A note on abelian envelopes

Anya Nordskova Vakgroep Wiskunde, Universiteit Hasselt, Agoralaan gebouw D, 3590 Diepenbeek, Belgium [email protected]
Abstract.

This is a short note bridging the gap between two notions of universal abelian categories associated to exact categories, namely, Rump’s quotient categories and Bodzenta-Bondal’s abelian envelopes. The established connection allows us to draw several easy conclusions, in particular, we answer some (technical) questions raised by Bodzenta and Bondal in [2].

1. Introduction

In [13] Rump defines the left (resp. right) quotient category Ql(\EuScriptE)subscript𝑄𝑙\EuScript𝐸Q_{l}(\EuScript E)italic_Q start_POSTSUBSCRIPT italic_l end_POSTSUBSCRIPT ( italic_E ) (resp. Qr(\EuScriptE)subscript𝑄𝑟\EuScript𝐸Q_{r}(\EuScript E)italic_Q start_POSTSUBSCRIPT italic_r end_POSTSUBSCRIPT ( italic_E )) for any exact category \EuScriptE\EuScript𝐸\EuScript Eitalic_E. More generally, he considers left (right) exact categories, which are defined by imposing only half of the axioms of the usual definition, see Section 2 for details. The category Ql(\EuScriptE)subscript𝑄𝑙\EuScript𝐸Q_{l}(\EuScript E)italic_Q start_POSTSUBSCRIPT italic_l end_POSTSUBSCRIPT ( italic_E ) comes with an exact full embedding il:\EuScriptEQl(\EuScriptE):subscript𝑖𝑙\EuScript𝐸subscript𝑄𝑙\EuScript𝐸i_{l}:\EuScript E\to Q_{l}(\EuScript E)italic_i start_POSTSUBSCRIPT italic_l end_POSTSUBSCRIPT : italic_E → italic_Q start_POSTSUBSCRIPT italic_l end_POSTSUBSCRIPT ( italic_E ) and is left abelian (a natural “one-sided” generalisation of the notion of abelian categories). It was shown in [13] that Ql(\EuScriptE)subscript𝑄𝑙\EuScript𝐸Q_{l}(\EuScript E)italic_Q start_POSTSUBSCRIPT italic_l end_POSTSUBSCRIPT ( italic_E ) together with ilsubscript𝑖𝑙i_{l}italic_i start_POSTSUBSCRIPT italic_l end_POSTSUBSCRIPT is universal among left abelian categories \EuScriptA\EuScript𝐴\EuScript Aitalic_A and right exact functors \EuScriptE\EuScriptA\EuScript𝐸\EuScript𝐴\EuScript E\to\EuScript Aitalic_E → italic_A, see Theorem 2.6 for the precise statement. In [2] Bodzenta and Bondal defined right (resp. left) abelian envelopes \EuScriptAr(\EuScriptE)\EuScriptsubscript𝐴𝑟\EuScript𝐸\EuScript A_{r}(\EuScript E)italic_A start_POSTSUBSCRIPT italic_r end_POSTSUBSCRIPT ( italic_E ) (resp. \EuScriptAl(\EuScriptE)\EuScriptsubscript𝐴𝑙\EuScript𝐸\EuScript A_{l}(\EuScript E)italic_A start_POSTSUBSCRIPT italic_l end_POSTSUBSCRIPT ( italic_E )) of exact categories. By definition, a right abelian envelope \EuScriptAr(\EuScriptE)\EuScriptsubscript𝐴𝑟\EuScript𝐸\EuScript A_{r}(\EuScript E)italic_A start_POSTSUBSCRIPT italic_r end_POSTSUBSCRIPT ( italic_E ) of an exact category \EuScriptE\EuScript𝐸\EuScript Eitalic_E is an abelian category together with a right exact functor \EuScriptE\EuScriptAr(\EuScriptE)\EuScript𝐸\EuScriptsubscript𝐴𝑟\EuScript𝐸\EuScript E\to\EuScript A_{r}(\EuScript E)italic_E → italic_A start_POSTSUBSCRIPT italic_r end_POSTSUBSCRIPT ( italic_E ), satisfying a universal property for all abelian categories \EuScriptA\EuScript𝐴\EuScript Aitalic_A and right exact functors \EuScriptE\EuScriptA\EuScript𝐸\EuScript𝐴\EuScript E\to\EuScript Aitalic_E → italic_A (Definition 3.1). Unlike Rump’s quotient categories, abelian envelopes do not always exist.

Even though the definitions look quite similar, the precise relation between the two notions is not immediately clear. For any exact category \EuScriptE\EuScript𝐸\EuScript Eitalic_E, it is evident (see [2]) from the two universal properties that \EuScriptAr(\EuScriptE)\EuScriptsubscript𝐴𝑟\EuScript𝐸\EuScript A_{r}(\EuScript E)italic_A start_POSTSUBSCRIPT italic_r end_POSTSUBSCRIPT ( italic_E ) exists and coincides with Ql(\EuScriptE)subscript𝑄𝑙\EuScript𝐸Q_{l}(\EuScript E)italic_Q start_POSTSUBSCRIPT italic_l end_POSTSUBSCRIPT ( italic_E ) if the latter is abelian. The main conclusion of this note is that the converse also holds, i.e. if \EuScriptAr(\EuScriptE)\EuScriptsubscript𝐴𝑟\EuScript𝐸\EuScript A_{r}(\EuScript E)italic_A start_POSTSUBSCRIPT italic_r end_POSTSUBSCRIPT ( italic_E ) exists then Ql(\EuScriptE)subscript𝑄𝑙\EuScript𝐸Q_{l}(\EuScript E)italic_Q start_POSTSUBSCRIPT italic_l end_POSTSUBSCRIPT ( italic_E ) is abelian and they coincide. We make this connection via the explicit description of right abelian envelopes obtained by Bodzenta and Bondal. Namely, they show that \EuScriptAr(\EuScriptE)\EuScriptsubscript𝐴𝑟\EuScript𝐸\EuScript A_{r}(\EuScript E)italic_A start_POSTSUBSCRIPT italic_r end_POSTSUBSCRIPT ( italic_E ), if exists, is isomorphic to the subcategory of compact objects Lex(\EuScriptE)cLexsuperscript\EuScript𝐸𝑐{\mathrm{Lex}}(\EuScript E)^{c}roman_Lex ( italic_E ) start_POSTSUPERSCRIPT italic_c end_POSTSUPERSCRIPT of left exact contravariant functors from \EuScriptE\EuScript𝐸\EuScript Eitalic_E to abelian groups AbAb{\mathrm{Ab}}roman_Ab. Here we observe that Lex(\EuScriptE)cLexsuperscript\EuScript𝐸𝑐{\mathrm{Lex}}(\EuScript E)^{c}roman_Lex ( italic_E ) start_POSTSUPERSCRIPT italic_c end_POSTSUPERSCRIPT is always left abelian and together with the Yoneda embedding i:\EuScriptELex(\EuScriptE)c:𝑖\EuScript𝐸Lexsuperscript\EuScript𝐸𝑐i:\EuScript E\to{\mathrm{Lex}}(\EuScript E)^{c}italic_i : italic_E → roman_Lex ( italic_E ) start_POSTSUPERSCRIPT italic_c end_POSTSUPERSCRIPT it satisfies the universal property of Ql(\EuScriptE)subscript𝑄𝑙\EuScript𝐸Q_{l}(\EuScript E)italic_Q start_POSTSUBSCRIPT italic_l end_POSTSUBSCRIPT ( italic_E ) for any exact category \EuScriptE\EuScript𝐸\EuScript Eitalic_E. It follows in particular that the universal functor iR:\EuScriptE\EuScriptAr(\EuScriptE):subscript𝑖𝑅\EuScript𝐸\EuScriptsubscript𝐴𝑟\EuScript𝐸i_{R}:\EuScript E\to\EuScript A_{r}(\EuScript E)italic_i start_POSTSUBSCRIPT italic_R end_POSTSUBSCRIPT : italic_E → italic_A start_POSTSUBSCRIPT italic_r end_POSTSUBSCRIPT ( italic_E ) is always fully faithful and exact (not only right exact), which answers the questions in Remark 4.6 and after Example 4.7 in [2].

It is worth noting that although abelian envelopes of [2] essentially turn out to be a special case of Rump’s quotient categories, this special case has many advantages, e.g. when considering derived categories (some of this is also discussed in [14]). We end the note with some simple applications and examples.

Acknowledgements. I would like to express my gratitude to Michel Van den Bergh and Alexey Bondal for many interesting and helpful discussions.

1.1. Conventions.

Throughout the note we will assume that all categories are skeletally small. All general “one-sided” (e.g. about right abelian envelopes) statements have natural dual counterparts which we do not mention.

2. Preliminaries and Rump’s quotient categories

We recall the definitions of left exact and left abelian categories.

Definition 2.1.

A left exact category is an additive category with a distinguished class of cokernels (deflations) satisfying the following properties:

  1. 1)

    The class of deflations is closed under compositions and contains identity morphisms.

  2. 2)

    The pullback of a deflation along any morphism exists and is again a deflation.

  3. 3)

    If the composition A𝑓B𝑔C𝑓𝐴𝐵𝑔𝐶A\xrightarrow{f}B\xrightarrow{g}Citalic_A start_ARROW overitalic_f → end_ARROW italic_B start_ARROW overitalic_g → end_ARROW italic_C is a deflation and g𝑔gitalic_g has a kernel, then g𝑔gitalic_g is a deflation.

