohomological operators \(\varvec{\chi } = \chi _1, \ldots , \chi _c\) in degree 2 such that for XY in \({\text {D}}_{b}({\text {mod}}(R))\), the graded modules \({\text {Ext}}^{*}_{R}(X,Y)\) are finitely generated over the noetherian graded ring \(R[\varvec{\chi }]\); see [4, Theorem (4.2)]. In particular, the category \({\text {D}}_{b}({\text {mod}}(R))\) is \(R[\varvec{\chi }]\)-linear and Ext-finite.

The ring of cohomological operators coincides with the Hochschild cohomology

see [2, Section 3].

Corollary 3.9

Let \(R = Q/(\varvec{f})\) be the quotient of a regular ring Q by a regular sequence \(\varvec{f} = f_1, \ldots , f_c\) with cohomological operators \(\varvec{\chi }\). If \({\text {D}}_{b}({\text {mod}}(R))\) is strongly generated, then any graded \(R[\varvec{\chi }]\)-linear functor \(\textsf{f} :{\text {D}}_{b}({\text {mod}}(R)) \rightarrow {\text {grMod}}(R[\varvec{\chi }])\) is graded representable if and only if \(\textsf{f}\) is locally finite. \(\square \)

3.10. For Corollaries 3.6 and 3.9, it is crucial that the ring action on the derived category is graded since the Ext-modules need not be not bounded. In particular, Corollaries 3.6 and 3.9 are not consequences of [21, 4.3], but require Theorem 2.7.