Introduction
According to the discussion in [5 Section 7.11] , there are two kinds of quasi-coherent module objects over an ind-scheme. In the terminology of the present author’s book [33 Chapters 2–3] , they are called the quasi-coherent torsion sheaves and the pro-quasi-coherent pro-sheaves . In this paper, we study the pro-quasi-coherent pro-sheaves.
The category of arbitrary pro-quasi-coherent pro-sheaves on ind-scheme or a formal scheme is not well-behaved from the homological algebra standpoint [5 Section 7.11.3] , [33 Examples 3.1 and 3.8(1)] , [30 Example 2.7(1)] . The first approximate explanation is that the functors of inverse image of quasi-coherent sheaves with respect to closed immersions of schemes are not exact (and the functor of projective limit of countable sequences of abelian groups is not exact, either). In this paper, we restrict ourselves to the category of flat pro-quasi-coherent pro-sheaves , which is a well-behaved exact category where one can do homological algebra.
The identification of flat pro-quasi-coherent pro-sheaves on a strict ind-affine ℵ 0 \aleph_{0} ℵ 0 -ind-scheme U = Spf R \mathcal{U} = \operatorname{Spf} R U = Spf R with flat contramodules over the related topological commutative ring R R R was pointed out in [33 Example 3.8(2)] . It became a geometric motivation for the study of flat contramodules over complete, separated topological rings with a countable base of neighborhoods of zero consisting of open two-sided ideals in the paper [35 Sections 7–13] (see [35 Example 8.9] ). In the present paper, go back to the geometric language and generalize some contramodule-theoretic results of [35 ] (as well as [29 Appendix E] , [42 ] , etc.) to flat pro-quasi-coherent pro-sheaves over non-ind-affine ind-schemes.
Ind-schemes (and morphisms of ind-schemes) are represented by filtered inductive systems of schemes, with the transition morphisms usually presumed to be closed immersions. Various classes of ind-schemes and morphisms of ind-schemes are conventionally defined by imposing conditions on the filtered inductive systems representing such ind-schemes and morphisms.
A subtle issue in the foundations of the theory arises immediately. Suppose that we are given an ind-scheme X \mathcal{X} X , represented by a directed diagram of schemes ( Y δ ′ → Y δ ′ ′ ) δ ′ ≤ δ ′ ′ ∈ Δ (Y_{\delta'} \to Y_{\delta''})_{\delta'\leq\delta''\in\Delta} ( Y δ ′ → Y δ ′′ ) δ ′ ≤ δ ′′ ∈ Δ , and this directed diagram has some good property P \mathbf{P} P . Suppose further that the same ind-scheme X \mathcal{X} X is represented by another directed diagram of schemes ( Z ϵ ′ → Z ϵ ′ ′ ) ϵ ′ ≤ ϵ ′ ′ ∈ E (Z_{\epsilon'} \to Z_{\epsilon''})_{\epsilon'\leq\epsilon''\in E} ( Z ϵ ′ → Z ϵ ′′ ) ϵ ′ ≤ ϵ ′′ ∈ E , and that second directed diagram representing X \mathcal{X} X has another good property Q \mathbf{Q} Q . Can one find a directed diagram ( X γ ′ → X γ ′ ′ ) γ ′ ≤ γ ′ ′ ∈ Γ (X_{\gamma'} \to X_{\gamma''})_{\gamma'\leq\gamma''\in\Gamma} ( X γ ′ → X γ ′′ ) γ ′ ≤ γ ′′ ∈ Γ representing X \mathfrak{X} X that has both the good properties P \mathbf{P} P and Q \mathbf{Q} Q ?
The same question applies to directed diagrams of morphisms of schemes representing morphisms of ind-schemes, and their properties. Without a positive answer to this kind of questions, requiring an ind-scheme or a morphism of ind-schemes to be “both ind-P and ind-Q” becomes ambiguous at best, or useless at worst.
In Section 2 of the present paper, which builds upon [33 Chapter 1] , we discuss the foundations of the notion of an ind-scheme, concentrating in particular on the class of questions described above. The exact category X \mathfrak{X} X –Proqc f l \operatorname{Proqc}_{\mathrm{fl}} Proqc fl of flat pro-quasi-coherent pro-sheaves on an ind-scheme X \mathfrak{X} X and its full subcategory X \mathfrak{X} X –Proqc v f l ⊂ X \operatorname{Proqc}_{\mathrm{vfl}} \subset\mathfrak{X} Proqc vfl ⊂ X –Proqc f l \operatorname{Proqc}_{\mathrm{fl}} Proqc fl of very flat pro-quasi-coherent pro-sheaves are defined in the next Section 3.
In the subsequent Sections 4–6, we recall various techniques from (set-theoretical and other) homological algebra and apply them to flat pro-quasi-coherent pro-sheaves. In particular, we prove that the category X \mathfrak{X} X –Proqc f l \operatorname{Proqc}_{\mathrm{fl}} Proqc fl is accessible (ℵ 1 \aleph_1 ℵ 1 -accessible for an ind-quasi-compact ind-quasi-separated ℵ 0 \aleph_0 ℵ 0 -ind-scheme X \mathfrak{X} X ), and that the classes of flat and very flat pro-quasi-coherent pro-sheaves X \mathfrak{X} X –Proqc f l \operatorname{Proqc}_{\mathrm{fl}} Proqc fl and X \mathfrak{X} X –Proqc v f l \operatorname{Proqc}_{\mathrm{vfl}} Proqc vfl are deconstructible. We also spell out an argument proving Ext 1 \operatorname{Ext}^{1} Ext 1 -orthogonality of flat pro-quasi-coherent pro-sheaves with certain levelwise conditions along the directed diagram of schemes representing X \mathfrak{X} X , which was invented in the recent book manuscript [41 Section 8.5] . All of these results are also extended to some classes of complexes of flat and very flat pro-quasi-coherent pro-sheaves.
Acknowledgement
Acknowledgement. I am grateful to Jan Šťovíček and Michal Hrbek for helpful discussions. The author is supported by the GAČR project 26-22734S and the Institute of Mathematics, Czech Academy of Sciences (research plan RVO: 67985840).
Preliminaries on Ind-Schemes
The basics of ind-schemes
We can recommend the notes [53 ] , the unfinished manuscript [5 Section 7.11] , and the book [33 ] as the reference sources on ind-schemes.
In this section, we consider diagrams indexed by directed posets. The terms “filtered inductive system”, “direct system”, and “directed diagram” are used as synonymous. A substantial discussion of the difference between and the essential equivalence of the directed and the filtered colimits can be found in the book [1 Theorem 1.5] and the paper [2 ] .
Given a category K \mathsf{K} K , the category Ind K \operatorname{Ind}\mathsf{K} Ind K of ind-objects in K \mathsf{K} K is defined as the following full subcategory of the category Sets K o p \operatorname{Sets}^{\mathsf{K}^{\mathrm{op}}} Sets K op of contravariant functors from K \mathsf{K} K to the category of sets Sets \operatorname{Sets} Sets . To every object K ∈ K K \in\mathsf{K} K ∈ K , the representable functor Mor K ( − , K ) : K o p ⟶ Sets \operatorname{Mor}_{\mathsf{K}}(-,K)\colon\mathsf{K}^{\mathrm{op}} \longrightarrow\operatorname{Sets} Mor K ( − , K ) : K op ⟶ Sets is assigned (where Mor K ( − , − ) \operatorname{Mor}_{\mathsf{K}}(-,-) Mor K ( − , − ) denotes the set of morphisms in K \mathsf{K} K ). The full subcategory Ind K ⊂ Sets K o p \operatorname{Ind}\mathsf{K} \subset\operatorname{Sets}^{\mathsf{K}^{\mathrm{op}}} Ind K ⊂ Sets K op consists of all the filtered inductive limits of representable functors. Let Γ \Gamma Γ be a directed poset and ( K γ → K δ ) γ < δ ∈ Γ (K_\gamma\to K_\delta)_{\gamma<\delta\in\Gamma} ( K γ → K δ ) γ < δ ∈ Γ be a Γ \Gamma Γ -indexed diagram in K \mathbf{K} K . Then the corresponding ind-object, denoted by “l i m → γ ∈ Γ K γ \varinjlim_{\gamma\in\Gamma} K_\gamma lim γ ∈ Γ K γ ” ∈ Ind K \in\operatorname{Ind}\mathbf{K} ∈ Ind K , represents the functor K o p → S e t s \mathbf{K}^{\mathrm{op}} \to\mathbf{Sets} K op → Sets given by the rule L ↦ l i m → γ ∈ Γ Mor K ( L , K γ ) L \mapsto\varinjlim_{\gamma\in\Gamma}\operatorname{Mor}_{\mathbf{K}}(L,K_\gamma) L ↦ lim γ ∈ Γ Mor K ( L , K γ ) . Accordingly, the sets of morphisms in Ind K \operatorname{Ind}\mathbf{K} Ind K are computed by the formula
Mor Ind K ( “ l i m → δ ∈ Δ L δ ” , “ l i m → γ ∈ Γ K γ ” ) = l i m ← δ ∈ Δ l i m → γ ∈ Γ Mor K ( L δ , K γ ) (1) \operatorname{Mor}_{\operatorname{Ind}\mathbf{K}}\left(\text{“}\varinjlim_{\delta\in\Delta} L_\delta\text{”},\text{ “}\varinjlim_{\gamma\in\Gamma} K_\gamma\text{”}\right)=\varprojlim_{\delta\in\Delta}\varinjlim_{\gamma\in\Gamma}\operatorname{Mor}_{\mathbf{K}}(L_\delta,K_\gamma) \tag*{(1)} Mor Ind K ( “ δ ∈ Δ lim L δ ” , “ γ ∈ Γ lim K γ ” ) = δ ∈ Δ lim γ ∈ Γ lim Mor K ( L δ , K γ ) (1)
for any two directed posets Δ \Delta Δ and Γ \Gamma Γ and any two diagrams ( L δ ∈ K ) δ ∈ Δ (L_\delta\in\mathbf{K})_{\delta\in\Delta} ( L δ ∈ K ) δ ∈ Δ and ( K γ ∈ K ) γ ∈ Γ (K_\gamma\in\mathbf{K})_{\gamma\in\Gamma} ( K γ ∈ K ) γ ∈ Γ in the category K \mathbf{K} K .
The standard references on ind-objects in categories are [3 Section I.8] and [22 Section 6] . See also [33 Section 1.1] , [32 Section 9] , and/or [20 ] .
C S c h o p ⟶ S e t s \mathrm{CSch}^{\mathrm{op}} \longrightarrow\mathrm{Sets} CSch op ⟶ Sets given by the same rule Z ⟼ l i m → γ ∈ Γ MorSch ( Z , X γ ) Z \longmapsto\varinjlim_{\gamma\in\Gamma} \operatorname{MorSch}(Z,X_\gamma) Z ⟼ lim γ ∈ Γ MorSch ( Z , X γ ) . Then the category of ind-schemes is equivalent to the category of functors C S c h o p ⟶ S e t s \mathrm{CSch}^{\mathrm{op}} \longrightarrow\mathrm{Sets} CSch op ⟶ Sets represented by directed diagrams of schemes.
The point is that, for any scheme X X X , the functor C S c h o p ⟶ S e t s \mathrm{CSch}^{\mathrm{op}} \longrightarrow\mathrm{Sets} CSch op ⟶ Sets represented by X X X is a sheaf in the Zariski topology on the category of concentrated schemes. Moreover, the filtered inductive limit functors in the categories of presheaves of sets and sheaves of sets on C S c h \mathrm{CSch} CSch agree, due to the finite nature of the Zariski topology on concentrated schemes. Finally, the category of sheaves of sets on C S c h \mathrm{CSch} CSch is equivalent to the category of sheaves of sets on A f f S c h \mathrm{AffSch} AffSch . This argument can be found in [33 Section 1.2] .
It follows from the discussion in the preceding two paragraphs that the full subcategory of ind-schemes representable by filtered inductive systems of concentrated schemes is equivalent to the category Ind ( C S c h ) \operatorname{Ind}(\mathrm{CSch}) Ind ( CSch ) of ind-objects in C S c h \mathrm{CSch} CSch . Ind-schemes representable by filtered inductive systems of concentrated schemes are said to be ind-concentrated (or ind-quasi-compact ind-quasi-separated ). In this paper, we are only interested in ind-concentrated ind-schemes.
An ind-scheme is said to be strict if it is representable by a filtered inductive system ( X γ ⟶ X δ ) γ < δ ∈ Γ (X_\gamma\longrightarrow X_\delta)_{\gamma<\delta\in\Gamma} ( X γ ⟶ X δ ) γ < δ ∈ Γ whose transition morphisms i γ δ : X γ ⟶ X δ i_{\gamma\delta}:X_\gamma\longrightarrow X_\delta i γ δ : X γ ⟶ X δ are closed immersions of schemes. An ℵ 0 \aleph_0 ℵ 0 -ind-scheme is an ind-scheme representable by a direct system ( X γ ) γ ∈ Γ (X_\gamma)_{\gamma\in\Gamma} ( X γ ) γ ∈ Γ indexed by a countable directed poset Γ \Gamma Γ (or equivalently, by the directed poset of positive integers Γ = Z ≥ 1 \Gamma=\mathbb{Z}_{\ge1} Γ = Z ≥ 1 ). In this paper, we are only interested in strict ℵ 0 \aleph_0 ℵ 0 -ind-schemes.