Kernels of deflations will be referred to as inflations. A kernel-cokernel pair consisting of an inflation and a deflation is called a conflation.

Definition 2.2.

An additive category 𝒜𝒜\mathcal{A}caligraphic_A with cokernels is said to be left abelian if for every f:AB:𝑓𝐴𝐵f:A\to Bitalic_f : italic_A → italic_B, g:DB:𝑔𝐷𝐵g:D\to Bitalic_g : italic_D → italic_B with cg=0𝑐𝑔0cg=0italic_c italic_g = 0, c:=coker(f)assign𝑐coker𝑓c:={\mathrm{coker}}(f)italic_c := roman_coker ( italic_f ), there exists a cokernel d:ED:𝑑𝐸𝐷d:E\to Ditalic_d : italic_E → italic_D such that gd𝑔𝑑gditalic_g italic_d factors through f𝑓fitalic_f.

A𝐴{A}italic_AB𝐵{B}italic_BC𝐶{C}italic_C00{0}E𝐸{E}italic_ED𝐷{D}italic_D00{0}f𝑓\scriptstyle{f}italic_fc𝑐\scriptstyle{c}italic_cd𝑑\scriptstyle{d}italic_dg𝑔\scriptstyle{g}italic_g00\scriptstyle{0}

Dually one can define right exact and right abelian categories. Note that a category is abelian if and only if it is right and left abelian. An exact category in the sense of Quillen [10] is the same as a left and right exact category.

Definition 2.3.

An additive functor F:\EuScriptE𝒟:𝐹\EuScript𝐸𝒟F:\EuScript E\to\mathcal{D}italic_F : italic_E → caligraphic_D from a left exact category to an additive category is said to be right exact if F(g)=cokerF(f)𝐹𝑔coker𝐹𝑓F(g)={\mathrm{coker}}F(f)italic_F ( italic_g ) = roman_coker italic_F ( italic_f ) for every conflation A𝑓B𝑔C{A\xhookrightarrow{f}B\xrightarrow[]{g}\mathrel{\mkern-14.0mu}\rightarrow C}italic_A start_ARROW overitalic_f ↪ end_ARROW italic_B start_ARROW overitalic_g → end_ARROW → italic_C in \EuScriptE\EuScript𝐸\EuScript Eitalic_E. An additive functor F:\EuScriptE𝒟:𝐹\EuScript𝐸𝒟F:\EuScript E\to\mathcal{D}italic_F : italic_E → caligraphic_D from a right exact category to an additive category is left exact if Fop:\EuScriptEop𝒟op:superscript𝐹𝑜𝑝\EuScriptsuperscript𝐸𝑜𝑝superscript𝒟𝑜𝑝{F^{op}:\EuScript E^{op}\to\mathcal{D}^{op}}italic_F start_POSTSUPERSCRIPT italic_o italic_p end_POSTSUPERSCRIPT : italic_E start_POSTSUPERSCRIPT italic_o italic_p end_POSTSUPERSCRIPT → caligraphic_D start_POSTSUPERSCRIPT italic_o italic_p end_POSTSUPERSCRIPT is right exact.

Let \EuScriptE\EuScript𝐸\EuScript Eitalic_E be an additive category. By mod(\EuScriptE)mod\EuScript𝐸\mathop{\mathrm{mod}}\nolimits(\EuScript E)roman_mod ( italic_E ) we denote the category of additive contravariant functors F:\EuScriptEopAb:𝐹\EuScriptsuperscript𝐸𝑜𝑝Ab{F:\EuScript E^{op}\to{\mathrm{Ab}}}italic_F : italic_E start_POSTSUPERSCRIPT italic_o italic_p end_POSTSUPERSCRIPT → roman_Ab admitting a presentation

Hom\EuScriptE(,X)𝑓Hom\EuScriptE(,Y)F0𝑓subscriptHom\EuScript𝐸𝑋subscriptHom\EuScript𝐸𝑌𝐹0{\mathrm{Hom}}_{\EuScript E}(-,X)\xrightarrow{f}{\mathrm{Hom}}_{\EuScript E}(-% ,Y)\twoheadrightarrow F\to 0roman_Hom start_POSTSUBSCRIPT italic_E end_POSTSUBSCRIPT ( - , italic_X ) start_ARROW overitalic_f → end_ARROW roman_Hom start_POSTSUBSCRIPT italic_E end_POSTSUBSCRIPT ( - , italic_Y ) ↠ italic_F → 0

for some morphism f:XY:𝑓𝑋𝑌f:X\to Yitalic_f : italic_X → italic_Y in \EuScriptE\EuScript𝐸\EuScript Eitalic_E. If \EuScriptE\EuScript𝐸\EuScript Eitalic_E is a left exact category, denote by def(\EuScriptE)def\EuScript𝐸{\mathrm{def}}(\EuScript E)roman_def ( italic_E ) the full subcategory of mod(\EuScriptE)mod\EuScript𝐸\mathop{\mathrm{mod}}\nolimits(\EuScript E)roman_mod ( italic_E ) of functors of the form F=coker(Hom(,f))𝐹cokerHom𝑓F={\mathrm{coker}}({\mathrm{Hom}}(-,f))italic_F = roman_coker ( roman_Hom ( - , italic_f ) ) with f𝑓fitalic_f a deflation in \EuScriptE\EuScript𝐸\EuScript Eitalic_E.

One can show that mod(\EuScriptE)mod\EuScript𝐸\mathop{\mathrm{mod}}\nolimits(\EuScript E)roman_mod ( italic_E ) is a left abelian category and def(\EuScriptE)def\EuScript𝐸{\mathrm{def}}(\EuScript E)roman_def ( italic_E ) is a thick subcategory of mod(\EuScriptE)mod\EuScript𝐸\mathop{\mathrm{mod}}\nolimits(\EuScript E)roman_mod ( italic_E ), i.e. a Serre subcategory closed under subobjects and epimorphic images, for any left exact category \EuScriptE\EuScript𝐸\EuScript Eitalic_E. The localization theory for left abelian categories developed by Rump [11], [13] allows to define the quotient, which is again a left abelian category:

Definition 2.4 (Rump, [13]).

The left quotient category Ql(\EuScriptE)subscript𝑄𝑙\EuScript𝐸Q_{l}(\EuScript E)italic_Q start_POSTSUBSCRIPT italic_l end_POSTSUBSCRIPT ( italic_E ) of \EuScriptE\EuScript𝐸\EuScript Eitalic_E is defined by

Ql(\EuScriptE):=mod(\EuScriptE)/def(\EuScriptE)assignsubscript𝑄𝑙\EuScript𝐸mod\EuScript𝐸def\EuScript𝐸Q_{l}(\EuScript E):=\mathop{\mathrm{mod}}\nolimits(\EuScript E)/{\mathrm{def}}% (\EuScript E)italic_Q start_POSTSUBSCRIPT italic_l end_POSTSUBSCRIPT ( italic_E ) := roman_mod ( italic_E ) / roman_def ( italic_E )
Proposition 2.5 (Rump, [13], Proposition 5).

The Yoneda embedding \EuScriptEmod(\EuScriptE)\EuScript𝐸mod\EuScript𝐸\EuScript E\hookrightarrow\mathop{\mathrm{mod}}\nolimits(\EuScript E)italic_E ↪ roman_mod ( italic_E ) induces an exact full embedding i:\EuScriptEQl(\EuScriptE):𝑖\EuScript𝐸subscript𝑄𝑙\EuScript𝐸i:\EuScript E\hookrightarrow Q_{l}(\EuScript E)italic_i : italic_E ↪ italic_Q start_POSTSUBSCRIPT italic_l end_POSTSUBSCRIPT ( italic_E ) which reflects conflations.

Left quotient categories satisfy the following universal property.

Theorem 2.6 (Rump, [13], Corollary 2 of Proposition 6).

For any left exact category \EuScriptE\EuScript𝐸\EuScript Eitalic_E and any left abelian category 𝒜𝒜\mathcal{A}caligraphic_A, the embedding i:\EuScriptEQl(\EuScriptE):𝑖\EuScript𝐸subscript𝑄𝑙\EuScript𝐸i:\EuScript E\hookrightarrow Q_{l}(\EuScript E)italic_i : italic_E ↪ italic_Q start_POSTSUBSCRIPT italic_l end_POSTSUBSCRIPT ( italic_E ) induces an equivalence of categories

()i:Funcokercoker(Ql(\EuScriptE),𝒜)Rex(\EuScriptE,𝒜):𝑖subscriptFunmaps-tocokercokersubscript𝑄𝑙\EuScript𝐸𝒜Rex\EuScript𝐸𝒜(-)\circ i:{\mathrm{Fun}}_{{\mathrm{coker}}\mapsto{\mathrm{coker}}}(Q_{l}(% \EuScript E),\mathcal{A})\xrightarrow{\cong}{\mathrm{Rex}}(\EuScript E,% \mathcal{A})( - ) ∘ italic_i : roman_Fun start_POSTSUBSCRIPT roman_coker ↦ roman_coker end_POSTSUBSCRIPT ( italic_Q start_POSTSUBSCRIPT italic_l end_POSTSUBSCRIPT ( italic_E ) , caligraphic_A ) start_ARROW over≅ → end_ARROW roman_Rex ( italic_E , caligraphic_A )

where by FuncokercokersubscriptFunmaps-tocokercoker{\mathrm{Fun}}_{{\mathrm{coker}}\mapsto{\mathrm{coker}}}roman_Fun start_POSTSUBSCRIPT roman_coker ↦ roman_coker end_POSTSUBSCRIPT we denote the category of additive functors respecting cokernels and by RexRex{\mathrm{Rex}}roman_Rex the category of right exact functors.

Remark 2.7.

Note that Funcokercoker(,𝒞)=Rex(,𝒞)subscriptFunmaps-tocokercoker𝒞Rex𝒞{\mathrm{Fun}}_{{\mathrm{coker}}\mapsto{\mathrm{coker}}}(\mathcal{B},\mathcal{% C})={\mathrm{Rex}}(\mathcal{B},\mathcal{C})roman_Fun start_POSTSUBSCRIPT roman_coker ↦ roman_coker end_POSTSUBSCRIPT ( caligraphic_B , caligraphic_C ) = roman_Rex ( caligraphic_B , caligraphic_C ) if \mathcal{B}caligraphic_B and 𝒞𝒞\mathcal{C}caligraphic_C are abelian.

The key to proving this universal property is the fact that the left abelian category Ql(\EuScriptE)subscript𝑄𝑙\EuScript𝐸Q_{l}(\EuScript E)italic_Q start_POSTSUBSCRIPT italic_l end_POSTSUBSCRIPT ( italic_E ) together with the full embedding i𝑖iitalic_i is a dense extension of \EuScriptE\EuScript𝐸\EuScript Eitalic_E.

Definition 2.8 (Rump, [13], Definition 3).

Let \EuScriptE\EuScript𝐸\EuScript Eitalic_E be a left exact category. A right exact full embedding of \EuScriptE\EuScript𝐸\EuScript Eitalic_E into an additive category i:\EuScriptE𝒜:𝑖\EuScript𝐸𝒜i:\EuScript E\to\mathcal{A}italic_i : italic_E → caligraphic_A is a dense extension if every object A𝒜𝐴𝒜A\in\mathcal{A}italic_A ∈ caligraphic_A has a presentation E0𝑒E1𝑝A0𝑒subscript𝐸0subscript𝐸1𝑝𝐴0E_{0}\xrightarrow{e}E_{1}\xrightarrow{p}A\to 0italic_E start_POSTSUBSCRIPT 0 end_POSTSUBSCRIPT start_ARROW overitalic_e → end_ARROW italic_E start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT start_ARROW overitalic_p → end_ARROW italic_A → 0 with E0,E1\EuScriptEsubscript𝐸0subscript𝐸1\EuScript𝐸E_{0},E_{1}\in\EuScript Eitalic_E start_POSTSUBSCRIPT 0 end_POSTSUBSCRIPT , italic_E start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT ∈ italic_E, p=coker(e)𝑝coker𝑒p={\mathrm{coker}}(e)italic_p = roman_coker ( italic_e ), such that it satisfies the following two proprieties:

  1. 1)

    For every morphism f:EA:𝑓𝐸𝐴f:E\to Aitalic_f : italic_E → italic_A with E\EuScriptE𝐸\EuScript𝐸E\in\EuScript Eitalic_E ∈ italic_E there is a deflation d:EE:𝑑superscript𝐸𝐸d:E^{\prime}\twoheadrightarrow Eitalic_d : italic_E start_POSTSUPERSCRIPT ′ end_POSTSUPERSCRIPT ↠ italic_E in \EuScriptE\EuScript𝐸\EuScript Eitalic_E such that fd𝑓𝑑fditalic_f italic_d factors through p𝑝pitalic_p.

  2. 2)

    For every morphism f:EE1:𝑓𝐸subscript𝐸1f:E\to E_{1}italic_f : italic_E → italic_E start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT such that pf=0𝑝𝑓0pf=0italic_p italic_f = 0 there exists a deflation d:EE:𝑑superscript𝐸𝐸d:E^{\prime}\twoheadrightarrow Eitalic_d : italic_E start_POSTSUPERSCRIPT ′ end_POSTSUPERSCRIPT ↠ italic_E in \EuScriptE\EuScript𝐸\EuScript Eitalic_E such that fd𝑓𝑑fditalic_f italic_d factors through e𝑒eitalic_e.

A presentation satisfying these conditions will be referred to as a left exact presentation.

Rump establishes the following fact which we will use later.

Theorem 2.9 (Rump, [13], Corollary 5 of Proposition 6).

Let \EuScriptE\EuScript𝐸\EuScript Eitalic_E be a left exact category. Then the left quotient category \EuScriptEQl(\EuScriptE)\EuScript𝐸subscript𝑄𝑙\EuScript𝐸\EuScript E\hookrightarrow Q_{l}(\EuScript E)italic_E ↪ italic_Q start_POSTSUBSCRIPT italic_l end_POSTSUBSCRIPT ( italic_E ) is a dense extension of \EuScriptE\EuScript𝐸\EuScript Eitalic_E. Moreover, it is unique among dense extensions with cokernels in the following sense. If i:\EuScriptE𝒜:𝑖\EuScript𝐸𝒜i:\EuScript E\hookrightarrow\mathcal{A}italic_i : italic_E ↪ caligraphic_A is a dense extension of \EuScriptE\EuScript𝐸\EuScript Eitalic_E with cokernels, then there is a unique equivalence φ:𝒜Ql(\EuScriptE):𝜑similar-to-or-equals𝒜subscript𝑄𝑙\EuScript𝐸\varphi:\mathcal{A}\xrightarrow{\simeq}Q_{l}(\EuScript E)italic_φ : caligraphic_A start_ARROW over≃ → end_ARROW italic_Q start_POSTSUBSCRIPT italic_l end_POSTSUBSCRIPT ( italic_E ) making the diagram commute:

Ql()subscript𝑄𝑙{Q_{l}(\mathcal{E})}italic_Q start_POSTSUBSCRIPT italic_l end_POSTSUBSCRIPT ( caligraphic_E ){\mathcal{E}}caligraphic_E𝒜𝒜{\mathcal{A}}caligraphic_AYsubscript𝑌\scriptstyle{Y_{\mathcal{E}}}italic_Y start_POSTSUBSCRIPT caligraphic_E end_POSTSUBSCRIPTi𝑖\scriptstyle{i}italic_iφ𝜑\scriptstyle{\varphi}italic_φ

3. Connection to abelian envelopes of Bodzenta and Bondal

Now we recall the definition of right abelian envelopes of exact categories.

Definition 3.1 (Bodzenta-Bondal, [2]).

Let \EuScriptE\EuScript𝐸\EuScript Eitalic_E be an exact category. A right abelian envelope of \EuScriptE\EuScript𝐸\EuScript Eitalic_E is an abelian category \EuScriptAr(\EuScriptE)\EuScriptsubscript𝐴𝑟\EuScript𝐸\EuScript A_{r}(\EuScript E)italic_A start_POSTSUBSCRIPT italic_r end_POSTSUBSCRIPT ( italic_E ) together with a right exact functor iR:\EuScriptE\EuScriptAr(\EuScriptE):subscript𝑖𝑅\EuScript𝐸\EuScriptsubscript𝐴𝑟\EuScript𝐸i_{R}:\EuScript E\to\EuScript A_{r}(\EuScript E)italic_i start_POSTSUBSCRIPT italic_R end_POSTSUBSCRIPT : italic_E → italic_A start_POSTSUBSCRIPT italic_r end_POSTSUBSCRIPT ( italic_E ) such that for any abelian category \mathcal{B}caligraphic_B there is an equivalence of categories:

()iR:Rex(\EuScriptAr(\EuScriptE),)Rex(\EuScriptE,):subscript𝑖𝑅Rex\EuScriptsubscript𝐴𝑟\EuScript𝐸Rex\EuScript𝐸(-)\circ i_{R}:{\mathrm{Rex}}(\EuScript A_{r}(\EuScript E),\mathcal{B})% \xrightarrow{\cong}{\mathrm{Rex}}(\EuScript E,\mathcal{B})( - ) ∘ italic_i start_POSTSUBSCRIPT italic_R end_POSTSUBSCRIPT : roman_Rex ( italic_A start_POSTSUBSCRIPT italic_r end_POSTSUBSCRIPT ( italic_E ) , caligraphic_B ) start_ARROW over≅ → end_ARROW roman_Rex ( italic_E , caligraphic_B )

Left abelian envelopes are defined dually. Unlike the quotient categories of Rump, right/left abelian envelopes do not always exist.

Definition 3.2.

An object X𝑋Xitalic_X of a category 𝒞𝒞\mathcal{C}caligraphic_C is compact (also finitely presented in the literature) if Hom𝒞(X,)𝐻𝑜subscript𝑚𝒞𝑋Hom_{\mathcal{C}}(X,-)italic_H italic_o italic_m start_POSTSUBSCRIPT caligraphic_C end_POSTSUBSCRIPT ( italic_X , - ) commutes with all filtered colimits.

The following facts were established in [2]:

Proposition 3.3 ([2], Lemma 4.5 and the discussion after the definition of envelopes).

Let \EuScriptE\EuScript𝐸\EuScript Eitalic_E be an exact category.

  1. 1)

    If the left quotient category Ql(\EuScriptE)subscript𝑄𝑙\EuScript𝐸Q_{l}(\EuScript E)italic_Q start_POSTSUBSCRIPT italic_l end_POSTSUBSCRIPT ( italic_E ) is abelian, then the right abelian envelope \EuScriptAr(\EuScriptE)\EuScriptsubscript𝐴𝑟\EuScript𝐸\EuScript A_{r}(\EuScript E)italic_A start_POSTSUBSCRIPT italic_r end_POSTSUBSCRIPT ( italic_E ) exists and coincides with Ql(\EuScriptE)subscript𝑄𝑙\EuScript𝐸Q_{l}(\EuScript E)italic_Q start_POSTSUBSCRIPT italic_l end_POSTSUBSCRIPT ( italic_E ).