The previous two paragraphs raise an immediate question. Is imposing such and other similar conditions on a filtered inductive system ( X γ ⟶ X δ ) γ < δ ∈ Γ (X_\gamma\longrightarrow X_\delta)_{\gamma<\delta\in\Gamma} ( X γ ⟶ X δ ) γ < δ ∈ Γ separately equivalent to imposing them simultaneously ? Given an ind-scheme X \mathfrak{X} X , suppose that it is represented by a filtered inductive system of schemes ( Y δ ) γ ∈ Δ (Y_\delta)_{\gamma\in\Delta} ( Y δ ) γ ∈ Δ having some good properties, and the same ind-scheme X \mathfrak{X} X is also represented by a filtered inductive system of schemes ( Z λ ) γ ∈ Λ (Z_\lambda)_{\gamma\in\Lambda} ( Z λ ) γ ∈ Λ having some other good properties. Does it follow that one can represent X \mathfrak{X} X by a filtered inductive system ( X γ ) γ ∈ Γ (X_\gamma)_{\gamma\in\Gamma} ( X γ ) γ ∈ Γ combining both sets of good properties at the same time?
We continue the discussion of this issue started in [33 Section 1.2] . It turns out that some of such questions can be answered positively with the naïve definition of an ind-scheme as an object of Ind ( S c h ) \operatorname{Ind}(\mathrm{Sch}) Ind ( Sch ) , but get negative answers with the more advanced definition of an ind-scheme as an object of S e t s A f f S c h o p \mathrm{Sets}^{\mathrm{AffSch}^{\mathrm{op}}} Sets AffSch op or S e t s C S c h o p \mathrm{Sets}^{\mathrm{CSch}^{\mathrm{op}}} Sets CSch op .
Let us start with restating a particular lemma from The Stacks project, which is highly relevant to the basics of ind-schemes.
Lemma 2.2. Let X ⟶ Y ⟶ Z X \longrightarrow Y \longrightarrow Z X ⟶ Y ⟶ Z be morphisms of schemes.
(a) Assume that the composition X ⟶ Z X \longrightarrow Z X ⟶ Z is a locally closed immersion. Then the morphism X ⟶ Y X \longrightarrow Y X ⟶ Y is a locally closed immersion.
(b) Assume that the composition X ⟶ Z X \longrightarrow Z X ⟶ Z is a quasi-compact locally closed immersion, while the morphism Y ⟶ Z Y \longrightarrow Z Y ⟶ Z is quasi-separated. Then X ⟶ Y X \longrightarrow Y X ⟶ Y is a quasi-compact locally closed immersion.
(c) Assume that the composition X ⟶ Z X \longrightarrow Z X ⟶ Z is a closed immersion, while the morphism Y ⟶ Z Y \longrightarrow Z Y ⟶ Z is separated. Then X ⟶ Y X \longrightarrow Y X ⟶ Y is a closed immersion.
(d) In particular, if the composition X ⟶ Z X \longrightarrow Z X ⟶ Z is a closed immersion and Y Y Y is a quasi-separated scheme, then the morphism X ⟶ Y X \longrightarrow Y X ⟶ Y is a quasi-compact locally closed immersion.
(e) In particular, if the composition X ⟶ Z X \longrightarrow Z X ⟶ Z is a closed immersion and Y Y Y is a separated scheme, then the morphism X ⟶ Y X \longrightarrow Y X ⟶ Y is a closed immersion.
Proof. Parts (a–c) are [21 Definition Tag 01IO and Lemma Tag 07RK] . Parts (a) and (c) were also restated in [33 Lemma 1.2(a–b)] . Part (d) follows from (b), since any closed immersion of schemes is an affine (hence quasi-compact) morphism, and any morphism from a quasi-separated scheme is quasi-separated [21 Lemma Tag 01KV] . Part (e) similarly follows (c); see [33 Lemma 1.2(c)] . □
Corollary 2.3. (a) Suppose that l i m → δ ∈ Δ Y δ ≃ l i m → λ ∈ Λ Z λ \underset{\delta\in\Delta}{\varinjlim} Y_\delta\simeq\underset{\lambda\in\Lambda}{\varinjlim} Z_\lambda δ ∈ Δ lim Y δ ≃ λ ∈ Λ lim Z λ in Ind ( Sch ) \operatorname{Ind}(\operatorname{Sch}) Ind ( Sch ) , where ( Y δ ) δ ∈ Δ (Y_\delta)_{\delta\in\Delta} ( Y δ ) δ ∈ Δ is a direct system of quasi-separated schemes, while ( Z λ ) λ ∈ Λ (Z_\lambda)_{\lambda\in\Lambda} ( Z λ ) λ ∈ Λ is a direct system of closed immersions of schemes. Then all the schemes Z λ Z_\lambda Z λ are quasi-separated.
(b) Suppose that l i m → δ ∈ Δ Y δ ≃ l i m → λ ∈ Λ Z λ \underset{\delta\in\Delta}{\varinjlim} Y_\delta\simeq\underset{\lambda\in\Lambda}{\varinjlim} Z_\lambda δ ∈ Δ lim Y δ ≃ λ ∈ Λ lim Z λ in Ind ( Sch ) \operatorname{Ind}(\operatorname{Sch}) Ind ( Sch ) , where ( Y δ ) δ ∈ Δ (Y_\delta)_{\delta\in\Delta} ( Y δ ) δ ∈ Δ is a direct system of semi-separated schemes, while ( Z λ ) λ ∈ Λ (Z_\lambda)_{\lambda\in\Lambda} ( Z λ ) λ ∈ Λ is a direct system of closed immersions of schemes. Then all the schemes Z λ Z_\lambda Z λ are semi-separated.
(c) Suppose that l i m → δ ∈ Δ Y δ ≃ l i m → λ ∈ Λ Z λ \underset{\delta\in\Delta}{\varinjlim} Y_\delta\simeq\underset{\lambda\in\Lambda}{\varinjlim} Z_\lambda δ ∈ Δ lim Y δ ≃ λ ∈ Λ lim Z λ in Ind ( Sch ) \operatorname{Ind}(\operatorname{Sch}) Ind ( Sch ) , where ( Y δ ) δ ∈ Δ (Y_\delta)_{\delta\in\Delta} ( Y δ ) δ ∈ Δ is a direct system of Noetherian schemes, while ( Z λ ) λ ∈ Λ (Z_\lambda)_{\lambda\in\Lambda} ( Z λ ) λ ∈ Λ is a direct system of closed immersions of schemes. Then all the schemes Z λ Z_\lambda Z λ are Noetherian.
(d) Suppose that l i m → δ ∈ Δ Y δ ≃ l i m → λ ∈ Λ Z λ \underset{\delta\in\Delta}{\varinjlim} Y_\delta\simeq\underset{\lambda\in\Lambda}{\varinjlim} Z_\lambda δ ∈ Δ lim Y δ ≃ λ ∈ Λ lim Z λ in Ind ( Sch ) \operatorname{Ind}(\operatorname{Sch}) Ind ( Sch ) , where ( Y δ ) δ ∈ Δ (Y_\delta)_{\delta\in\Delta} ( Y δ ) δ ∈ Δ is a direct system of quasi-compact quasi-separated schemes, while ( Z λ ) λ ∈ Λ (Z_\lambda)_{\lambda\in\Lambda} ( Z λ ) λ ∈ Λ is a direct system of closed immersions of schemes. Then all the schemes Z λ Z_\lambda Z λ are quasi-compact and quasi-separated.
(e) Suppose that l i m → δ ∈ Δ Y δ ≃ l i m → λ ∈ Λ Z λ \underset{\delta\in\Delta}{\varinjlim} Y_\delta\simeq\underset{\lambda\in\Lambda}{\varinjlim} Z_\lambda δ ∈ Δ lim Y δ ≃ λ ∈ Λ lim Z λ in Ind ( Sch ) \operatorname{Ind}(\operatorname{Sch}) Ind ( Sch ) , where ( Y δ ) δ ∈ Δ (Y_\delta)_{\delta\in\Delta} ( Y δ ) δ ∈ Δ is a direct system of affine schemes, while ( Z λ ) λ ∈ Λ (Z_\lambda)_{\lambda\in\Lambda} ( Z λ ) λ ∈ Λ is a direct system of closed immersions of schemes. Then all the schemes Z λ Z_\lambda Z λ are affine.
Proof. Put X = l i m → δ ∈ Δ Y δ = l i m → λ ∈ Λ Z λ ∈ Ind ( Sch ) X = \underset{\delta\in\Delta}{\varinjlim} Y_\delta= \underset{\lambda\in\Lambda}{\varinjlim} Z_\lambda\in\operatorname{Ind}(\operatorname{Sch}) X = δ ∈ Δ lim Y δ = λ ∈ Λ lim Z λ ∈ Ind ( Sch ) . Let λ ′ ∈ Λ \lambda' \in\Lambda λ ′ ∈ Λ be an index. By the definition of the category Ind ( Sch ) \operatorname{Ind}(\operatorname{Sch}) Ind ( Sch ) , the natural morphism Z λ ′ ⟶ X Z_{\lambda'} \longrightarrow X Z λ ′ ⟶ X factorizes through the natural morphism Y δ ⟶ X Y_\delta\longrightarrow X Y δ ⟶ X for some δ ∈ Δ \delta\in\Delta δ ∈ Δ , while the morphism Y δ ⟶ X Y_\delta\longrightarrow X Y δ ⟶ X factorizes through the natural morphism Z λ ′ ′ ⟶ X Z_{\lambda''} \longrightarrow X Z λ ′′ ⟶ X for some λ ′ ′ ∈ Λ \lambda'' \in\Lambda λ ′′ ∈ Λ . Furthermore, all closed immersions of schemes are monomorphisms in Sch \operatorname{Sch} Sch ; so ( Z λ ) λ ∈ Λ (Z_\lambda)_{\lambda\in\Lambda} ( Z λ ) λ ∈ Λ is a direct system of monomorphisms in Sch \operatorname{Sch} Sch , and it follows that Z λ ′ ′ ⟶ X Z_{\lambda''} \longrightarrow X Z λ ′′ ⟶ X is a monomorphism in Ind ( Sch ) \operatorname{Ind}(\operatorname{Sch}) Ind ( Sch ) . Hence the composition Z λ ′ ⟶ Y δ ⟶ Z λ ′ ′ Z_{\lambda'} \longrightarrow Y_\delta\longrightarrow Z_{\lambda''} Z λ ′ ⟶ Y δ ⟶ Z λ ′′ is equal to the transition morphism i λ ′ λ ′ ′ : Z λ ′ ⟶ Z λ ′ ′ i_{\lambda'\lambda''} : Z_{\lambda'} \longrightarrow Z_{\lambda''} i λ ′ λ ′′ : Z λ ′ ⟶ Z λ ′′ in the direct system ( Z λ ) λ ∈ Λ (Z_\lambda)_{\lambda\in\Lambda} ( Z λ ) λ ∈ Λ .
By Lemma 2.2(a), it follows that Z λ ′ ⟶ Y δ Z_{\lambda'} \longrightarrow Y_\delta Z λ ′ ⟶ Y δ is a locally closed immersion of schemes. Hence, if Y δ Y_\delta Y δ is a quasi-separated scheme, then so is Z λ ′ Z_{\lambda'} Z λ ′ . For the same reason, if Y δ Y_\delta Y δ is a semi-separated scheme, then so is Z λ ′ Z_{\lambda'} Z λ ′ . Similarly, if Y δ Y_\delta Y δ is a Noetherian scheme, then so is Z λ ′ Z_{\lambda'} Z λ ′ . Furthermore, if Y δ Y_\delta Y δ is a quasi-separated scheme, then Z λ ′ ⟶ Y δ Z_{\lambda'} \longrightarrow Y_\delta Z λ ′ ⟶ Y δ is a quasi-compact morphism by Lemma 2.2(d). Hence, if Y δ Y_\delta Y δ is a quasi-compact quasi-separated scheme, then Z λ ′ Z_{\lambda'} Z λ ′ is a quasi-compact quasi-separated scheme, too. Finally, if Y δ Y_\delta Y δ is an affine scheme, then Z λ ′ ⟶ Y δ Z_{\lambda'} \longrightarrow Y_\delta Z λ ′ ⟶ Y δ is a closed immersion by Lemma 2.2(e). Hence Z λ ′ Z_{\lambda'} Z λ ′ is an affine scheme, too. □
Parts (b), (c), and (e) of Corollary 2.3 are also applicable to direct systems of concentrated schemes and isomorphisms in Ind ( CSch ) \operatorname{Ind}(\operatorname{CSch}) Ind ( CSch ) . Notice that, for any full subcategory L ⊂ K \mathsf{L} \subset\mathsf{K} L ⊂ K , the category of ind-objects Ind L \operatorname{Ind}\mathsf{L} Ind L is a full subcategory in Ind K \operatorname{Ind}\mathsf{K} Ind K . In particular, Ind ( CSch ) \operatorname{Ind}(\operatorname{CSch}) Ind ( CSch ) is a full subcategory in Ind ( Sch ) \operatorname{Ind}(\operatorname{Sch}) Ind ( Sch ) ; so the assertions specialize to the case of concentrated schemes automatically.