  2. 2)

    Let Lex(\EuScriptE):=Lex(\EuScriptEop,Ab)assignLex\EuScript𝐸Lex\EuScriptsuperscript𝐸𝑜𝑝Ab{\mathrm{Lex}}(\EuScript E):={\mathrm{Lex}}(\EuScript E^{op},{\mathrm{Ab}})roman_Lex ( italic_E ) := roman_Lex ( italic_E start_POSTSUPERSCRIPT italic_o italic_p end_POSTSUPERSCRIPT , roman_Ab ) be the category of contravariant left exact functors from \EuScriptE\EuScript𝐸\EuScript Eitalic_E to the category AbAb{\mathrm{Ab}}roman_Ab of abelian groups. Denote by Lex(\EuScriptE)cLexsuperscript\EuScript𝐸𝑐{\mathrm{Lex}}(\EuScript E)^{c}roman_Lex ( italic_E ) start_POSTSUPERSCRIPT italic_c end_POSTSUPERSCRIPT the subcategory of compact objects in Lex(\EuScriptE)Lex\EuScript𝐸{\mathrm{Lex}}(\EuScript E)roman_Lex ( italic_E ). If \EuScriptAr(\EuScriptE)\EuScriptsubscript𝐴𝑟\EuScript𝐸\EuScript A_{r}(\EuScript E)italic_A start_POSTSUBSCRIPT italic_r end_POSTSUBSCRIPT ( italic_E ) exists, then it is equivalent to Lex(\EuScriptE)cLexsuperscript\EuScript𝐸𝑐{\mathrm{Lex}}(\EuScript E)^{c}roman_Lex ( italic_E ) start_POSTSUPERSCRIPT italic_c end_POSTSUPERSCRIPT. Moreover, the functor iRsubscript𝑖𝑅i_{R}italic_i start_POSTSUBSCRIPT italic_R end_POSTSUBSCRIPT is faithful.

It is remarked, however, that the authors do not know whether the functor iRsubscript𝑖𝑅i_{R}italic_i start_POSTSUBSCRIPT italic_R end_POSTSUBSCRIPT in 2) composed with the natural embedding of compact objects Lex(\EuScriptE)cLex(\EuScriptE)Lexsuperscript\EuScript𝐸𝑐Lex\EuScript𝐸{\mathrm{Lex}}(\EuScript E)^{c}\hookrightarrow{\mathrm{Lex}}(\EuScript E)roman_Lex ( italic_E ) start_POSTSUPERSCRIPT italic_c end_POSTSUPERSCRIPT ↪ roman_Lex ( italic_E ) is the Yoneda embedding Y\EuScriptE:\EuScriptELex(\EuScriptE):subscript𝑌\EuScript𝐸\EuScript𝐸Lex\EuScript𝐸Y_{\EuScript}E:\EuScript E\hookrightarrow{\mathrm{Lex}}(\EuScript E)italic_Y start_POSTSUBSCRIPT end_POSTSUBSCRIPT italic_E : italic_E ↪ roman_Lex ( italic_E ). Indeed, this follows from their proof only provided that iRsubscript𝑖𝑅i_{R}italic_i start_POSTSUBSCRIPT italic_R end_POSTSUBSCRIPT is full, but this is also stated as unknown. In addition, it is natural to ask if \EuScriptAr(\EuScriptE)\EuScriptsubscript𝐴𝑟\EuScript𝐸\EuScript A_{r}(\EuScript E)italic_A start_POSTSUBSCRIPT italic_r end_POSTSUBSCRIPT ( italic_E ) always exists when the category Lex(\EuScriptE)cLexsuperscript\EuScript𝐸𝑐{\mathrm{Lex}}(\EuScript E)^{c}roman_Lex ( italic_E ) start_POSTSUPERSCRIPT italic_c end_POSTSUPERSCRIPT is abelian. We answer these questions by establishing a more precise connection between Bodzenta-Bondal’s abelian envelopes and Rump’s quotient categories:

Theorem 3.4.

Let \EuScriptE\EuScript𝐸\EuScript Eitalic_E be an exact category. Then the category Lex(\EuScriptE)cLexsuperscript\EuScript𝐸𝑐{\mathrm{Lex}}(\EuScript E)^{c}roman_Lex ( italic_E ) start_POSTSUPERSCRIPT italic_c end_POSTSUPERSCRIPT together with the Yoneda embedding Y\EuScriptEsubscript𝑌\EuScript𝐸Y_{\EuScript}Eitalic_Y start_POSTSUBSCRIPT end_POSTSUBSCRIPT italic_E is equivalent to the left quotient category Ql(\EuScriptE)subscript𝑄𝑙\EuScript𝐸Q_{l}(\EuScript E)italic_Q start_POSTSUBSCRIPT italic_l end_POSTSUBSCRIPT ( italic_E ).

Note that it is claimed in particular that the Yoneda embedding Y\EuScriptE:\EuScriptELex(\EuScriptE)c:subscript𝑌\EuScript𝐸\EuScript𝐸Lexsuperscript\EuScript𝐸𝑐Y_{\EuScript}E:\EuScript E\to{\mathrm{Lex}}(\EuScript E)^{c}italic_Y start_POSTSUBSCRIPT end_POSTSUBSCRIPT italic_E : italic_E → roman_Lex ( italic_E ) start_POSTSUPERSCRIPT italic_c end_POSTSUPERSCRIPT maps \EuScriptE\EuScript𝐸\EuScript Eitalic_E to compact objects in Lex(\EuScriptE)Lex\EuScript𝐸{\mathrm{Lex}}(\EuScript E)roman_Lex ( italic_E ) (this will be explained later). We conclude:

Corollary 3.5.

The following are equivalent for any exact category \EuScriptE\EuScript𝐸\EuScript Eitalic_E.

  1. 1)

    \EuScriptAr(\EuScriptE)\EuScriptsubscript𝐴𝑟\EuScript𝐸\EuScript A_{r}(\EuScript E)italic_A start_POSTSUBSCRIPT italic_r end_POSTSUBSCRIPT ( italic_E ) exists.

  2. 2)

    Ql(\EuScriptE)subscript𝑄𝑙\EuScript𝐸Q_{l}(\EuScript E)italic_Q start_POSTSUBSCRIPT italic_l end_POSTSUBSCRIPT ( italic_E ) is abelian.

  3. 3)

    Lex(\EuScriptE)cLexsuperscript\EuScript𝐸𝑐{\mathrm{Lex}}(\EuScript E)^{c}roman_Lex ( italic_E ) start_POSTSUPERSCRIPT italic_c end_POSTSUPERSCRIPT is abelian.

In this case the universal functor iR:\EuScriptE\EuScriptAr(\EuScriptE):subscript𝑖𝑅\EuScript𝐸\EuScriptsubscript𝐴𝑟\EuScript𝐸i_{R}:\EuScript E\to\EuScript A_{r}(\EuScript E)italic_i start_POSTSUBSCRIPT italic_R end_POSTSUBSCRIPT : italic_E → italic_A start_POSTSUBSCRIPT italic_r end_POSTSUBSCRIPT ( italic_E ) is the Yoneda embedding Y\EuScriptE:\EuScriptELex(\EuScriptE)c:subscript𝑌\EuScript𝐸\EuScript𝐸Lexsuperscript\EuScript𝐸𝑐Y_{\EuScript}E:\EuScript E\to{\mathrm{Lex}}(\EuScript E)^{c}italic_Y start_POSTSUBSCRIPT end_POSTSUBSCRIPT italic_E : italic_E → roman_Lex ( italic_E ) start_POSTSUPERSCRIPT italic_c end_POSTSUPERSCRIPT. In particular, iRsubscript𝑖𝑅i_{R}italic_i start_POSTSUBSCRIPT italic_R end_POSTSUBSCRIPT is exact and fully faithful.

Remark 3.6.
  1. 1)