Lemma 2.4. (a) Suppose that l i m → δ ∈ Δ Y δ ≃ l i m → λ ∈ Λ Z λ \varinjlim_{\delta\in\Delta} Y_\delta\simeq\varinjlim_{\lambda\in\Lambda} Z_\lambda lim δ ∈ Δ Y δ ≃ lim λ ∈ Λ Z λ in Ind ( Sch ) \operatorname{Ind}(\operatorname{Sch}) Ind ( Sch ) , where ( Y δ ) δ ∈ Δ (Y_\delta)_{\delta\in\Delta} ( Y δ ) δ ∈ Δ is a direct system of schemes indexed by a countable directed poset Δ \Delta Δ , while ( Z λ ) λ ∈ Λ (Z_\lambda)_{\lambda\in\Lambda} ( Z λ ) λ ∈ Λ is a direct system of closed immersions of schemes. Then there exists a countable directed subposet Λ ′ ⊂ Λ \Lambda' \subset\Lambda Λ ′ ⊂ Λ such that the induced morphism of ind-objects l i m → λ ′ ∈ Λ ′ Z λ ′ ⟶ l i m → λ ∈ Λ Z λ \varinjlim_{\lambda' \in\Lambda'} Z_{\lambda'} \longrightarrow\varinjlim_{\lambda\in\Lambda} Z_\lambda lim λ ′ ∈ Λ ′ Z λ ′ ⟶ lim λ ∈ Λ Z λ is an isomorphism in Ind ( Sch ) \operatorname{Ind}(\operatorname{Sch}) Ind ( Sch ) .
(b) Suppose that l i m → δ ∈ Δ Y δ ≃ l i m → λ ∈ Λ Z λ \varinjlim_{\delta\in\Delta} Y_\delta\simeq\varinjlim_{\lambda\in\Lambda} Z_\lambda lim δ ∈ Δ Y δ ≃ lim λ ∈ Λ Z λ in Ind ( CSch ) \operatorname{Ind}(\operatorname{CSch}) Ind ( CSch ) , where ( Y δ ) δ ∈ Δ (Y_\delta)_{\delta\in\Delta} ( Y δ ) δ ∈ Δ is a direct system of concentrated schemes indexed by a countable directed poset Δ \Delta Δ , while ( Z λ ) λ ∈ Λ (Z_\lambda)_{\lambda\in\Lambda} ( Z λ ) λ ∈ Λ is a direct system of closed immersions of concentrated schemes. Then there exists a countable directed subposet Λ ′ ⊂ Λ \Lambda' \subset\Lambda Λ ′ ⊂ Λ such that the induced morphism of ind-objects l i m → λ ′ ∈ Λ ′ Z λ ′ ⟶ l i m → λ ∈ Λ Z λ \varinjlim_{\lambda' \in\Lambda'} Z_{\lambda'} \longrightarrow\varinjlim_{\lambda\in\Lambda} Z_\lambda lim λ ′ ∈ Λ ′ Z λ ′ ⟶ lim λ ∈ Λ Z λ is an isomorphism in Ind ( CSch ) \operatorname{Ind}(\operatorname{CSch}) Ind ( CSch ) .
Proof. Quite generally, suppose that ( Y δ ) δ ∈ Δ (Y_\delta)_{\delta\in\Delta} ( Y δ ) δ ∈ Δ is a direct system of objects indexed by a countable directed poset Δ \Delta Δ in a category K \mathsf{K} K , while ( Z λ ) λ ∈ Λ (Z_\lambda)_{\lambda\in\Lambda} ( Z λ ) λ ∈ Λ is a direct system of monomorphisms in K \mathsf{K} K , and there is an isomorphism l i m → δ ∈ Δ Y δ ≃ X = l i m → λ ∈ Λ Z λ \varinjlim_{\delta\in\Delta} Y_\delta\simeq\mathfrak{X} = \varinjlim_{\lambda\in\Lambda} Z_\lambda lim δ ∈ Δ Y δ ≃ X = lim λ ∈ Λ Z λ in Ind ( K ) \operatorname{Ind}(\mathsf{K}) Ind ( K ) . Then there exists a countable directed subposet Λ ′ ⊂ Λ \Lambda' \subset\Lambda Λ ′ ⊂ Λ such that the induced morphism l i m → λ ′ ∈ Λ ′ Z λ ′ ⟶ l i m → λ ∈ Λ Z λ \varinjlim_{\lambda' \in\Lambda'} Z_{\lambda'} \longrightarrow\varinjlim_{\lambda\in\Lambda} Z_\lambda lim λ ′ ∈ Λ ′ Z λ ′ ⟶ lim λ ∈ Λ Z λ is an isomorphism in Ind ( K ) \operatorname{Ind}(\mathsf{K}) Ind ( K ) . Specifically, given an index δ ∈ Δ \delta\in\Delta δ ∈ Δ , choose an index λ ( δ ) ∈ Λ \lambda(\delta) \in\Lambda λ ( δ ) ∈ Λ such that the natural morphism Y δ ⟶ X Y_\delta\longrightarrow\mathfrak{X} Y δ ⟶ X factorizes through the natural morphism Z λ ( δ ) ⟶ X Z_{\lambda(\delta)} \longrightarrow\mathfrak{X} Z λ ( δ ) ⟶ X . Then any countable directed subposet Λ ′ ⊂ Λ \Lambda' \subset\Lambda Λ ′ ⊂ Λ such that for every δ ∈ Δ \delta\in\Delta δ ∈ Δ there exists λ ′ ∈ Λ ′ \lambda' \in\Lambda' λ ′ ∈ Λ ′ with λ ( δ ) ≤ λ ′ \lambda(\delta) \leq\lambda' λ ( δ ) ≤ λ ′ in Λ \Lambda Λ has the desired property. Indeed, for any object W ∈ K W \in\mathsf{K} W ∈ K , any morphism W ⟶ X W \longrightarrow\mathfrak{X} W ⟶ X in Ind ( K ) \operatorname{Ind}(\mathsf{K}) Ind ( K ) factorizes through Y δ ⟶ X Y_\delta\longrightarrow\mathfrak{X} Y δ ⟶ X for some δ ∈ Δ \delta\in\Delta δ ∈ Δ , hence also through Z λ ′ ⟶ X Z_{\lambda'} \longrightarrow\mathfrak{X} Z λ ′ ⟶ X for some λ ′ ∈ Λ ′ \lambda' \in\Lambda' λ ′ ∈ Λ ′ . So, setting X ′ = l i m → λ ′ ∈ Λ ′ Z λ ′ ∈ Ind ( K ) \mathfrak{X}' = \varinjlim_{\lambda' \in\Lambda'} Z_{\lambda'} \in\operatorname{Ind}(\mathsf{K}) X ′ = lim λ ′ ∈ Λ ′ Z λ ′ ∈ Ind ( K ) , the map of sets Mor ( W , X ′ ) ⟶ Mor ( W , X ) \operatorname{Mor}(W,\mathfrak{X}') \longrightarrow\operatorname{Mor}(W,\mathfrak{X}) Mor ( W , X ′ ) ⟶ Mor ( W , X ) is surjective. The injectivity of this map follows from the assumption that ( Z λ ) λ ∈ Λ (Z_\lambda)_{\lambda\in\Lambda} ( Z λ ) λ ∈ Λ is a direct system of monomorphisms. □
Examples 2.5. The following counterexamples show that the notion of a strict ind-quasi-compact ind-quasi-separated ind-scheme, viewed as an object of Sets AffSch o p \operatorname{Sets}^{\operatorname{AffSch}^{\mathrm{op}}} Sets AffSch op or Sets CSch o p \operatorname{Sets}^{\operatorname{CSch}^{\mathrm{op}}} Sets CSch op , is not as well-behaved as one might hope for.
(1) Let X = ⋃ γ ∈ Γ U γ X = \bigcup_{\gamma\in\Gamma} U_\gamma X = ⋃ γ ∈ Γ U γ be a quasi-separated scheme representable as a union of its quasi-compact open subschemes U γ U_\gamma U γ indexed by a directed poset Γ \Gamma Γ such that U γ ′ ⊂ U γ ′ ′ U_{\gamma'} \subset U_{\gamma''} U γ ′ ⊂ U γ ′′ whenever γ ′ < γ ′ ′ ∈ Γ \gamma' < \gamma'' \in\Gamma γ ′ < γ ′′ ∈ Γ . Assume further that, for every γ ∈ Γ \gamma\in\Gamma γ ∈ Γ , the open subscheme U γ U_\gamma U γ is dense in X X X . An example of a separated locally Noetherian scheme X X X of this kind (of Krull dimension 2 and with Γ = Z ≥ 1 \Gamma= \mathbb{Z}_{\geq1} Γ = Z ≥ 1 ) can be found in [19 end of Section II.7] .
Denote by X ∈ Sets CSch o p \mathfrak{X} \in\operatorname{Sets}^{\operatorname{CSch}^{\mathrm{op}}} X ∈ Sets CSch op the functor represented by X X X . Then any morphism into X \mathfrak{X} X from a quasi-compact quasi-separated scheme W W W factorizes uniquely as W ⟶ U γ ⟶ X W \longrightarrow U_\gamma\longrightarrow\mathfrak{X} W ⟶ U γ ⟶ X for large enough γ ∈ Γ \gamma\in\Gamma γ ∈ Γ ; hence we have X = l i m → γ ∈ Γ U γ \mathfrak{X} = \varinjlim_{\gamma\in\Gamma} U_\gamma X = lim γ ∈ Γ U γ in Sets CSch o p \operatorname{Sets}^{\operatorname{CSch}^{\mathrm{op}}} Sets CSch op . On the other hand, one obviously has X = l i m → Δ X \mathfrak{X} = \varinjlim_{\Delta} X X = lim Δ X , where Δ = { ∗ } \Delta= \{*\} Δ = { ∗ } denotes the
singleton poset. So we have a representation of the functor X ∈ S e t s C S c h o p \mathcal{X} \in\mathrm{Sets}^{\mathrm{CSch}^{\mathrm{op}}} X ∈ Sets CSch op by a directed diagram of quasi-compact quasi-separated schemes U γ U_{\gamma} U γ (indexed by Γ \Gamma Γ ), and another representation of X \mathcal{X} X by a directed diagram of closed immersions of quasi-separated schemes (indexed by Δ \Delta Δ ).
However, it is intuitively pretty clear that X \mathcal{X} X cannot be represented by a directed diagram of closed immersions of quasi-compact quasi-separated schemes. Indeed, assume for the sake of contradiction that X = l i m → λ ∈ Λ Z λ \mathcal{X} = \varinjlim_{\lambda\in\Lambda} Z_{\lambda} X = lim λ ∈ Λ Z λ is such a representation. Pick an index γ 0 ∈ Γ \gamma_{0} \in\Gamma γ 0 ∈ Γ . Then the natural morphism U γ 0 ⟶ X U_{\gamma_{0}} \longrightarrow\mathcal{X} U γ 0 ⟶ X factorizes through the natural morphism Z λ 0 ⟶ X Z_{\lambda_{0}} \longrightarrow\mathcal{X} Z λ 0 ⟶ X for some λ 0 ∈ Λ \lambda_{0} \in\Lambda λ 0 ∈ Λ . Pick an index λ ′ ≥ λ 0 \lambda' \geq\lambda_{0} λ ′ ≥ λ 0 in Λ \Lambda Λ . Then the natural morphism Z λ ′ ⟶ X Z_{\lambda'} \longrightarrow\mathcal{X} Z λ ′ ⟶ X factorizes through the natural morphism U γ ′ ⟶ X U_{\gamma'} \longrightarrow\mathcal{X} U γ ′ ⟶ X for some γ ′ ∈ Γ \gamma' \in\Gamma γ ′ ∈ Γ , γ ′ ≥ γ 0 \gamma' \geq\gamma_{0} γ ′ ≥ γ 0 . Now we have a commutative square diagram of morphisms of schemes
U γ 0 → U γ ′ ↓ ↑ Z λ 0 → Z λ ′ \begin{CD}
U_{\gamma_{0}} @>>> U_{\gamma'} \\
@VVV @AAA \\
Z_{\lambda_{0}} @>>> Z_{\lambda'}
\end{CD} U γ 0 ↓ ⏐ Z λ 0 U γ ′ ⏐ ↑ Z λ ′
with an open immersion U γ 0 ⟶ U γ ′ U_{\gamma_{0}} \longrightarrow U_{\gamma'} U γ 0 ⟶ U γ ′ and a closed immersion Z λ 0 ⟶ Z λ ′ Z_{\lambda_{0}} \longrightarrow Z_{\lambda'} Z λ 0 ⟶ Z λ ′ . The commutativity of the square diagram follows from the fact that ( U γ ) γ ∈ Γ (U_{\gamma})_{\gamma\in\Gamma} ( U γ ) γ ∈ Γ is a direct system of monomorphisms in CSch.
Arguing similarly to the proof of Corollary 2.3 and using Lemma 2.2 (a), we see that U γ 0 ⟶ Z λ 0 U_{\gamma_{0}} \longrightarrow Z_{\lambda_{0}} U γ 0 ⟶ Z λ 0 and Z λ ′ ⟶ U γ ′ Z_{\lambda'} \longrightarrow U_{\gamma'} Z λ ′ ⟶ U γ ′ are locally closed immersions of schemes. We recall that, by assumption, U γ 0 U_{\gamma_{0}} U γ 0 is dense in U γ ′ U_{\gamma'} U γ ′ . Hence Z λ 0 Z_{\lambda_{0}} Z λ 0 is dense in Z λ ′ Z_{\lambda'} Z λ ′ ; so the closed immersion of schemes Z λ 0 ⟶ Z λ ′ Z_{\lambda_{0}} \longrightarrow Z_{\lambda'} Z λ 0 ⟶ Z λ ′ is bijective (in fact, a homeomorphism of the underlying topological spaces). We have proved that this holds for all λ ′ ≥ λ 0 \lambda' \geq\lambda_{0} λ ′ ≥ λ 0 .