    Theorem 3.4 claims that mod(\EuScriptE)/def(\EuScriptE)Lex(\EuScriptE)cmod\EuScript𝐸def\EuScript𝐸Lexsuperscript\EuScript𝐸𝑐\mathop{\mathrm{mod}}\nolimits(\EuScript E)/{\mathrm{def}}(\EuScript E)\cong{% \mathrm{Lex}}(\EuScript E)^{c}roman_mod ( italic_E ) / roman_def ( italic_E ) ≅ roman_Lex ( italic_E ) start_POSTSUPERSCRIPT italic_c end_POSTSUPERSCRIPT for any exact category \EuScriptE\EuScript𝐸\EuScript Eitalic_E. One can also show this directly. Let us sketch the argument in the case when mod(\EuScriptE)mod\EuScript𝐸\mathop{\mathrm{mod}}\nolimits(\EuScript E)roman_mod ( italic_E ) is abelian (equivalently, when \EuScriptE\EuScript𝐸\EuScript Eitalic_E has weak kernels, see the discussion after Definition 3.13). Recall that Lex(\EuScriptE)Lex\EuScript𝐸{\mathrm{Lex}}(\EuScript E)roman_Lex ( italic_E ) is the quotient of the category of all additive contravariant functors Mod(\EuScriptE)Mod\EuScript𝐸{\mathrm{Mod}}(\EuScript E)roman_Mod ( italic_E ) from \EuScriptE\EuScript𝐸\EuScript Eitalic_E to AbAb{\mathrm{Ab}}roman_Ab by the Serre subcategory of weakly effaceable functors Eff(\EuScriptE)Eff\EuScript𝐸{\mathrm{Eff}}(\EuScript E)roman_Eff ( italic_E ), see e.g. [8]. By the discussion after Theorem 3.9, any compact object in both Mod(\EuScriptE)Mod\EuScript𝐸{\mathrm{Mod}}(\EuScript E)roman_Mod ( italic_E ) and Lex(\EuScriptE)Lex\EuScript𝐸{\mathrm{Lex}}(\EuScript E)roman_Lex ( italic_E ) is the cokernel of a morphism in \EuScriptE\EuScript𝐸\EuScript Eitalic_E. Hence the localisation functor, which is an equivalence on \EuScriptE\EuScript𝐸\EuScript Eitalic_E, induces an exact essentially surjective functor Q:mod(\EuScriptE)(Mod\EuScriptE)cLex(\EuScriptE)c:𝑄mod\EuScript𝐸superscriptMod\EuScript𝐸𝑐Lexsuperscript\EuScript𝐸𝑐{Q:\mathop{\mathrm{mod}}\nolimits(\EuScript E)\cong({\mathrm{Mod}}{\EuScript E% })^{c}\to{\mathrm{Lex}}(\EuScript E)^{c}}italic_Q : roman_mod ( italic_E ) ≅ ( roman_Mod italic_E ) start_POSTSUPERSCRIPT italic_c end_POSTSUPERSCRIPT → roman_Lex ( italic_E ) start_POSTSUPERSCRIPT italic_c end_POSTSUPERSCRIPT. It is known that Eff(\EuScriptE)mod(\EuScriptE)=def(\EuScriptE)Eff\EuScript𝐸mod\EuScript𝐸def\EuScript𝐸{\mathrm{Eff}}(\EuScript E)\cap\mathop{\mathrm{mod}}\nolimits(\EuScript E)={% \mathrm{def}}(\EuScript E)roman_Eff ( italic_E ) ∩ roman_mod ( italic_E ) = roman_def ( italic_E ) (e.g. [5], Proposition 3.20 and [15], Lemma 9). Thus the kernel of Q𝑄Qitalic_Q is def(\EuScriptE)def\EuScript𝐸{\mathrm{def}}(\EuScript E)roman_def ( italic_E ).

  2. 2)

    Let lex(\EuScriptE)lex\EuScript𝐸{\mathrm{lex}}(\EuScript E)roman_lex ( italic_E ) be the full subcategory of mod(\EuScriptE)mod\EuScript𝐸\mathop{\mathrm{mod}}\nolimits(\EuScript E)roman_mod ( italic_E ) consisting of left exact functors. We warn the reader that lex(\EuScriptE)lex\EuScript𝐸{\mathrm{lex}}(\EuScript E)roman_lex ( italic_E ) is not necessarily equivalent to Lex(\EuScriptE)cLexsuperscript\EuScript𝐸𝑐{\mathrm{Lex}}(\EuScript E)^{c}roman_Lex ( italic_E ) start_POSTSUPERSCRIPT italic_c end_POSTSUPERSCRIPT even when \EuScriptE\EuScript𝐸\EuScript Eitalic_E has weak kernels. To be more precise, in this case lex(\EuScriptE)lex\EuScript𝐸{\mathrm{lex}}(\EuScript E)roman_lex ( italic_E ) is the subcategory of def(\EuScriptE)def\EuScript𝐸{\mathrm{def}}(\EuScript E)roman_def ( italic_E )-closed objects in mod(\EuScriptE)mod\EuScript𝐸\mathop{\mathrm{mod}}\nolimits(\EuScript E)roman_mod ( italic_E ), i.e. lex(\EuScriptE)={Xmod(\EuScriptE)|Hom(def(\EuScriptE),X)=Ext1(def(\EuScriptE),X)=0}lex\EuScript𝐸conditional-set𝑋mod\EuScript𝐸Homdef\EuScript𝐸𝑋superscriptExt1def\EuScript𝐸𝑋0{\mathrm{lex}}(\EuScript E)=\{X\in\mathop{\mathrm{mod}}\nolimits(\EuScript E)% \ |\ {\mathrm{Hom}}({\mathrm{def}}(\EuScript E),X)=\mathop{\mathrm{Ext}}% \nolimits^{1}({\mathrm{def}}(\EuScript E),X)=0\}roman_lex ( italic_E ) = { italic_X ∈ roman_mod ( italic_E ) | roman_Hom ( roman_def ( italic_E ) , italic_X ) = roman_Ext start_POSTSUPERSCRIPT 1 end_POSTSUPERSCRIPT ( roman_def ( italic_E ) , italic_X ) = 0 } (see [9], Proposition 2.8). Hence, if we assume additionally that the localisation functor L:mod(\EuScriptE)mod(\EuScriptE)/def(\EuScriptE):𝐿mod\EuScript𝐸mod\EuScript𝐸def\EuScript𝐸L:\mathop{\mathrm{mod}}\nolimits(\EuScript E)\to\mathop{\mathrm{mod}}\nolimits% (\EuScript E)/{\mathrm{def}}(\EuScript E)italic_L : roman_mod ( italic_E ) → roman_mod ( italic_E ) / roman_def ( italic_E ) has a right adjoint, then mod(\EuScriptE)/def(\EuScriptE)lex(\EuScriptE)mod\EuScript𝐸def\EuScript𝐸lex\EuScript𝐸\mathop{\mathrm{mod}}\nolimits(\EuScript E)/{\mathrm{def}}(\EuScript E)\cong{% \mathrm{lex}}(\EuScript E)roman_mod ( italic_E ) / roman_def ( italic_E ) ≅ roman_lex ( italic_E ). However, this is not always the case ([Ibid., Example 2.10]). Note that if \EuScriptE\EuScript𝐸\EuScript Eitalic_E has enough projectives, then L𝐿Litalic_L indeed has a right adjoint ([Ibid., Proposition 2.17]).

  3. 3)

    Note an immediately consequence of the fact that the universal functors iR:\EuScriptE\EuScriptAr(\EuScriptE):subscript𝑖𝑅\EuScript𝐸\EuScriptsubscript𝐴𝑟\EuScript𝐸i_{R}:\EuScript E\to\EuScript A_{r}(\EuScript E)italic_i start_POSTSUBSCRIPT italic_R end_POSTSUBSCRIPT : italic_E → italic_A start_POSTSUBSCRIPT italic_r end_POSTSUBSCRIPT ( italic_E ), iL:\EuScriptE\EuScriptAl(\EuScriptE):subscript𝑖𝐿\EuScript𝐸\EuScriptsubscript𝐴𝑙\EuScript𝐸i_{L}:\EuScript E\to\EuScript A_{l}(\EuScript E)italic_i start_POSTSUBSCRIPT italic_L end_POSTSUBSCRIPT : italic_E → italic_A start_POSTSUBSCRIPT italic_l end_POSTSUBSCRIPT ( italic_E ) are exact and not only right and left exact respectively. Namely, if \EuScriptE\EuScript𝐸\EuScript Eitalic_E has both abelian envelopes, then the two universal properties give rise to a pair of adjoint functors between \EuScriptAr(\EuScriptE)\EuScriptsubscript𝐴𝑟\EuScript𝐸\EuScript A_{r}(\EuScript E)italic_A start_POSTSUBSCRIPT italic_r end_POSTSUBSCRIPT ( italic_E ) and \EuScriptAl(\EuScriptE)\EuScriptsubscript𝐴𝑙\EuScript𝐸\EuScript A_{l}(\EuScript E)italic_A start_POSTSUBSCRIPT italic_l end_POSTSUBSCRIPT ( italic_E ) (see also [14]).

We will need some basic facts about Grothendieck categories and compact objects in the proof of Theorem 3.4.

Definition 3.7.

A category 𝒜𝒜\mathcal{A}caligraphic_A is locally finitely presented if it has filtered colimits, the full subcategory of compact objects 𝒜csuperscript𝒜𝑐\mathcal{A}^{c}caligraphic_A start_POSTSUPERSCRIPT italic_c end_POSTSUPERSCRIPT is skeletally small and every object of 𝒜𝒜\mathcal{A}caligraphic_A is a filtered colimit of compact objects. A Grothendieck category is an abelian category which has a generator and small colimits and satisfies (AB5), i.e. small filtered colimits of exact sequences are exact.

Remark 3.8.

An abelian category is locally finitely presented if and only if it is a Grothendieck category with a generating set of compact objects (see [3], Satz 1.5, [4], 2.4).

Theorem 3.9 (Breitsprecher, [3], Satz 1.11).

Let 𝒞𝒞\mathcal{C}caligraphic_C be a locally finitely presented Grothendieck category and 𝒮𝒞𝒮𝒞\mathcal{S}\subset\mathcal{C}caligraphic_S ⊂ caligraphic_C a class of compact generators. Then X𝒞𝑋𝒞X\in\mathcal{C}italic_X ∈ caligraphic_C is compact if and only if there is a presentation i=0nSii=0mSiX0superscriptsubscriptdirect-sum𝑖0𝑛subscript𝑆𝑖superscriptsubscriptdirect-sum𝑖0𝑚superscriptsubscript𝑆𝑖𝑋0\bigoplus_{i=0}^{n}S_{i}\to\bigoplus_{i=0}^{m}S_{i}^{\prime}\to X\to 0⨁ start_POSTSUBSCRIPT italic_i = 0 end_POSTSUBSCRIPT start_POSTSUPERSCRIPT italic_n end_POSTSUPERSCRIPT italic_S start_POSTSUBSCRIPT italic_i end_POSTSUBSCRIPT → ⨁ start_POSTSUBSCRIPT italic_i = 0 end_POSTSUBSCRIPT start_POSTSUPERSCRIPT italic_m end_POSTSUPERSCRIPT italic_S start_POSTSUBSCRIPT italic_i end_POSTSUBSCRIPT start_POSTSUPERSCRIPT ′ end_POSTSUPERSCRIPT → italic_X → 0 with Si,Si𝒮subscript𝑆𝑖superscriptsubscript𝑆𝑖𝒮S_{i},S_{i}^{\prime}\in\mathcal{S}italic_S start_POSTSUBSCRIPT italic_i end_POSTSUBSCRIPT , italic_S start_POSTSUBSCRIPT italic_i end_POSTSUBSCRIPT start_POSTSUPERSCRIPT ′ end_POSTSUPERSCRIPT ∈ caligraphic_S.