Put μ 0 = λ 0 \mu_{0} = \lambda_{0} μ 0 = λ 0 , and pick an index μ ′ ∈ Λ \mu' \in\Lambda μ ′ ∈ Λ , μ ′ ≥ μ 0 \mu' \geq\mu_{0} μ ′ ≥ μ 0 . Then the natural morphism Z μ 0 ⟶ X Z_{\mu_{0}} \longrightarrow\mathcal{X} Z μ 0 ⟶ X factorizes through the natural morphism U δ 0 ⟶ X U_{\delta_{0}} \longrightarrow\mathcal{X} U δ 0 ⟶ X for some δ 0 ∈ Γ \delta_{0} \in\Gamma δ 0 ∈ Γ , and the natural morphism Z μ ′ ⟶ X Z_{\mu'} \longrightarrow\mathcal{X} Z μ ′ ⟶ X factorizes through the natural morphism U δ ′ ⟶ X U_{\delta'} \longrightarrow\mathcal{X} U δ ′ ⟶ X for some δ ′ ∈ Γ \delta' \in\Gamma δ ′ ∈ Γ , δ ′ ≥ δ 0 \delta' \geq\delta_{0} δ ′ ≥ δ 0 . Now we have another commutative square diagram of morphisms of schemes
U δ 0 → U δ ′ ↑ ↑ Z μ 0 → Z μ ′ \begin{CD}
U_{\delta_{0}} @>>> U_{\delta'} \\
@AAA @AAA \\
Z_{\mu_{0}} @>>> Z_{\mu'}
\end{CD} U δ 0 ⏐ ↑ Z μ 0 U δ ′ ⏐ ↑ Z μ ′
with an open immersion U δ 0 ⟶ U δ ′ U_{\delta_{0}} \longrightarrow U_{\delta'} U δ 0 ⟶ U δ ′ and a bijective closed immersion Z μ 0 ⟶ Z μ ′ Z_{\mu_{0}} \longrightarrow Z_{\mu'} Z μ 0 ⟶ Z μ ′ . Once again, both Z μ 0 ⟶ U δ 0 Z_{\mu_{0}} \longrightarrow U_{\delta_{0}} Z μ 0 ⟶ U δ 0 and Z μ ′ ⟶ U δ ′ Z_{\mu'} \longrightarrow U_{\delta'} Z μ ′ ⟶ U δ ′ are locally closed immersions of schemes (but we do not need to use this fact).
It follows that the morphism of schemes Z μ ′ ⟶ U δ ′ Z_{\mu'} \longrightarrow U_{\delta'} Z μ ′ ⟶ U δ ′ factorizes uniquely through the open immersion U δ 0 ⟶ U δ ′ U_{\delta_{0}} \longrightarrow U_{\delta'} U δ 0 ⟶ U δ ′ . As this holds for all μ ′ ≥ μ 0 \mu' \geq\mu_{0} μ ′ ≥ μ 0 and X = l i m → μ ′ ≥ μ 0 Z μ ′ \mathcal{X} = \varinjlim_{\mu' \geq\mu_{0}} Z_{\mu'} X = lim μ ′ ≥ μ 0 Z μ ′ , we arrive to the conclusion that the morphism U δ 0 ⟶ X U_{\delta_{0}} \longrightarrow\mathcal{X} U δ 0 ⟶ X is an isomorphism; an obvious contradiction.
(2) Let U U U be an affine scheme with an open subscheme V ⊂ U V \subset U V ⊂ U that is not quasi-compact. Then V V V can be represented as a union of quasi-compact open subschemes W γ ⊂ V W_\gamma\subset V W γ ⊂ V indexed by a directed poset Γ \Gamma Γ such that W γ ′ ⊂ W γ ′ ′ W_{\gamma'} \subset W_{\gamma''} W γ ′ ⊂ W γ ′′ whenever γ ′ ≤ γ ′ ′ ∈ Γ \gamma' \le\gamma'' \in\Gamma γ ′ ≤ γ ′′ ∈ Γ . Denote by X X X the scheme obtained by gluing two copies of U U U along V V V , and by Y γ Y_\gamma Y γ the scheme obtained by gluing two copies of U U U along W γ W_\gamma W γ . So the schemes Y γ Y_\gamma Y γ are quasi-separated while the scheme X X X is not quasi-separated. Denote by S e t s C S c h o p \mathrm{Sets}^{\mathrm{CSch}^{\mathrm{op}}} Sets CSch op the functor represented by X X X . The schemes Y γ Y_\gamma Y γ form a directed diagram indexed by Γ \Gamma Γ , with a natural morphism l i m → γ ∈ Γ Y γ ⟶ X \varinjlim_{\gamma\in\Gamma} Y_\gamma\longrightarrow X lim γ ∈ Γ Y γ ⟶ X in S e t s C S c h o p \mathrm{Sets}^{\mathrm{CSch}^{\mathrm{op}}} Sets CSch op .
We claim that the morphism of functors l i m → γ ∈ Γ Y γ ⟶ X \varinjlim_{\gamma\in\Gamma} Y_\gamma\longrightarrow X lim γ ∈ Γ Y γ ⟶ X is an isomorphism in S e t s C S c h o p \mathrm{Sets}^{\mathrm{CSch}^{\mathrm{op}}} Sets CSch op . Indeed, let Z Z Z be a quasi-compact quasi-separated scheme, and let f : Z ⟶ X f: Z \longrightarrow X f : Z ⟶ X be a morphism of schemes. Denote by U ′ U' U ′ , U ′ ′ ⊂ X U'' \subset X U ′′ ⊂ X the two copies of U U U embedded naturally into X X X ; so U ′ ∩ U ′ ′ = V ⊂ X U' \cap U'' = V \subset X U ′ ∩ U ′′ = V ⊂ X . Then the open subschemes f − 1 ( U ′ ) f^{-1}(U') f − 1 ( U ′ ) and f − 1 ( U ′ ′ ) ⊂ Z f^{-1}(U'') \subset Z f − 1 ( U ′′ ) ⊂ Z are usually not quasi-compact. Pick quasi-compact open coverings f − 1 ( U ′ ) = ⋃ α T α ′ f^{-1}(U') = \bigcup_\alpha T'_\alpha f − 1 ( U ′ ) = ⋃ α T α ′ and f − 1 ( U ′ ′ ) = ⋃ β T β ′ ′ f^{-1}(U'') = \bigcup_\beta T''_\beta f − 1 ( U ′′ ) = ⋃ β T β ′′ , and put T α , β = T α ′ ∪ T β ′ ′ ⊂ Z T_{\alpha,\beta} = T'_\alpha\cup T''_\beta\subset Z T α , β = T α ′ ∪ T β ′′ ⊂ Z . Then Z = ⋃ α , β T α , β Z = \bigcup_{\alpha,\beta} T_{\alpha,\beta} Z = ⋃ α , β T α , β is an open covering of Z Z Z . As such, it has a finite subcovering; so we have Z = ⋃ i = 1 m T α i , β i Z = \bigcup_{i=1}^m T_{\alpha_i,\beta_i} Z = ⋃ i = 1 m T α i , β i for some integer m ≥ 0 m \ge0 m ≥ 0 and some indices α i \alpha_i α i , β i \beta_i β i .
Now the intersections T α ′ ∩ T β ′ ′ ⊂ Z T'_\alpha\cap T''_\beta\subset Z T α ′ ∩ T β ′′ ⊂ Z are quasi-compact (since the scheme Z Z Z is quasi-separated). Hence there exists an index γ ∈ Γ \gamma\in\Gamma γ ∈ Γ such that for all 1 ≤ i ≤ m 1 \le i \le m 1 ≤ i ≤ m one has f ( T α i ′ ∩ T β i ′ ′ ) ⊂ W γ ⊂ V = U ′ ∩ U ′ ′ f(T'_{\alpha_i} \cap T''_{\beta_i}) \subset W_\gamma\subset V = U' \cap U'' f ( T α i ′ ∩ T β i ′′ ) ⊂ W γ ⊂ V = U ′ ∩ U ′′ .
Let T T T be a scheme represented as the union of two open subschemes, T = T ′ ∪ T ′ ′ T = T' \cup T'' T = T ′ ∪ T ′′ . To specify a morphism of schemes g : T ⟶ X g: T \longrightarrow X g : T ⟶ X such that g ( T ′ ) ⊂ U ′ g(T') \subset U' g ( T ′ ) ⊂ U ′ and g ( T ′ ′ ) ⊂ U ′ ′ g(T'') \subset U'' g ( T ′′ ) ⊂ U ′′ , it suffices to pinpoint two morphisms g ′ : T ′ ⟶ U ′ g': T' \longrightarrow U' g ′ : T ′ ⟶ U ′ and g ′ ′ : T ′ ′ ⟶ U ′ ′ g'': T'' \longrightarrow U'' g ′′ : T ′′ ⟶ U ′′ that both take the intersection T ′ ∩ T ′ ′ T' \cap T'' T ′ ∩ T ′′ into V = U ′ ∩ U ′ ′ ⊂ X V = U' \cap U'' \subset X V = U ′ ∩ U ′′ ⊂ X and agree on that intersection. Morphisms of schemes h : T ′ ∪ T ′ ′ ⟶ Y γ h: T' \cup T'' \longrightarrow Y_\gamma h : T ′ ∪ T ′′ ⟶ Y γ are described similarly, except that the intersection is now U γ ′ ∩ U γ ′ ′ = W γ U'_\gamma\cap U''_\gamma= W_\gamma U γ ′ ∩ U γ ′′ = W γ , with the obvious notation U γ ′ U'_\gamma U γ ′ and U γ ′ ′ U''_\gamma U γ ′′ for the two copies of U U U embedded naturally into Y γ Y_\gamma Y γ .
It follows easily that the morphism Z ⟶ X Z \longrightarrow X Z ⟶ X factorizes uniquely as Z ⟶ Y γ ⟶ X Z \longrightarrow Y_\gamma\longrightarrow X Z ⟶ Y γ ⟶ X , where the index γ \gamma γ satisfies the condition above, and the morphism h : Z ⟶ Y γ h: Z \longrightarrow Y_\gamma h : Z ⟶ Y γ satisfies the conditions that h ( T α i ′ ) ⊂ U γ ′ h(T'_{\alpha_i}) \subset U'_\gamma h ( T α i ′ ) ⊂ U γ ′ and h ( T β i ′ ′ ) ⊂ U γ ′ ′ h(T''_{\beta_i}) \subset U''_\gamma h ( T β i ′′ ) ⊂ U γ ′′ for all $1 \le i \le m.
This shows that the morphism of set-valued functors l i m → γ ∈ Γ Y γ ⟶ X \varinjlim_{\gamma\in\Gamma} Y_\gamma\longrightarrow X lim γ ∈ Γ Y γ ⟶ X is surjective.
To prove the injectivity, suppose that we are given an index δ ∈ Γ \delta\in\Gamma δ ∈ Γ and two morphisms of schemes h 1 , h 2 : Z ⟶ Y δ h_1, h_2: Z \longrightarrow Y_\delta h 1 , h 2 : Z ⟶ Y δ such that the two compositions Z ⇉ Y δ ⟶ X Z \rightrightarrows Y_\delta\longrightarrow X Z ⇉ Y δ ⟶ X agree. Denote these two equal compositions by f : Z ⟶ X f: Z \longrightarrow X f : Z ⟶ X . For every index ϵ ≥ δ \epsilon\ge\delta ϵ ≥ δ , denote the compositions Z ⇉ Y δ ⟶ Y ϵ Z \rightrightarrows Y_\delta\longrightarrow Y_\epsilon Z ⇉ Y δ ⟶ Y ϵ by h 1 , ϵ , h 2 , ϵ : Z ⟶ Y ϵ h_{1,\epsilon}, h_{2,\epsilon}: Z \longrightarrow Y_\epsilon h 1 , ϵ , h 2 , ϵ : Z ⟶ Y ϵ . Then we have f − 1 ( U ′ ) = ⋃ ϵ ≥ δ h 1 , ϵ − 1 ( U ϵ ′ ) = ⋃ ϵ ≥ δ h 2 , ϵ − 1 ( U ϵ ′ ) f^{-1}(U') = \bigcup_{\epsilon\ge\delta} h_{1,\epsilon}^{-1}(U'_\epsilon) = \bigcup_{\epsilon\ge\delta} h_{2,\epsilon}^{-1}(U'_\epsilon) f − 1 ( U ′ ) = ⋃ ϵ ≥ δ h 1 , ϵ − 1 ( U ϵ ′ ) = ⋃ ϵ ≥ δ h 2 , ϵ − 1 ( U ϵ ′ ) and f − 1 ( U ′ ′ ) = ⋃ ϵ ≥ δ h 1 , ϵ − 1 ( U ϵ ′ ′ ) = ⋃ ϵ ≥ δ h 2 , ϵ − 1 ( U ϵ ′ ′ ) f^{-1}(U'') = \bigcup_{\epsilon\ge\delta} h_{1,\epsilon}^{-1}(U''_\epsilon) = \bigcup_{\epsilon\ge\delta} h_{2,\epsilon}^{-1}(U''_\epsilon) f − 1 ( U ′′ ) = ⋃ ϵ ≥ δ h 1 , ϵ − 1 ( U ϵ ′′ ) = ⋃ ϵ ≥ δ h 2 , ϵ − 1 ( U ϵ ′′ ) .