Let \EuScriptE\EuScript𝐸\EuScript Eitalic_E be an exact category. It is well known that Lex(\EuScriptE)Lex\EuScript𝐸{\mathrm{Lex}}(\EuScript E)roman_Lex ( italic_E ) is a Grothendieck category and the Yoneda functor Y\EuScriptE:\EuScriptELex(\EuScriptE):subscript𝑌\EuScript𝐸\EuScript𝐸Lex\EuScript𝐸Y_{\EuScript}E:\EuScript E\to{\mathrm{Lex}}(\EuScript E)italic_Y start_POSTSUBSCRIPT end_POSTSUBSCRIPT italic_E : italic_E → roman_Lex ( italic_E ), sending E\EuScriptE𝐸\EuScript𝐸E\in\EuScript Eitalic_E ∈ italic_E to Hom\EuScriptE(,E)subscriptHom\EuScript𝐸𝐸{\mathrm{Hom}}_{\EuScript E}(-,E)roman_Hom start_POSTSUBSCRIPT italic_E end_POSTSUBSCRIPT ( - , italic_E ), is fully faithful and exact. One can show that \EuScriptE\EuScript𝐸\EuScript Eitalic_E is a class of compact generators of Lex(\EuScriptE)Lex\EuScript𝐸{\mathrm{Lex}}(\EuScript E)roman_Lex ( italic_E ). The fact that representable functors are compact follows from the fact that they are compact in Mod(\EuScriptE)Mod\EuScript𝐸{\mathrm{Mod}}(\EuScript E)roman_Mod ( italic_E ) and the subcategory Lex(\EuScriptE)Lex\EuScript𝐸{\mathrm{Lex}}(\EuScript E)roman_Lex ( italic_E ) is closed under filtered colimits (see e.g. [7]). In particular, the Yoneda embedding Y\EuScriptEsubscript𝑌\EuScript𝐸Y_{\EuScript}Eitalic_Y start_POSTSUBSCRIPT end_POSTSUBSCRIPT italic_E maps \EuScriptE\EuScript𝐸\EuScript Eitalic_E into Lex(\EuScriptE)cLexsuperscript\EuScript𝐸𝑐{\mathrm{Lex}}(\EuScript E)^{c}roman_Lex ( italic_E ) start_POSTSUPERSCRIPT italic_c end_POSTSUPERSCRIPT and every object of Lex(\EuScriptE)cLexsuperscript\EuScript𝐸𝑐{\mathrm{Lex}}(\EuScript E)^{c}roman_Lex ( italic_E ) start_POSTSUPERSCRIPT italic_c end_POSTSUPERSCRIPT is the cokernel of a morphism in \EuScriptE\EuScript𝐸\EuScript Eitalic_E by Theorem 3.9.

Proposition 3.10 (e.g. Lowen-Kaledin, [7], Proposition 2.17).

Let \EuScriptE\EuScript𝐸\EuScript Eitalic_E be an exact category. Let p:XE:𝑝𝑋𝐸p:X\twoheadrightarrow Eitalic_p : italic_X ↠ italic_E be an epimorphism in Lex(\EuScriptE)Lex\EuScript𝐸{\mathrm{Lex}}(\EuScript E)roman_Lex ( italic_E ) with E\EuScriptE𝐸\EuScript𝐸E\in\EuScript Eitalic_E ∈ italic_E. Then there exist E\EuScriptEsuperscript𝐸\EuScript𝐸E^{\prime}\in\EuScript Eitalic_E start_POSTSUPERSCRIPT ′ end_POSTSUPERSCRIPT ∈ italic_E and a morphism g:EX:𝑔superscript𝐸𝑋g:E^{\prime}\to Xitalic_g : italic_E start_POSTSUPERSCRIPT ′ end_POSTSUPERSCRIPT → italic_X in Lex(\EuScriptE)Lex\EuScript𝐸{\mathrm{Lex}}(\EuScript E)roman_Lex ( italic_E ) such that the composition fg:EE:𝑓𝑔superscript𝐸𝐸fg:E^{\prime}\twoheadrightarrow Eitalic_f italic_g : italic_E start_POSTSUPERSCRIPT ′ end_POSTSUPERSCRIPT ↠ italic_E is again an epimorphism.

The following result implies that Lex(\EuScriptE)cLexsuperscript\EuScript𝐸𝑐{\mathrm{Lex}}(\EuScript E)^{c}roman_Lex ( italic_E ) start_POSTSUPERSCRIPT italic_c end_POSTSUPERSCRIPT is left abelian for any exact category \EuScriptE\EuScript𝐸\EuScript Eitalic_E.

Theorem 3.11 (Breitsprecher [3], see also Crawley-Boevey [4], Rump [12]).

Let 𝒜𝒜\mathcal{A}caligraphic_A be an additive locally finitely presented category. Then 𝒜𝒜\mathcal{A}caligraphic_A is abelian if and only if 𝒜csuperscript𝒜𝑐\mathcal{A}^{c}caligraphic_A start_POSTSUPERSCRIPT italic_c end_POSTSUPERSCRIPT is left abelian.

Proof of Theorem 3.4.

Since the category Lex(\EuScriptE)cLexsuperscript\EuScript𝐸𝑐{\mathrm{Lex}}(\EuScript E)^{c}roman_Lex ( italic_E ) start_POSTSUPERSCRIPT italic_c end_POSTSUPERSCRIPT has cokernels, due to Theorem 2.9 it is sufficient to show that Y\EuScriptE:\EuScriptELex(\EuScriptE)c:subscript𝑌\EuScript𝐸\EuScript𝐸Lexsuperscript\EuScript𝐸𝑐Y_{\EuScript}E:\EuScript E\to{\mathrm{Lex}}(\EuScript E)^{c}italic_Y start_POSTSUBSCRIPT end_POSTSUBSCRIPT italic_E : italic_E → roman_Lex ( italic_E ) start_POSTSUPERSCRIPT italic_c end_POSTSUPERSCRIPT is a dense extension. As discussed above, every XLex(\EuScriptE)c𝑋Lexsuperscript\EuScript𝐸𝑐X\in{\mathrm{Lex}}(\EuScript E)^{c}italic_X ∈ roman_Lex ( italic_E ) start_POSTSUPERSCRIPT italic_c end_POSTSUPERSCRIPT has a presentation

E0𝑎E1𝑝X0𝑎subscript𝐸0subscript𝐸1𝑝𝑋0E_{0}\xrightarrow{a}E_{1}\xrightarrow{p}X\to 0italic_E start_POSTSUBSCRIPT 0 end_POSTSUBSCRIPT start_ARROW overitalic_a → end_ARROW italic_E start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT start_ARROW overitalic_p → end_ARROW italic_X → 0

with E0,E1\EuScriptEsubscript𝐸0subscript𝐸1\EuScript𝐸E_{0},E_{1}\in\EuScript Eitalic_E start_POSTSUBSCRIPT 0 end_POSTSUBSCRIPT , italic_E start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT ∈ italic_E. One can show that this is always a left exact presentation, i.e. it satisfies the two conditions in the definition of a dense extension. Indeed, let f:EX:𝑓𝐸𝑋f:E\to Xitalic_f : italic_E → italic_X be a morphism with E\EuScriptE𝐸\EuScript𝐸E\in\EuScript Eitalic_E ∈ italic_E. Consider the pullback of p𝑝pitalic_p and f𝑓fitalic_f in the ambient abelian category Lex(\EuScriptE)Lex\EuScript𝐸{\mathrm{Lex}}(\EuScript E)roman_Lex ( italic_E ). Since p𝑝pitalic_p is an epimorphism, so is its pullback q:YE:𝑞𝑌𝐸q:Y\to Eitalic_q : italic_Y → italic_E. By Proposition 3.10 there exist E\EuScriptEsuperscript𝐸\EuScript𝐸E^{\prime}\in\EuScript Eitalic_E start_POSTSUPERSCRIPT ′ end_POSTSUPERSCRIPT ∈ italic_E and a morphism t:EY:𝑡superscript𝐸𝑌t:E^{\prime}\to Yitalic_t : italic_E start_POSTSUPERSCRIPT ′ end_POSTSUPERSCRIPT → italic_Y such that d:=qtassign𝑑𝑞𝑡d:=qtitalic_d := italic_q italic_t is again an epimorphism. Observe that d𝑑ditalic_d is the desired deflation. The second condition is easily established in the same fashion.