In the notation above, we have f − 1 ( U ′ ) = ⋃ α T α ′ f^{-1}(U') = \bigcup_\alpha T'_\alpha f − 1 ( U ′ ) = ⋃ α T α ′ and f − 1 ( U ′ ′ ) = ⋃ β T β ′ ′ f^{-1}(U'') = \bigcup_\beta T''_\beta f − 1 ( U ′′ ) = ⋃ β T β ′′ , and then Z = ⋃ i = 1 m T α i , β i Z = \bigcup_{i=1}^m T_{\alpha_i,\beta_i} Z = ⋃ i = 1 m T α i , β i for T α , β = T α ′ ∪ T β ′ ′ T_{\alpha,\beta} = T'_\alpha\cup T''_\beta T α , β = T α ′ ∪ T β ′′ and some indices α i \alpha_i α i , β i \beta_i β i , and finally f ( T α i ′ ∩ T β i ′ ′ ) ⊂ W γ ⊂ V = U ′ ∩ U ′ ′ f(T'_{\alpha_i} \cap T''_{\beta_i}) \subset W_\gamma\subset V = U' \cap U'' f ( T α i ′ ∩ T β i ′′ ) ⊂ W γ ⊂ V = U ′ ∩ U ′′ for all 1 ≤ i ≤ m 1 \le i \le m 1 ≤ i ≤ m and some γ ∈ Γ \gamma\in\Gamma γ ∈ Γ . Let ϵ ∈ Γ \epsilon\in\Gamma ϵ ∈ Γ be an index such that ϵ ≥ δ \epsilon\ge\delta ϵ ≥ δ and ϵ ≥ γ \epsilon\ge\gamma ϵ ≥ γ , and also T α i ′ ⊂ h 1 , ϵ − 1 ( U ϵ ′ ) T'_{\alpha_i} \subset h_{1,\epsilon}^{-1}(U'_\epsilon) T α i ′ ⊂ h 1 , ϵ − 1 ( U ϵ ′ ) , T α i ′ ⊂ h 2 , ϵ − 1 ( U ϵ ′ ) T'_{\alpha_i} \subset h_{2,\epsilon}^{-1}(U'_\epsilon) T α i ′ ⊂ h 2 , ϵ − 1 ( U ϵ ′ ) and T β i ′ ′ ⊂ h 1 , ϵ − 1 ( U ϵ ′ ′ ) T''_{\beta_i} \subset h_{1,\epsilon}^{-1}(U''_\epsilon) T β i ′′ ⊂ h 1 , ϵ − 1 ( U ϵ ′′ ) , T β i ′ ′ ⊂ h 2 , ϵ − 1 ( U ϵ ′ ′ ) T''_{\beta_i} \subset h_{2,\epsilon}^{-1}(U''_\epsilon) T β i ′′ ⊂ h 2 , ϵ − 1 ( U ϵ ′′ ) for all $1 \le i \le m.
We have a natural morphism Y ϵ ⟶ U Y_\epsilon\longrightarrow U Y ϵ ⟶ U , which factorizes naturally as Y ϵ ⟶ X ⟶ U Y_\epsilon\longrightarrow X \longrightarrow U Y ϵ ⟶ X ⟶ U . Now let T T T be a scheme and g 1 g_1 g 1 , g 2 : T ⇉ Y ϵ g_2: T \rightrightarrows Y_\epsilon g 2 : T ⇉ Y ϵ be two morphisms of schemes. To check that g 1 = g 2 g_1 = g_2 g 1 = g 2 , it suffices to cover T T T by two open subschemes, T = T ′ ∪ T ′ ′ T = T' \cup T'' T = T ′ ∪ T ′′ , such that g j ( T ′ ) ⊂ U ϵ ′ g_j(T') \subset U'_{\epsilon} g j ( T ′ ) ⊂ U ϵ ′ and g j ( T ′ ′ ) ⊂ U ϵ ′ ′ g_j(T'') \subset U''_{\epsilon} g j ( T ′′ ) ⊂ U ϵ ′′ for j = 1 , 2 j=1,2 j = 1 , 2 , and the two compositions T ⇒ Y ϵ → U T \Rightarrow Y_{\epsilon} \to U T ⇒ Y ϵ → U are equal to each other.
It follows easily, by applying this criterion, that h 1 , ϵ = h 2 , ϵ h_{1,\epsilon}=h_{2,\epsilon} h 1 , ϵ = h 2 , ϵ for the index ϵ ∈ Γ \epsilon\in\Gamma ϵ ∈ Γ chosen as above. This proves that the morphism of set-valued functors “lim γ ∈ Γ Y γ → X \lim\limits_{\gamma\in\Gamma} Y_{\gamma} \to\mathfrak{X} γ ∈ Γ lim Y γ → X ” is also injective, as desired.
On the other hand, one obviously has X = \mathfrak{X}= X = “lim Δ X \lim\limits_{\Delta} X Δ lim X ”, where Δ = { ∗ } \Delta=\{\ast\} Δ = { ∗ } is the singleton poset. Thus we have a representation of the functor X ∈ S e t s C S c h o p \mathfrak{X}\in\mathrm{Sets}^{\mathrm{CSch}^{\mathrm{op}}} X ∈ Sets CSch op by a directed diagram of quasi-compact quasi-separated schemes Y γ Y_{\gamma} Y γ (indexed by Γ \Gamma Γ ), and another representation of X \mathfrak{X} X by a directed diagram of closed immersions of quasi-compact schemes (indexed by Δ \Delta Δ ). We leave it as an exercise for the reader to check that X \mathfrak{X} X cannot be represented by a directed diagram of closed immersions of quasi-compact quasi-separated schemes.
Based on the discussion above in this section, for the rest of this paper we will mean by “ind-scheme” an ind-object in the category of concentrated schemes C S c h \mathrm{CSch} CSch . A “strict ind-scheme” is an ind-scheme representable by a directed diagram of closed immersions of concentrated schemes. An “ℵ 0 \aleph_0 ℵ 0 -ind-scheme” is an ind-scheme representable by a countable directed diagram of concentrated schemes. By Lemma 2.4 (b), any strict ℵ 0 \aleph_0 ℵ 0 -ind-scheme is representable by a countable directed diagram of closed immersions of concentrated schemes.
We refer to [33 Section 1.2] for a discussion of the issue of uniqueness of a directed diagram of closed immersions of schemes representing a given strict ind-scheme. To give a precise formulation, the arguments in [33 Section 1.2] prove that any isomorphism in I n d ( C S c h ) \mathrm{Ind}(\mathrm{CSch}) Ind ( CSch ) or I n d ( S c h ) \mathrm{Ind}(\mathrm{Sch}) Ind ( Sch ) between two ind-schemes represented by directed diagrams of closed immersions of schemes is also an isomorphism of ind-objects in the category of (concentrated or arbitrary) schemes with closed immersions as morphisms. No other morphisms of schemes than closed immersions can appear as components of such an isomorphism, the “components” being understood in the sense of the formula (1) .
An ind-scheme is said to be ind-semi-separated if it is representable by a directed diagram of semi-separated schemes, ind-affine if it is representable by a directed diagram of affine schemes, and ind-Noetherian if it is representable by a directed diagram of Noetherian schemes. By Corollary 2.3 (b,c,e) and the subsequent paragraph, the notions of a strict ind-semi-separated ind-scheme, strict ind-affine ind-scheme, and strict ind-Noetherian ind-scheme are well-behaved, in the sense of the discussion above in this section. So are the notions of a strict ind-semi-separated, strict ind-affine, or strict ind-Noetherian ℵ 0 \aleph_0 ℵ 0 -ind-scheme, by Lemma 2.4 (b).
Finite diagrams and fibered products of ind-schemes
Let D D D be a finite poset. Given a category K K K , we denote by K D K^D K D the category of D D D -indexed diagrams in K K K . We are particularly interested in two special cases: for the ordered set of two elements D = ( ∙ → ∙ ) D=(\bullet\to\bullet) D = ( ∙ → ∙ ) , the category K D K^D K D is the category of morphisms in K K K ; while for the poset D = ( ∙ → ∙ ← ∙ ) D=(\bullet\to\bullet\leftarrow\bullet) D = ( ∙ → ∙ ← ∙ ) , the category K D K^D K D is the category of cospans in K K K .
The limits of cospans are called the pullbacks or the fibered products . We denote the fibered product of a cospan X ⟶ S ⟵ Y X \longrightarrow S \longleftarrow Y X ⟶ S ⟵ Y by X × S Y X \times_{S} Y X × S Y .
Proposition 2.6. Let K K K be a category and D D D be a finite poset. Then the natural functor Ind ( K D ) ⟶ Ind ( K ) D \operatorname{Ind}(K^{D}) \longrightarrow\operatorname{Ind}(K)^{D} Ind ( K D ) ⟶ Ind ( K ) D is a category equivalence.
Proof. This is [3 Exposé I, Proposition 8.8.5] , [25 page 55] , or a special case of [20 Theorem 1.3] . □
In particular, by Proposition 2.6, any morphism f : Y ⟶ X f\colon\mathcal{Y} \longrightarrow\mathcal{X} f : Y ⟶ X in Ind K \operatorname{Ind} K Ind K can be represented by a directed diagram of morphisms ( f ξ : Y ξ ⟶ X ξ ) ξ ∈ Ξ (f_{\xi}\colon Y_{\xi} \longrightarrow X_{\xi})_{\xi\in\Xi} ( f ξ : Y ξ ⟶ X ξ ) ξ ∈ Ξ in K K K , indexed by some directed poset Ξ \Xi Ξ . So we have f = l i m → ξ ∈ Ξ f ξ f = \varinjlim_{\xi\in\Xi} f_{\xi} f = lim ξ ∈ Ξ f ξ . Given two directed diagrams ( X γ ) γ ∈ Γ (X_{\gamma})_{\gamma\in\Gamma} ( X γ ) γ ∈ Γ and ( Y δ ) δ ∈ Δ (Y_{\delta})_{\delta\in\Delta} ( Y δ ) δ ∈ Δ in K K K representing the ind-objects X \mathcal{X} X and Y ∈ Ind K \mathcal{Y} \in\operatorname{Ind} K Y ∈ Ind K , an explicit construction of a directed poset Ξ \Xi Ξ and a Ξ \Xi Ξ -indexed diagram of morphisms ( f ξ : Y ξ ⟶ X ξ ) (f_{\xi}\colon Y_{\xi} \longrightarrow X_{\xi}) ( f ξ : Y ξ ⟶ X ξ ) in K K K representing the morphism f f f can be found, e. g., in [32 Section 9] .
The nature of the explicit construction mentioned in the previous paragraph is such that all the objects in the diagrams ( X ξ ) ξ ∈ Ξ (X_{\xi})_{\xi\in\Xi} ( X ξ ) ξ ∈ Ξ and ( Y ξ ) ξ ∈ Ξ (Y_{\xi})_{\xi\in\Xi} ( Y ξ ) ξ ∈ Ξ appear already in the diagrams ( X γ ) γ ∈ Γ (X_{\gamma})_{\gamma\in\Gamma} ( X γ ) γ ∈ Γ and ( Y δ ) δ ∈ Δ (Y_{\delta})_{\delta\in\Delta} ( Y δ ) δ ∈ Δ , and all the transition morphisms in the diagrams ( X ξ ) ξ ∈ Ξ (X_{\xi})_{\xi\in\Xi} ( X ξ ) ξ ∈ Ξ and ( Y ξ ) ξ ∈ Ξ (Y_{\xi})_{\xi\in\Xi} ( Y ξ ) ξ ∈ Ξ appear already in the diagrams ( X γ ) γ ∈ Γ (X_{\gamma})_{\gamma\in\Gamma} ( X γ ) γ ∈ Γ and ( Y δ ) δ ∈ Δ (Y_{\delta})_{\delta\in\Delta} ( Y δ ) δ ∈ Δ , respectively. If ( X γ ) γ ∈ Γ (X_{\gamma})_{\gamma\in\Gamma} ( X γ ) γ ∈ Γ and ( Y δ ) δ ∈ Δ (Y_{\delta})_{\delta\in\Delta} ( Y δ ) δ ∈ Δ are countable diagrams of monomorphisms in K K K , then so are the diagrams ( X ξ ) ξ ∈ Ξ (X_{\xi})_{\xi\in\Xi} ( X ξ ) ξ ∈ Ξ and ( Y ξ ) ξ ∈ Ξ (Y_{\xi})_{\xi\in\Xi} ( Y ξ ) ξ ∈ Ξ .
Furthermore, let X ⟶ S ⟵ Y \mathcal{X} \longrightarrow\mathcal{S} \longleftarrow\mathcal{Y} X ⟶ S ⟵ Y be a cospan in Ind K \operatorname{Ind} K Ind K . By Proposition 2.6, the cospan X ⟶ S ⟵ Y \mathcal{X} \longrightarrow\mathcal{S} \longleftarrow\mathcal{Y} X ⟶ S ⟵ Y can be represented by a directed diagram of cospans ( X ξ ⟶ S ξ ⟵ Y ξ ) (X_{\xi} \longrightarrow S_{\xi} \longleftarrow Y_{\xi}) ( X ξ ⟶ S ξ ⟵ Y ξ ) in K K K , indexed by some directed poset Ξ \Xi Ξ . Given three directed diagrams ( X γ ) γ ∈ Γ (X_{\gamma})_{\gamma\in\Gamma} ( X γ ) γ ∈ Γ , ( Y δ ) δ ∈ Δ (Y_{\delta})_{\delta\in\Delta} ( Y δ ) δ ∈ Δ , and ( S ϵ ) ϵ ∈ E (S_{\epsilon})_{\epsilon\in E} ( S ϵ ) ϵ ∈ E in K K K representing the ind-objects X \mathcal{X} X , Y \mathcal{Y} Y , and S \mathcal{S} S , an explicit construction of a directed poset Ξ \Xi Ξ and a Ξ \Xi Ξ -indexed diagram of cospans ( X ξ ⟶ S ξ ⟵ Y ξ ) (X_{\xi} \longrightarrow S_{\xi} \longleftarrow Y_{\xi}) ( X ξ ⟶ S ξ ⟵ Y ξ ) in K K K representing the cospan X ⟶ S ⟵ Y \mathcal{X} \longrightarrow\mathcal{S} \longleftarrow\mathcal{Y} X ⟶ S ⟵ Y in Ind K \operatorname{Ind} K Ind K can be found, e. g., in [33 Section 1.1] .
Once again, all the objects in the diagrams ( X ξ ) ξ ∈ Ξ (X_{\xi})_{\xi\in\Xi} ( X ξ ) ξ ∈ Ξ , ( Y ξ ) ξ ∈ Ξ (Y_{\xi})_{\xi\in\Xi} ( Y ξ ) ξ ∈ Ξ , and ( S ξ ) ξ ∈ Ξ (S_{\xi})_{\xi\in\Xi} ( S ξ ) ξ ∈ Ξ appear already in the diagrams ( X γ ) γ ∈ Γ (X_{\gamma})_{\gamma\in\Gamma} ( X γ ) γ ∈ Γ , ( Y δ ) δ ∈ Δ (Y_{\delta})_{\delta\in\Delta} ( Y δ ) δ ∈ Δ , and ( S ϵ ) ϵ ∈ E (S_{\epsilon})_{\epsilon\in E} ( S ϵ ) ϵ ∈ E , and all the transition morphisms in the diagrams ( X ξ ) ξ ∈ Ξ (X_{\xi})_{\xi\in\Xi} ( X ξ ) ξ ∈ Ξ , ( Y ξ ) ξ ∈ Ξ (Y_{\xi})_{\xi\in\Xi} ( Y ξ ) ξ ∈ Ξ , and ( S ξ ) ξ ∈ Ξ (S_{\xi})_{\xi\in\Xi} ( S ξ ) ξ ∈ Ξ appear already in the diagrams ( X γ ) γ ∈ Γ (X_{\gamma})_{\gamma\in\Gamma} ( X γ ) γ ∈ Γ , ( Y δ ) δ ∈ Δ (Y_{\delta})_{\delta\in\Delta} ( Y δ ) δ ∈ Δ , and ( S ϵ ) ϵ ∈ E (S_{\epsilon})_{\epsilon\in E} ( S ϵ ) ϵ ∈ E , respectively. If ( X γ ) γ ∈ Γ (X_{\gamma})_{\gamma\in\Gamma} ( X γ ) γ ∈ Γ , ( Y δ ) δ ∈ Δ (Y_{\delta})_{\delta\in\Delta} ( Y δ ) δ ∈ Δ , and ( S ϵ ) ϵ ∈ E (S_{\epsilon})_{\epsilon\in E} ( S ϵ ) ϵ ∈ E are countable diagrams of monomorphisms in K K K , then so are the diagrams ( X ξ ) ξ ∈ Ξ (X_{\xi})_{\xi\in\Xi} ( X ξ ) ξ ∈ Ξ , ( Y ξ ) ξ ∈ Ξ (Y_{\xi})_{\xi\in\Xi} ( Y ξ ) ξ ∈ Ξ , and ( S ξ ) ξ ∈ Ξ (S_{\xi})_{\xi\in\Xi} ( S ξ ) ξ ∈ Ξ .
Now let K K K be a category with fibered coproducts. Let X ⟶ S ⟵ Y \mathcal{X} \longrightarrow\mathcal{S} \longleftarrow\mathcal{Y} X ⟶ S ⟵ Y be a cospan in Ind K \operatorname{Ind} K Ind K , represented by a directed diagram of cospans ( X ξ ⟶ S ξ ⟵ Y ξ ) (X_{\xi} \longrightarrow S_{\xi} \longleftarrow Y_{\xi}) ( X ξ ⟶ S ξ ⟵ Y ξ ) in K K K , indexed by a directed poset Ξ \Xi Ξ . Then one can easily see that the directed diagram ( X ξ × S ξ Y ξ ) ξ ∈ Ξ (X_{\xi} \times_{S_{\xi}} Y_{\xi})_{\xi\in\Xi} ( X ξ × S ξ Y ξ ) ξ ∈ Ξ in K K K represents the fibered coproduct X × S Y \mathcal{X} \times_{\mathcal{S}} \mathcal{Y} X × S Y in Ind K \operatorname{Ind} K Ind K . It suffices to compute morphisms in Ind K \operatorname{Ind} K Ind K by the formula (1) and use the fact that, in the category of sets, the finite limits commute with both the directed limits and the directed colimits. So Ind K \operatorname{Ind} K Ind K is a category with fibered coproducts as well.
Recall that the full subcategory of concentrated schemes CSch \operatorname{CSch} CSch is closed under fibered products in Sch \operatorname{Sch} Sch [21 Remark Tag 0816] . Hence the category of concentrated schemes K = C S c h \mathrm{K}=\mathrm{CSch} K = CSch has fibered products. It follows that the category of (ind-quasi-compact ind-quasi-separated) ind-schemes Ind ( C S c h ) \operatorname{Ind}(\mathrm{CSch}) Ind ( CSch ) is a category with fibered products, which can be computed termwise on the cospans of directed diagrams in C S c h \mathrm{CSch} CSch . Furthermore, the full subcategories of strict ind-schemes and strict ℵ 0 \aleph_{0} ℵ 0 -ind-schemes are closed under fibered products in Ind ( C S c h ) \operatorname{Ind}(\mathrm{CSch}) Ind ( CSch ) .
Ind-affine ind-schemes
Given a scheme X X X , we denote, as usual, by O X \mathcal{O}_{X} O X the structure sheaf of X X X . In particular, for an affine scheme U U U , the notation R = O ( U ) = O U ( U ) R=\mathcal{O}(U)=\mathcal{O}_{U}(U) R = O ( U ) = O U ( U ) stands for the commutative ring such that U = Spec R U=\operatorname{Spec}R U = Spec R .
Following Corollary 2.3(e) and the subsequent paragraph, any strict ind-affine ind-scheme can be represented by a directed diagram of closed immersions of affine schemes. Furthermore, by Lemma 2.4(b), any strict ind-affine ℵ 0 \aleph_{0} ℵ 0 -ind-scheme can be represented by a countable directed diagram of closed immersions of affine schemes. This was already mentioned at the end of Section 2.1.
What are the strict ind-affine ind-schemes? Let U = ‘ ‘ l i m ′ ′ γ ∈ Γ U γ \mathcal{U}=\underset{\gamma\in\Gamma}{\operatorname{``lim''}}\ U_{\gamma} U = γ ∈ Γ ‘‘li m ′′ U γ be an ind-affine ind-scheme represented by a directed diagram of closed immersions of affine schemes. So, for every pair of indices γ ′ ≤ γ ′ ′ ∈ Γ \gamma'\leq\gamma''\in\Gamma γ ′ ≤ γ ′′ ∈ Γ , the transition map O ( U γ ′ ′ ) ⟶ O ( U γ ′ ) \mathcal{O}(U_{\gamma''})\longrightarrow\mathcal{O}(U_{\gamma'}) O ( U γ ′′ ) ⟶ O ( U γ ′ ) is a surjective homomorphism of commutative rings.
Let us say that the ind-affine ind-scheme U \mathcal{U} U is limit-epimorphic if the projection map lim ⟵ γ ∈ Γ O ( U γ ) ⟶ O ( U δ ) \underset{\gamma\in\Gamma}{\lim\limits_{\longleftarrow}}\mathcal{O}(U_{\gamma})\longrightarrow\mathcal{O}(U_{\delta}) γ ∈ Γ ⟵ lim O ( U γ ) ⟶ O ( U δ ) is surjective for every index δ ∈ Γ \delta\in\Gamma δ ∈ Γ . One can easily see that this property does not depend on the choice of a directed diagram of closed immersions of affine schemes ( U γ ) γ ∈ Γ (U_{\gamma})_{\gamma\in\Gamma} ( U γ ) γ ∈ Γ representing the given strict ind-affine ind-scheme U \mathcal{U} U . By the standard property of countable projective limits, any strict ind-affine ℵ 0 \aleph_{0} ℵ 0 -ind-scheme is limit-epimorphic.
Lemma 2.7. The full subcategory of limit-epimorphic strict ind-affine ind-schemes in Ind ( C S c h ) \operatorname{Ind}(\mathrm{CSch}) Ind ( CSch ) is naturally anti-equivalent to the category of complete, separated topological commutative rings where open ideals form a base of neighborhoods of zero (with continuous ring homomorphisms as the morphisms). In particular, the full subcategory of strict ind-affine ℵ 0 \aleph_{0} ℵ 0 -ind-schemes in Ind ( C S c h ) \operatorname{Ind}(\mathrm{CSch}) Ind ( CSch ) is naturally anti-equivalent to the category of complete, separated topological rings having a countable base of neighborhoods of zero consisting of open ideals.
Proof. This is [5 Section 7.11.2(i)] or [33 Examples 1.6 and the subsequent paragraph] . Another relevant reference is [32 Theorem 9.4(a)] .
To spell out some details, the desired equivalence of categories assigns to a limit-epimorphic strict ind-affine ind-scheme U = ‘ ‘ l i m ′ ′ γ ∈ Γ U γ \mathcal{U}=\underset{\gamma\in\Gamma}{\operatorname{``lim''}}\ U_{\gamma} U = γ ∈ Γ ‘‘li m ′′ U γ as above the topological commutative ring R = lim ⟵ γ ∈ Γ O ( U γ ) \mathcal{R}=\underset{\gamma\in\Gamma}{\lim\limits_{\longleftarrow}}\mathcal{O}(U_{\gamma}) R = γ ∈ Γ ⟵ lim O ( U γ ) with the topology of projective limit of discrete rings O ( U γ ) \mathcal{O}(U_{\gamma}) O ( U γ ) . The inverse functor assigns to a complete, separated topological commutative ring R \mathcal{R} R with a topology base of open ideals the ind-affine ind-scheme U = ‘ ‘ l i m ′ ′ J ⊂ R Spec R / J \mathcal{U}=\underset{\mathcal{J}\subset\mathcal{R}}{\operatorname{``lim''}}\operatorname{Spec}\mathcal{R}/\mathcal{J} U = J ⊂ R ‘‘li m ′′ Spec R / J . Here J \mathcal{J} J ranges the poset of all open ideals J ⊂ R \mathcal{J}\subset\mathcal{R} J ⊂ R (ordered by reverse inclusion), and Spec R / J \operatorname{Spec}\mathcal{R}/\mathcal{J} Spec R / J is the spectrum of the discrete commutative ring R / J \mathcal{R}/\mathcal{J} R / J .
We denote the equivalence of categories from Lemma 2.7 by R = O ( U ) \mathcal{R} = \mathcal{O}(\mathcal{U}) R = O ( U ) and U = Spf R \mathcal{U} = \operatorname{Spf} \mathcal{R} U = Spf R (where U \mathcal{U} U and R \mathcal{R} R are the ind-affine ind-scheme and the topological ring corresponding to each other under this category equivalence).
Classes of morphisms of ind-schemes
Let P \mathbf{P} P be a property of morphisms in CSch that is preserved by the pullbacks in CSch. We will say that a morphism f : Y ⟶ X f\colon\mathcal{Y} \longrightarrow\mathcal{X} f : Y ⟶ X in Ind ( CSch ) \operatorname{Ind}(\operatorname{CSch}) Ind ( CSch ) is lax-ind-P \mathbf{P} P if f f f can be represented by a directed diagram of morphisms ( f γ : Y γ → X γ ) γ ∈ Γ (f_\gamma\colon Y_\gamma\to X_\gamma)_{\gamma\in\Gamma} ( f γ : Y γ → X γ ) γ ∈ Γ in CSch, i. e., f = ‘ ‘ lim ⟶ f γ " f = ``\lim\limits_{\longrightarrow} f_\gamma" f = ‘‘ ⟶ lim f γ " , in such a way that all the morphisms f γ f_\gamma f γ have property P \mathbf{P} P [53 Definition 1.19] .
Lemma 2.8. Let P \mathbf{P} P be a property of morphisms in CSch that is stable under pullbacks, and let X ∈ Ind ( CSch ) \mathcal{X} \in\operatorname{Ind}(\operatorname{CSch}) X ∈ Ind ( CSch ) be a strict ind-scheme. Then a morphism f : Y ⟶ X f\colon\mathcal{Y} \longrightarrow\mathcal{X} f : Y ⟶ X in Ind ( CSch ) \operatorname{Ind}(\operatorname{CSch}) Ind ( CSch ) is lax-ind-P \mathbf{P} P if and only if f f f can be represented by a directed diagram of morphisms ( f ξ : Y ξ → X ξ ) ξ ∈ Ξ (f_\xi\colon Y_\xi\to X_\xi)_{\xi\in\Xi} ( f ξ : Y ξ → X ξ ) ξ ∈ Ξ in CSch, i. e., f = ‘ ‘ lim ⟶ f ξ " f = ``\lim\limits_{\longrightarrow} f_\xi" f = ‘‘ ⟶ lim f ξ " , in such a way that all the morphisms f ξ f_\xi f ξ have property P \mathbf{P} P and all the transition morphisms in the directed diagram ( X ξ ) ξ ∈ Ξ (X_\xi)_{\xi\in\Xi} ( X ξ ) ξ ∈ Ξ are closed immersions.