E0subscript𝐸0{E_{0}}italic_E start_POSTSUBSCRIPT 0 end_POSTSUBSCRIPTE1subscript𝐸1{E_{1}}italic_E start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPTX𝑋{X}italic_X00{0}Esuperscript𝐸{E^{\prime}}italic_E start_POSTSUPERSCRIPT ′ end_POSTSUPERSCRIPTY𝑌{Y}italic_YE𝐸{E}italic_Ea𝑎\scriptstyle{a}italic_ap𝑝\scriptstyle{p}italic_pt𝑡\scriptstyle{t}italic_td𝑑\scriptstyle{d}italic_dq𝑞\scriptstyle{q}italic_qg𝑔\scriptstyle{g}italic_gf𝑓\scriptstyle{f}italic_f

Remark 3.12.

It is also possible to show that Y\EuScriptE:\EuScriptELex(\EuScriptE)c:subscript𝑌\EuScript𝐸\EuScript𝐸Lexsuperscript\EuScript𝐸𝑐Y_{\EuScript}E:\EuScript E\to{\mathrm{Lex}}(\EuScript E)^{c}italic_Y start_POSTSUBSCRIPT end_POSTSUBSCRIPT italic_E : italic_E → roman_Lex ( italic_E ) start_POSTSUPERSCRIPT italic_c end_POSTSUPERSCRIPT satisfies the universal property of Theorem 2.6 directly. The proof would be essentially the same as Rump’s proof for Ql(\EuScriptE)subscript𝑄𝑙\EuScript𝐸Q_{l}(\EuScript E)italic_Q start_POSTSUBSCRIPT italic_l end_POSTSUBSCRIPT ( italic_E ) in [13].

Let us now discuss the conditions on \EuScriptE\EuScript𝐸\EuScript Eitalic_E under which it possesses a right/left abelian envelope.

Definition 3.13 (Rump, [14], Definition 3).

Let \EuScriptE\EuScript𝐸\EuScript Eitalic_E be a left exact category. A morphism g:CB:𝑔𝐶𝐵g:C\to Bitalic_g : italic_C → italic_B in \EuScriptE\EuScript𝐸\EuScript Eitalic_E is said to be an Ext-kernel of f:BA:𝑓𝐵𝐴f:B\to Aitalic_f : italic_B → italic_A if fg=0𝑓𝑔0fg=0italic_f italic_g = 0 and for any g:CB:superscript𝑔superscript𝐶𝐵g^{\prime}:C^{\prime}\to Bitalic_g start_POSTSUPERSCRIPT ′ end_POSTSUPERSCRIPT : italic_C start_POSTSUPERSCRIPT ′ end_POSTSUPERSCRIPT → italic_B such that fg=0𝑓superscript𝑔0fg^{\prime}=0italic_f italic_g start_POSTSUPERSCRIPT ′ end_POSTSUPERSCRIPT = 0 there exists a deflation d:DC:𝑑𝐷superscript𝐶d:D\to C^{\prime}italic_d : italic_D → italic_C start_POSTSUPERSCRIPT ′ end_POSTSUPERSCRIPT such gdsuperscript𝑔𝑑g^{\prime}ditalic_g start_POSTSUPERSCRIPT ′ end_POSTSUPERSCRIPT italic_d factors through g𝑔gitalic_g.

C𝐶{C}italic_CB𝐵{B}italic_BA𝐴{A}italic_AEsuperscript𝐸{E^{\prime}}italic_E start_POSTSUPERSCRIPT ′ end_POSTSUPERSCRIPTCsuperscript𝐶{C^{\prime}}italic_C start_POSTSUPERSCRIPT ′ end_POSTSUPERSCRIPTg𝑔\scriptstyle{g}italic_gf𝑓\scriptstyle{f}italic_fd𝑑\scriptstyle{d}italic_dgsuperscript𝑔\scriptstyle{g^{\prime}}italic_g start_POSTSUPERSCRIPT ′ end_POSTSUPERSCRIPT00\scriptstyle{0}

A left exact category \EuScriptE\EuScript𝐸\EuScript Eitalic_E is said to be left Ext-coherent if every morphism in \EuScriptE\EuScript𝐸\EuScript Eitalic_E has an Ext-kernel.

Dually one can define Ext-cokernels and right Ext-coherent categories. Recall that g:CB:𝑔𝐶𝐵g:C\to Bitalic_g : italic_C → italic_B is called a weak kernel of f:BA:𝑓𝐵𝐴f:B\to Aitalic_f : italic_B → italic_A if fg=0𝑓𝑔0fg=0italic_f italic_g = 0 and every gsuperscript𝑔g^{\prime}italic_g start_POSTSUPERSCRIPT ′ end_POSTSUPERSCRIPT with fg=0𝑓superscript𝑔0fg^{\prime}=0italic_f italic_g start_POSTSUPERSCRIPT ′ end_POSTSUPERSCRIPT = 0 factors through g𝑔gitalic_g. An additive category with weak kernels is called left coherent (in the literature the name right coherent is commonly used, but here we consistently follow the conventions of Rump). Unlike Ext-coherence, this property does not depend on an exact structure. Now thanks to the following result we can add another equivalent condition to the list in Corollary 3.5:

Theorem 3.14 (Rump, [14], Proposition 6).

Let \EuScriptE\EuScript𝐸\EuScript Eitalic_E be an exact category. Then Ql(\EuScriptE)subscript𝑄𝑙\EuScript𝐸Q_{l}(\EuScript E)italic_Q start_POSTSUBSCRIPT italic_l end_POSTSUBSCRIPT ( italic_E ) is abelian (i.e. \EuScriptAr(\EuScriptE)\EuScriptsubscript𝐴𝑟\EuScript𝐸\EuScript A_{r}(\EuScript E)italic_A start_POSTSUBSCRIPT italic_r end_POSTSUBSCRIPT ( italic_E ) exists) if and only if \EuScriptE\EuScript𝐸\EuScript Eitalic_E is left Ext-coherent.

Let us state a few immediate corollaries.

Theorem 3.15.
  1. 1)

    Let (\EuScriptE,ϵ)\EuScript𝐸italic-ϵ(\EuScript E,\epsilon)( italic_E , italic_ϵ ) be an exact category, where ϵitalic-ϵ\epsilonitalic_ϵ denotes the class of conflations. Let (\EuScriptE,ϵ)\EuScript𝐸superscriptitalic-ϵ(\EuScript E,\epsilon^{\prime})( italic_E , italic_ϵ start_POSTSUPERSCRIPT ′ end_POSTSUPERSCRIPT ) be an exact category with the same underlying additive category \EuScriptE\EuScript𝐸\EuScript Eitalic_E, but a stronger exact structure, i.e. ϵϵitalic-ϵsuperscriptitalic-ϵ\epsilon\subseteq\epsilon^{\prime}italic_ϵ ⊆ italic_ϵ start_POSTSUPERSCRIPT ′ end_POSTSUPERSCRIPT. If (\EuScriptE,ϵ)\EuScript𝐸italic-ϵ(\EuScript E,\epsilon)( italic_E , italic_ϵ ) possess a right abelian envelope, then so does (\EuScriptE,ϵ)\EuScript𝐸superscriptitalic-ϵ(\EuScript E,\epsilon^{\prime})( italic_E , italic_ϵ start_POSTSUPERSCRIPT ′ end_POSTSUPERSCRIPT ).

  2. 2)

    If \EuScriptE\EuScript𝐸\EuScript Eitalic_E is left coherent, then \EuScriptE\EuScript𝐸\EuScript Eitalic_E has a right abelian envelope (for the case of the split exact structure this is stated in [2], Proposition 4.14).

  3. 3)

    Let \EuScriptE\EuScript𝐸\EuScript Eitalic_E be an additive category with the split exact structure. Then \EuScriptE\EuScript𝐸\EuScript Eitalic_E has a right abelian envelope if and only if it is left coherent.

Proof.
  1. 1)

    Indeed, this is an immediate consequence of Theorem 3.14, since adding more deflations can only create more Ext-kernels.

  2. 2)

    A weak kernel is always an Ext-kernel, so left coherent categories are left Ext-coherent. Another way to see it is by observing that if \EuScriptE\EuScript𝐸\EuScript Eitalic_E is left-coherent, then Ql(\EuScriptE)subscript𝑄𝑙\EuScript𝐸Q_{l}(\EuScript E)italic_Q start_POSTSUBSCRIPT italic_l end_POSTSUBSCRIPT ( italic_E ) is abelian as a quotient of the abelian category mod(\EuScriptE)mod\EuScript𝐸\mathop{\mathrm{mod}}\nolimits(\EuScript E)roman_mod ( italic_E ) by a Serre subcategory.

  3. 3)

    Observe that for \EuScriptE\EuScript𝐸\EuScript Eitalic_E endowed with the split exact structure, Ext-kernels are exactly the same as weak kernels.

4. Examples

Proposition 4.1.
  1. 1)

    Let R𝑅Ritalic_R be a commutative local Cohen-Macauley ring. Then the category of maximal Cohen-Macauley modules MCM(R)MCM𝑅{\mathrm{MCM}}(R)roman_MCM ( italic_R ) has both a right and a left abelian envelope for any exact structure.

  2. 2)

    Let X𝑋Xitalic_X be a noetherian scheme with a resolution property (i.e. every coherent sheaf is a quotient of a locally free one). Let \EuScriptE:=Bun(X)assign\EuScript𝐸Bun𝑋\EuScript E:={\mathrm{Bun}}(X)italic_E := roman_Bun ( italic_X ) be the category of locally free sheaves on X𝑋Xitalic_X, considered as a fully exact subcategory of Coh(X)Coh𝑋{\mathrm{Coh}}(X)roman_Coh ( italic_X ). The category \EuScriptE\EuScript𝐸\EuScript Eitalic_E has a right abelian envelope, namely, Coh(X)Coh𝑋{\mathrm{Coh}}(X)roman_Coh ( italic_X ). The category Bun(X)Bun𝑋{\mathrm{Bun}}(X)roman_Bun ( italic_X ) endowed with any stronger exact structure (e.g. obtained by removing points from X𝑋Xitalic_X) also possess a right abelian envelope.