Proof. Let f = ‘ ‘ lim ⟶ g γ " f = ``\lim\limits_{\longrightarrow} g_\gamma" f = ‘‘ ⟶ lim g γ " be a representation of f f f by a directed diagram of morphisms ( g γ : Y γ → X γ ) γ ∈ Γ (g_\gamma\colon Y_\gamma\to X_\gamma)_{\gamma\in\Gamma} ( g γ : Y γ → X γ ) γ ∈ Γ in CSch such that all the morphisms g γ g_\gamma g γ have property P \mathbf{P} P , and let X ≃ ‘ ‘ lim ⟶ X δ ′ " \mathcal{X} \simeq``\lim\limits_{\longrightarrow} X'_\delta" X ≃ ‘‘ ⟶ lim X δ ′ " be a representation of X \mathcal{X} X by a directed diagram of closed immersions in CSch. Put X ′ = ‘ ‘ lim ⟶ X δ ′ " \mathcal{X}' = ``\lim\limits_{\longrightarrow} X'_\delta" X ′ = ‘‘ ⟶ lim X δ ′ " ; so we have an (iso)morphism of ind-schemes X ′ ⟶ X \mathcal{X}' \longrightarrow\mathcal{X} X ′ ⟶ X . Following the discussion in § 2.2 and the explicit constructions of representing direct systems of diagrams in [32 Section 9] and [33 Section 1.1] , one can produce a directed poset Ξ \Xi Ξ and a Ξ \Xi Ξ -indexed direct system of cospans ( Y ξ → X ξ ← X ξ ′ ) (Y_\xi\to X_\xi\leftarrow X'_\xi) ( Y ξ → X ξ ← X ξ ′ ) in CSch such that the morphisms g ξ : Y ξ ⟶ X ξ g_\xi\colon Y_\xi\longrightarrow X_\xi g ξ : Y ξ ⟶ X ξ have property P \mathbf{P} P , the transition morphisms in the direct system ( X ξ ′ ) ξ ∈ Ξ (X'_\xi)_{\xi\in\Xi} ( X ξ ′ ) ξ ∈ Ξ are closed immersions, f = ``\lim\limits_{\longrightarrow}_{\xi\in\Xi} g_\xi" , and the induced morphism of ind-schemes \mathcal{X}' = ``\lim\limits_{\longrightarrow}_{\xi\in\Xi} X'_\xi" \longrightarrow``\lim\limits_{\longrightarrow}_{\xi\in\Xi} X_\xi" = \mathcal{X} is our given (iso)morphism. Now we have \mathcal{Y} \simeq\mathcal{X}' \times_{\mathcal{X}} \mathcal{Y} = ``\lim\limits_{\longrightarrow}_{\xi\in\Xi}(X'_\xi\times_{X_\xi} Y_\xi)" by the same discussion in § 2.2 . It remains to point out that the induced morphism f ξ : X ξ ′ × X ξ Y ξ ⟶ X ξ ′ f_\xi\colon X'_\xi\times_{X_\xi} Y_\xi\longrightarrow X'_\xi f ξ : X ξ ′ × X ξ Y ξ ⟶ X ξ ′ has property P \mathbf{P} P as a pullback of the morphism g ξ g_\xi g ξ . □ \square □
More generally, it is clear from the discussion of the fibered products in § 2.2 that the class of all lax-ind-P \mathbf{P} P morphism of ind-schemes is closed under pullbacks. In other words, if Y ⟶ X \mathcal{Y} \longrightarrow\mathcal{X} Y ⟶ X is a lax-ind-P \mathbf{P} P morphism and X ′ ⟶ X \mathcal{X}' \longrightarrow\mathcal{X} X ′ ⟶ X is a morphism in Ind ( CSch ) \operatorname{Ind}(\operatorname{CSch}) Ind ( CSch ) , then the pullback X ′ × X Y ⟶ X ′ \mathcal{X}' \times_{\mathcal{X}} \mathcal{Y} \longrightarrow\mathcal{X}' X ′ × X Y ⟶ X ′ is also a lax-ind-P \mathbf{P} P morphism. An argument similar to the proof of Lemma 2.8 proves as much.
Remark 2.9. Lemma 2.8 only solves a half of the natural question about the definition of an ind-P \mathbf{P} P morphism of strict ind-schemes. Let X \mathcal{X} X and Y \mathcal{Y} Y be strict ind-schemes, and f : Y ⟶ X f\colon\mathcal{Y} \longrightarrow\mathcal{X} f : Y ⟶ X be a lax-ind-P \mathbf{P} P morphism. Does it follow that that f f f can be represented by a directed diagram of morphisms ( f γ : Y γ → X γ ) γ ∈ Γ (f_\gamma\colon Y_\gamma\to X_\gamma)_{\gamma\in\Gamma} ( f γ : Y γ → X γ ) γ ∈ Γ in CSch such that all the morphisms f γ f_\gamma f γ have property P \mathbf{P} P and the transition morphisms in both the directed diagrams ( X γ ) γ ∈ Γ (X_\gamma)_{\gamma\in\Gamma} ( X γ ) γ ∈ Γ and ( Y γ ) γ ∈ Γ (Y_\gamma)_{\gamma\in\Gamma} ( Y γ ) γ ∈ Γ are closed immersions of schemes? We do not know.
For this reason, from now on and for the rest of this paper we will generally presume all our ind-schemes to be strict, and all our directed diagrams in CSch representing ind-schemes to have closed immersions as transition morphisms.
In particular, modifying the definition above, we will say that a morphism f f f of strict ind-concentrated ind-schemes is i n d \mathit{ind} ind -P if it can be represented by a directed diagram of morphisms ( f γ : Y γ ⟶ X γ ) γ ∈ Γ (f_\gamma: Y_\gamma\longrightarrow X_\gamma)_{\gamma\in\Gamma} ( f γ : Y γ ⟶ X γ ) γ ∈ Γ in CSch such that all the morphisms f γ f_\gamma f γ have property P and the transition morphisms in both the directed diagrams ( X γ ) γ ∈ Γ (X_\gamma)_{\gamma\in\Gamma} ( X γ ) γ ∈ Γ and ( Y γ ) γ ∈ Γ (Y_\gamma)_{\gamma\in\Gamma} ( Y γ ) γ ∈ Γ are closed immersions of schemes.
Let f : Y ⟶ X f:\mathcal{Y} \longrightarrow\mathcal{X} f : Y ⟶ X be an ind-P morphism of strict ind-concentrated ind-schemes and g : X ′ ⟶ X g:\mathcal{X}' \longrightarrow\mathcal{X} g : X ′ ⟶ X be any morphism of strict ind-concentrated ind-schemes. Then the morphism of strict ind-concentrated ind-schemes f ′ : X ′ × X Y ⟶ X ′ f':\mathcal{X}'\times_{\mathcal{X}}\mathcal{Y} \longrightarrow\mathcal{X}' f ′ : X ′ × X Y ⟶ X ′ is also ind-P. Indeed, once again arguing as in the proof of Lemma 2.8, it remains to point out that the transition morphisms of schemes X ξ ′ × X ξ Y ξ ⟶ X ν ′ × X ν Y ν X'_\xi\times_{X_\xi}Y_\xi\longrightarrow X'_\nu\times_{X_\nu}Y_\nu X ξ ′ × X ξ Y ξ ⟶ X ν ′ × X ν Y ν are closed immersions for all ξ ≤ ν ∈ Ξ \xi\leq\nu\in\Xi ξ ≤ ν ∈ Ξ , since the transition morphisms X ξ ′ ⟶ X ν ′ X'_\xi\longrightarrow X'_\nu X ξ ′ ⟶ X ν ′ , X ξ ⟶ X ν X_\xi\longrightarrow X_\nu X ξ ⟶ X ν , and Y ξ ⟶ Y ν Y_\xi\longrightarrow Y_\nu Y ξ ⟶ Y ν are closed immersions.
Let X \mathcal{X} X and Y ∈ Ind ( C S c h ) \mathcal{Y}\in\operatorname{Ind}(\mathrm{CSch}) Y ∈ Ind ( CSch ) be two (ind-concentrated) ind-schemes. One says that a morphism of ind-schemes Y ⟶ X \mathcal{Y}\longrightarrow\mathcal{X} Y ⟶ X is “representable by concentrated schemes” [53 Lemma 1.7 and Definition 1.8] , [33 Section 1.3] if, for any concentrated scheme T ∈ C S c h T\in\mathrm{CSch} T ∈ CSch and any morphism of ind-schemes T ⟶ X T\longrightarrow\mathcal{X} T ⟶ X , the fibered product T × X Y T\times_{\mathcal{X}}\mathcal{Y} T × X Y computed in Ind ( C S c h ) \operatorname{Ind}(\mathrm{CSch}) Ind ( CSch ) is an object of the full subcategory C S c h ⊂ Ind ( C S c h ) \mathrm{CSch}\subset\operatorname{Ind}(\mathrm{CSch}) CSch ⊂ Ind ( CSch ) (i. e., a concentrated scheme).
Let \mathcal{X}=\text{“}\lim_{\longrightarrow}\limits_{\gamma\in\Gamma}X_\gamma\text{”} be an ind-scheme represented by a directed diagram in CSch, and let f : Y ⟶ X f:\mathcal{Y}\longrightarrow\mathcal{X} f : Y ⟶ X be a morphism of ind-schemes that is “representable by (concentrated) schemes”. Put Y γ = X γ × X Y Y_\gamma=X_\gamma\times_{\mathcal{X}}\mathcal{Y} Y γ = X γ × X Y , and let f γ : Y γ ⟶ X γ f_\gamma:Y_\gamma\longrightarrow X_\gamma f γ : Y γ ⟶ X γ be the induced morphism of schemes. Then we have \mathcal{Y}=\text{“}\lim_{\longrightarrow}\limits_{\gamma\in\Gamma}Y_\gamma\text{”} and f=\text{“}\lim_{\longrightarrow}\limits_{\gamma\in\Gamma}f_\gamma\text{”} . Clearly, Y γ = X γ × X δ Y δ Y_\gamma=X_\gamma\times_{X_\delta}Y_\delta Y γ = X γ × X δ Y δ for all γ ≤ δ ∈ Γ \gamma\leq\delta\in\Gamma γ ≤ δ ∈ Γ .
Conversely, let f : Y ⟶ X f:\mathcal{Y}\longrightarrow\mathcal{X} f : Y ⟶ X be a morphism of ind-schemes represented by a directed diagram ( f γ : Y γ ⟶ X γ ) γ ∈ Γ (f_\gamma:Y_\gamma\longrightarrow X_\gamma)_{\gamma\in\Gamma} ( f γ : Y γ ⟶ X γ ) γ ∈ Γ of morphisms in CSch such that one has Y γ = X γ × X δ Y δ Y_\gamma=X_\gamma\times_{X_\delta}Y_\delta Y γ = X γ × X δ Y δ for all γ ≤ δ ∈ Γ \gamma\leq\delta\in\Gamma γ ≤ δ ∈ Γ . Then it follows that Y γ = X γ × X Y Y_\gamma=X_\gamma\times_{\mathcal{X}}\mathcal{Y} Y γ = X γ × X Y . For any concentrated scheme T T T and any morphism of ind-schemes T ⟶ X T\longrightarrow\mathcal{X} T ⟶ X , there exists an index γ 0 ∈ Γ \gamma_0\in\Gamma γ 0 ∈ Γ such that the morphism T ⟶ X T\longrightarrow\mathcal{X} T ⟶ X factorizes as T ⟶ X γ 0 ⟶ X T\longrightarrow X_{\gamma_0}\longrightarrow\mathcal{X} T ⟶ X γ 0 ⟶ X . Then T × X Y = T × X γ 0 ( X γ 0 × X Y ) = T × X γ 0 Y γ 0 T\times_{\mathcal{X}}\mathcal{Y}=T\times_{X_{\gamma_0}}(X_{\gamma_0}\times_{\mathcal{X}}\mathcal{Y})=T\times_{X_{\gamma_0}}Y_{\gamma_0} T × X Y = T × X γ 0 ( X γ 0 × X Y ) = T × X γ 0 Y γ 0 is a concentrated scheme, so the morphism f f f is “representable by schemes”.
A morphism of ind-schemes f f f is said to have property P (as distinguished from ind-P) if f f f is simultaneously ind-P and “representable by schemes”.
It follows immediately from the definition that the class of all morphisms of ind-schemes that are “representable by (concentrated) schemes” is stable under pullbacks. Hence the class of all morphisms of ind-schemes having property P is preserved by the pullbacks, too.
Lemma 2.10. Let P be a property of morphisms in CSch that is stable under pullbacks. Then a morphism of ind-schemes f : Y ⟶ X f:\mathcal{Y}\longrightarrow\mathcal{X} f : Y ⟶ X has property P if and only if, for any concentrated scheme T T T and any morphism of ind-schemes T ⟶ X T\longrightarrow\mathcal{X} T ⟶ X , the
fibered product T × X Y T \times_{\mathfrak X} \mathfrak Y T × X Y is a concentrated scheme and the induced morphism of schemes T × X Y ⟶ T T \times_{\mathfrak X} \mathfrak Y \longrightarrow T T × X Y ⟶ T has property P.
Proof. The “if” implication follows from the representation of the morphism f f f as f = l i m → γ ∈ Γ f γ f = \varinjlim_{\gamma\in\Gamma} f_\gamma f = lim γ ∈ Γ f γ with f γ : Y γ = X γ × X Y ⟶ X γ f_\gamma: Y_\gamma= X_\gamma\times_{\mathfrak X} \mathfrak Y \longrightarrow X_\gamma f γ : Y γ = X γ × X Y ⟶ X γ , in the notation above.