  3. 3)

    Let X𝑋Xitalic_X be a smooth projective variety with dim(X)=1dimension𝑋1\dim(X)=1roman_dim ( italic_X ) = 1 or 2222. Then the category Bun(X)Bun𝑋{\mathrm{Bun}}(X)roman_Bun ( italic_X ) of locally free sheaves has a right abelian envelope when endowed with any exact structure. On the other hand, for dim(X)3dimension𝑋3\dim(X)\geq 3roman_dim ( italic_X ) ≥ 3 this is not true in general, e.g. for X=3𝑋superscript3X=\mathbb{P}^{3}italic_X = blackboard_P start_POSTSUPERSCRIPT 3 end_POSTSUPERSCRIPT.

  4. 4)

    Let X𝑋Xitalic_X be a smooth projective variety. Consider the subcategory \EuScriptE\EuScript𝐸\EuScript Eitalic_E of Bun(X)Bun𝑋{\mathrm{Bun}}(X)roman_Bun ( italic_X ) consisting of bundles of the form i𝒪X(i)disubscriptdirect-sum𝑖subscript𝒪𝑋superscript𝑖subscript𝑑𝑖\bigoplus_{i}\mathcal{O}_{X}(i)^{d_{i}}⨁ start_POSTSUBSCRIPT italic_i end_POSTSUBSCRIPT caligraphic_O start_POSTSUBSCRIPT italic_X end_POSTSUBSCRIPT ( italic_i ) start_POSTSUPERSCRIPT italic_d start_POSTSUBSCRIPT italic_i end_POSTSUBSCRIPT end_POSTSUPERSCRIPT. One can show that \EuScriptE\EuScript𝐸\EuScript Eitalic_E, unlike Bun(X)Bun𝑋{\mathrm{Bun}}(X)roman_Bun ( italic_X ), is always left coherent, hence it has a right abelian envelope with any exact structure. In particular,

    \EuScriptAr(\EuScriptE,ϵsplit)=mod(\EuScriptE)modfp(RX)Cohk(𝔸X)\EuScriptsubscript𝐴𝑟\EuScript𝐸subscriptitalic-ϵsplitmod\EuScript𝐸subscriptmod𝑓𝑝subscript𝑅𝑋subscriptCohsuperscriptk𝔸𝑋\EuScript A_{r}(\EuScript E,\epsilon_{\text{split}})=\mathop{\mathrm{mod}}% \nolimits(\EuScript E)\cong\mathop{\mathrm{mod}}\nolimits_{fp}(R_{X})\cong{% \mathrm{Coh}}_{\mathrm{k}^{*}}(\mathbb{A}X)italic_A start_POSTSUBSCRIPT italic_r end_POSTSUBSCRIPT ( italic_E , italic_ϵ start_POSTSUBSCRIPT split end_POSTSUBSCRIPT ) = roman_mod ( italic_E ) ≅ roman_mod start_POSTSUBSCRIPT italic_f italic_p end_POSTSUBSCRIPT ( italic_R start_POSTSUBSCRIPT italic_X end_POSTSUBSCRIPT ) ≅ roman_Coh start_POSTSUBSCRIPT roman_k start_POSTSUPERSCRIPT ∗ end_POSTSUPERSCRIPT end_POSTSUBSCRIPT ( blackboard_A italic_X )

    where RXsubscript𝑅𝑋R_{X}italic_R start_POSTSUBSCRIPT italic_X end_POSTSUBSCRIPT denotes the homogeneous coordinate ring of X𝑋Xitalic_X and Cohk(𝔸X)subscriptCohsuperscript𝑘𝔸𝑋{\mathrm{Coh}}_{k^{*}}(\mathbb{A}X)roman_Coh start_POSTSUBSCRIPT italic_k start_POSTSUPERSCRIPT ∗ end_POSTSUPERSCRIPT end_POSTSUBSCRIPT ( blackboard_A italic_X ) the category of equivariant sheaves on the affine cone over X𝑋Xitalic_X. Note that the right abelian envelope of \EuScriptE\EuScript𝐸\EuScript Eitalic_E with the exact structure inherited from Coh(X)Coh𝑋{\mathrm{Coh}}(X)roman_Coh ( italic_X ) is again Coh(X)Coh𝑋{\mathrm{Coh}}(X)roman_Coh ( italic_X ) because any coherent sheaf is a quotient of a vector bundle from \EuScriptE\EuScript𝐸\EuScript Eitalic_E (see [2], Theorem 4.11).

Proof.
  1. 1)

    This follows from Theorem 3.15.2) and the fact that MCM(R)MCM𝑅{\mathrm{MCM}}(R)roman_MCM ( italic_R ) has weak kernels and cokernels. The latter is established by Holm in [6], Proposition 4.2, basing on Auslander-Buchweitz’s maximal Cohen–Macaulay approximations [1].

  2. 2)

    The fact that \EuScriptAr(\EuScriptE)=Coh(X)\EuScriptsubscript𝐴𝑟\EuScript𝐸Coh𝑋\EuScript A_{r}(\EuScript E)={\mathrm{Coh}}(X)italic_A start_POSTSUBSCRIPT italic_r end_POSTSUBSCRIPT ( italic_E ) = roman_Coh ( italic_X ) when \EuScriptE\EuScript𝐸\EuScript Eitalic_E is endowed with the geometric exact structure is Corollary 4.12 of [2]. The rest follows from Theorem 3.15.1).

  3. 3)

    For dim(X)=1dimension𝑋1\dim(X)=1roman_dim ( italic_X ) = 1 or 2222, the category Bun(X)Bun𝑋{\mathrm{Bun}}(X)roman_Bun ( italic_X ) has kernels, hence it possesses a right abelian envelope for any exact structure. On the other hand, the category Bun(3)Bunsuperscript3{\mathrm{Bun}}(\mathbb{P}^{3})roman_Bun ( blackboard_P start_POSTSUPERSCRIPT 3 end_POSTSUPERSCRIPT ) does not have weak kernels (Bondal-Pavlov, via private communication). Hence, by Theorem 3.15.3), it has no right abelian envelope when considered with the split exact structure.

Remark 4.2.

I would like to thank Alexey Bondal for drawing my attention to the fact that Bun(3)Bunsuperscript3{\mathrm{Bun}}(\mathbb{P}^{3})roman_Bun ( blackboard_P start_POSTSUPERSCRIPT 3 end_POSTSUPERSCRIPT ) does not have weak kernels.

References

  • [1] Maurice Auslander and Ragnar-Olaf Buchweitz, The homological theory of maximal Cohen-Macaulay approximations, no. 38, 1989, Colloque en l’honneur de Pierre Samuel (Orsay, 1987), pp. 5–37.
  • [2] Agnieszka Bodzenta and Alexey Bondal, Abelian envelopes of exact categories and highest weight categories, Mathematische Zeitschrift 308 (2024), no. 8.
  • [3] Siegfried Breitsprecher, Lokal endlich präsentierbare Grothendieck-Kategorien, Mitt. Math. Sem. Giessen 85 (1970), 1–25. MR 262330
  • [4] William Crawley-Boevey, Locally finitely presented additive categories, Comm. Algebra 22 (1994), no. 5, 1641–1674.
  • [5] Ruben Henrard, Sondre Kvamme, and Adam-Christiaan van Roosmalen, Auslander’s formula and correspondence for exact categories, Adv. Math. 401 (2022), Paper No. 108296, 65.
  • [6] Henrik Holm, The category of maximal Cohen-Macaulay modules as a ring with several objects, Mediterr. J. Math. 13 (2016), no. 3, 885–898.
  • [7] Dmitry Kaledin and Wendy Lowen, Cohomology of exact categories and (non-)additive sheaves, Advances in Mathematics 272 (2015), 652–698.
  • [8] Bernhard Keller, Chain complexes and stable categories, Manuscripta Math. 67 (1990), no. 4, 379–417.
  • [9] Yasuaki Ogawa, Auslander’s defects over extriangulated categories: an application for the general heart construction, J. Math. Soc. Japan 73 (2021), no. 4, 1063–1089.
  • [10] Daniel Quillen, Higher algebraic K𝐾Kitalic_K-theory. I, Algebraic K𝐾Kitalic_K-theory, I: Higher K𝐾Kitalic_K-theories (Proc. Conf., Battelle Memorial Inst., Seattle, Wash., 1972), Lecture Notes in Math., vol. Vol. 341, Springer, Berlin-New York, 1973, pp. 85–147.
  • [11] Wolfgang Rump, One-sided Grothendieck quotients, Arch. Math. (Basel) 89 (2007), no. 2, 131–142.
  • [12] by same author, Locally finitely presented categories of sheaves, J. Pure Appl. Algebra 214 (2010), no. 2, 177–186.
  • [13] by same author, The abelian closure of an exact category, Journal of Pure and Applied Algebra 224 (2020), no. 10.
  • [14] by same author, Exact categories and infinite tilting, Comm. Algebra 49 (2021), no. 7, 3034–3065.
  • [15] Marco Schlichting, Negative K𝐾Kitalic_K-theory of derived categories, Math. Z. 253 (2006), no. 1, 97–134.