To prove the “only if”, pick a directed diagram of morphisms of concentrated schemes ( f γ : Y γ ⟶ X γ ) γ ∈ Γ (f_\gamma: Y_\gamma\longrightarrow X_\gamma)_{\gamma\in\Gamma} ( f γ : Y γ ⟶ X γ ) γ ∈ Γ such that all the morphisms f γ f_\gamma f γ have property P and f = l i m → γ ∈ Γ f γ f = \varinjlim_{\gamma\in\Gamma} f_\gamma f = lim γ ∈ Γ f γ . Let T ⟶ X T \longrightarrow\mathfrak X T ⟶ X be a morphism into X \mathfrak X X from a concentrated scheme T T T . Then there exists an index γ 0 ∈ Γ \gamma_0 \in\Gamma γ 0 ∈ Γ such that the morphism T ⟶ X T \longrightarrow\mathfrak X T ⟶ X factorizes as T ⟶ X γ 0 ⟶ X T \longrightarrow X_{\gamma_0} \longrightarrow\mathfrak X T ⟶ X γ 0 ⟶ X . Following the discussion in Section 2.2, we have T × X Y = l i m → γ ≥ γ 0 ( T × X γ Y γ ) T \times_{\mathfrak X} \mathfrak Y = \varinjlim_{\gamma\geq\gamma_0} (T \times_{X_\gamma} Y_\gamma) T × X Y = lim γ ≥ γ 0 ( T × X γ Y γ ) . Since the property P is stable under pullbacks in CSch, the morphisms of schemes T × X γ Y γ ⟶ T T \times_{X_\gamma} Y_\gamma\longrightarrow T T × X γ Y γ ⟶ T have property P. By assumption, the morphism of ind-schemes f f f is “representable by schemes”, so we have T × X Y ∈ CSch ⊂ Ind ( CSch ) T \times_{\mathfrak X} \mathfrak Y \in\operatorname{CSch} \subset\operatorname{Ind}(\operatorname{CSch}) T × X Y ∈ CSch ⊂ Ind ( CSch ) .
According to the convention of and the definition in Remark 2.9, we assume that the transitions morphisms in the directed diagrams ( X γ ) γ ∈ Γ (X_\gamma)_{\gamma\in\Gamma} ( X γ ) γ ∈ Γ and ( Y γ ) γ ∈ Γ (Y_\gamma)_{\gamma\in\Gamma} ( Y γ ) γ ∈ Γ are closed immersions of schemes. Then all the transition morphisms in the directed diagram ( T × X γ Y γ ) γ ∈ Γ (T \times_{X_\gamma} Y_\gamma)_{\gamma\in\Gamma} ( T × X γ Y γ ) γ ∈ Γ are closed immersions of schemes, too. Since T × X Y ∈ CSch T \times_{\mathfrak X} \mathfrak Y \in\operatorname{CSch} T × X Y ∈ CSch and the closed immersions are monomorphisms in CSch, it follows that there exists an index γ ∈ Γ \gamma\in\Gamma γ ∈ Γ for which T × X Y = T × X γ Y γ T \times_{\mathfrak X} \mathfrak Y = T \times_{X_\gamma} Y_\gamma T × X Y = T × X γ Y γ . Thus the morphism of schemes T × X Y ⟶ T T \times_{\mathfrak X} \mathfrak Y \longrightarrow T T × X Y ⟶ T has property P, as desired. □
We have proved that a morphism of ind-schemes f f f has property P if and only if it can be represented by a directed diagram ( f γ : Y γ ⟶ X γ ) γ ∈ Γ (f_\gamma: Y_\gamma\longrightarrow X_\gamma)_{\gamma\in\Gamma} ( f γ : Y γ ⟶ X γ ) γ ∈ Γ of morphisms in CSch such that Y γ = X γ × X Y Y_\gamma= X_\gamma\times_{\mathfrak X} \mathfrak Y Y γ = X γ × X Y for all γ ∈ Γ \gamma\in\Gamma γ ∈ Γ (or equivalently, Y γ = X γ × X δ Y δ Y_\gamma= X_\gamma\times_{X_\delta} Y_\delta Y γ = X γ × X δ Y δ for all γ ≤ δ ∈ Γ \gamma\leq\delta\in\Gamma γ ≤ δ ∈ Γ ) and all the morphisms f γ f_\gamma f γ have property P.
In particular, taking P to be the property of a morphism of schemes to be affine, we obtain the notions of an affine and ind-affine morphism of ind-schemes, as well as the notion of a lax-ind-affine morphism of ind-schemes. Every morphism of ind-affine ind-schemes is ind-affine, but it need not be affine.
Conversely, if f : V ⟶ U f : \mathfrak V \longrightarrow\mathfrak U f : V ⟶ U is a lax-ind-affine morphism of ind-schemes and the ind-scheme U \mathfrak U U is ind-affine, then the ind-scheme V \mathfrak V V is ind-affine, too. Indeed, by Lemma 2.8 we have f = l i m → γ ∈ Γ f γ f = \varinjlim_{\gamma\in\Gamma} f_\gamma f = lim γ ∈ Γ f γ , where ( f γ : V γ ⟶ U γ ) γ ∈ Γ (f_\gamma: V_\gamma\longrightarrow U_\gamma)_{\gamma\in\Gamma} ( f γ : V γ ⟶ U γ ) γ ∈ Γ is a directed diagram of affine morphisms of concentrated schemes and the transition morphisms in the direct directed diagram of schemes ( U γ ) γ ∈ Γ (U_\gamma)_{\gamma\in\Gamma} ( U γ ) γ ∈ Γ are closed immersions. By Corollary 2.3(e), all the schemes U γ U_\gamma U γ are affine, and it follows that all the schemes V γ V_\gamma V γ are affine, too.
Taking P to be the property of a morphism of schemes to be a closed immersion, we obtain the notions of a closed immersion and an ind-closed immersion of ind-schemes. In particular, a morphism Z ⟶ X Z \longrightarrow\mathfrak X Z ⟶ X from a concentrated scheme to an (ind-concentrated) ind-scheme is a closed immersion if and only if there exist a representation of X \mathfrak X X by a directed diagram (of closed immersions) of schemes ( X γ ) γ ∈ Γ (X_\gamma)_{\gamma\in\Gamma} ( X γ ) γ ∈ Γ , an index γ 0 ∈ Γ \gamma_0 \in\Gamma γ 0 ∈ Γ , and an isomorphism of schemes Z ≅ X γ 0 Z \cong X_{\gamma_0} Z ≅ X γ 0 such that the triangular diagram Z ≅ X γ 0 ⟶ X Z \cong X_{\gamma_0} \longrightarrow\mathfrak X Z ≅ X γ 0 ⟶ X is commutative.
A closed subscheme Z ⊂ X Z \subset\mathfrak X Z ⊂ X of an ind-scheme X \mathfrak X X is an equivalence class of pairs (a concentrated scheme Z Z Z , a closed immersion Z ⟶ X Z \longrightarrow\mathfrak X Z ⟶ X ) with respect to the obvious equivalence relation. Specifically, two closed immersions Z ′ ⟶ X Z^{\prime} \longrightarrow\mathfrak{X} Z ′ ⟶ X and Z ′ ′ ⟶ X Z^{\prime\prime} \longrightarrow\mathfrak{X} Z ′′ ⟶ X are said to be equivalent if there exists an isomorphism Z ′ ≃ Z ′ ′ Z^{\prime} \simeq Z^{\prime\prime} Z ′ ≃ Z ′′ making the triangular diagram Z ′ ≃ Z ′ ′ ⟶ X Z^{\prime} \simeq Z^{\prime\prime} \longrightarrow\mathfrak{X} Z ′ ≃ Z ′′ ⟶ X commutative.
Given two closed subschemes Z ′ Z^{\prime} Z ′ and Z ′ ′ Z^{\prime\prime} Z ′′ in X \mathfrak{X} X , one writes Z ′ ⊂ Z ′ ′ Z^{\prime} \subset Z^{\prime\prime} Z ′ ⊂ Z ′′ if the closed immersion morphism Z ′ ⟶ X Z^{\prime} \longrightarrow\mathfrak{X} Z ′ ⟶ X factorizes through the closed immersion morphism Z ′ ′ ⟶ X Z^{\prime\prime} \longrightarrow\mathfrak{X} Z ′′ ⟶ X . If this is the case, then the morphism of schemes Z ′ ⟶ Z ′ ′ Z^{\prime} \longrightarrow Z^{\prime\prime} Z ′ ⟶ Z ′′ is a closed immersion.
Example 2.11. Let k k k be a field and x x x be a variable. Then the inclusion i : k [ x ] ⟶ k [ [ x ] ] i:k[x] \longrightarrow k[[x]] i : k [ x ] ⟶ k [[ x ]] of the discrete ring of polynomials k [ x ] k[x] k [ x ] into the topological ring of formal power series k [ [ x ] ] k[[x]] k [[ x ]] (with the adic topology on k [ [ x ] ] k[[x]] k [[ x ]] ) induces, by the anti-equivalence of categories from Lemma 2.7, a morphism of ind-affine ind-schemes i : Spf k [ [ x ] ] ⟶ Spec k [ x ] i:\operatorname{Spf} k[[x]] \longrightarrow\operatorname{Spec} k[x] i : Spf k [[ x ]] ⟶ Spec k [ x ] . The morphism of ind-affine ind-schemes i i i is an ind-closed immersion, but not a closed immersion. In fact, if an ind-scheme X \mathfrak{X} X admits a closed immersion (or more generally, a “morphism representable by schemes”) into a concentrated scheme, then X \mathfrak{X} X is a concentrated scheme.
Open ind-subshemes
In the context of Section 2.4, take P \mathbf{P} P to be the property of a morphism of schemes to be an open immersion. This produces the notions of an open immersion and an ind-open immersion of ind-schemes.
Example 2.12. The following counterexample shows that an ind-open immersion of (ind-concentrated) ind-schemes need not be an open immersion. Let Ξ \Xi Ξ be an infinite set and ( U ξ = Spec R ξ ) ξ ∈ Ξ (U_{\xi}=\operatorname{Spec} R_{\xi})_{\xi\in\Xi} ( U ξ = Spec R ξ ) ξ ∈ Ξ be a family of nonempty affine schemes indexed by Ξ \Xi Ξ , as in Remark 2.1. Following the notation of that remark, put X = ∐ ξ ∈ Ξ U ξ X=\coprod_{\xi\in\Xi}U_{\xi} X = ∐ ξ ∈ Ξ U ξ and X = l i m → γ ∈ Γ ∏ ξ ∈ γ U ξ \mathfrak{X}=\varinjlim_{\gamma\in\Gamma}\prod_{\xi\in\gamma}U_{\xi} X = lim γ ∈ Γ ∏ ξ ∈ γ U ξ , where Γ \Gamma Γ is the directed poset of all finite subsets of Ξ \Xi Ξ . Then the non-quasi-compact scheme X X X represents the ind-concentrated ind-scheme X \mathfrak{X} X .
Consider the infinite product ring S = ∏ ξ ∈ Ξ R ξ S=\prod_{\xi\in\Xi}R_{\xi} S = ∏ ξ ∈ Ξ R ξ and the affine scheme V = Spec S V=\operatorname{Spec} S V = Spec S . Then, for every index ξ ∈ Ξ \xi\in\Xi ξ ∈ Ξ , the affine scheme U ξ U_{\xi} U ξ is an open-closed subscheme in the affine scheme V V V ; so V V V is the disjoint/disconnected union of the affine schemes U ξ U_{\xi} U ξ and V ∖ U ξ V\setminus U_{\xi} V ∖ U ξ . It follows that X X X is an open subscheme in V V V . The natural morphism of ind-schemes X ⟶ V \mathfrak{X}\longrightarrow V X ⟶ V is an ind-open immersion, but it is not an open immersion in our terminology, as it is not “representable by (concentrated) schemes”.
An open ind-subscheme U ⊂ X \mathcal{U}\subset\mathfrak{X} U ⊂ X of an ind-scheme X \mathfrak{X} X is an equivalence class of pairs (an ind-scheme U \mathcal{U} U , an open immersion U ⟶ X \mathcal{U}\longrightarrow\mathfrak{X} U ⟶ X ) with respect to the obvious equivalence relation. Specifically, two open immersions U ′ ⟶ X \mathcal{U}^{\prime}\longrightarrow\mathfrak{X} U ′ ⟶ X and U ′ ′ ⟶ X \mathcal{U}^{\prime\prime}\longrightarrow\mathfrak{X} U ′′ ⟶ X are said to be equivalent if there exists an isomorphism U ′ ≃ U ′ ′ \mathcal{U}^{\prime}\simeq\mathcal{U}^{\prime\prime} U ′ ≃ U ′′ making the triangular diagram U ′ ≃ U ′ ′ ⟶ X \mathcal{U}^{\prime}\simeq\mathcal{U}^{\prime\prime}\longrightarrow\mathfrak{X} U ′ ≃ U ′′ ⟶ X commutative. (Cf. [53 Definition 1.13] .)
The inclusion order on the set of all open ind-schemes U ⊂ X \mathcal{U}\subset\mathfrak{X} U ⊂ X is also defined in the obvious way. One has U ′ ⊂ U ′ ′ \mathcal{U}^{\prime}\subset\mathcal{U}^{\prime\prime} U ′ ⊂ U ′′ if the open immersion U ′ ⟶ X \mathcal{U}^{\prime}\longrightarrow\mathfrak{X} U ′ ⟶ X factorizes through the open immersion U ′ ′ ⟶ X \mathcal{U}^{\prime\prime}\longrightarrow\mathfrak{X} U ′′ ⟶ X . In this case, the morphism U ′ ⟶ U ′ ′ \mathcal{U}^{\prime}\longrightarrow\mathcal{U}^{\prime\prime} U ′ ⟶ U ′′ is an open immersion of ind-schemes as well.
For any scheme X X X , one can consider its underlying topological space (which we denote also by X X X ). Given a (strict ind-concentrated) ind-scheme X \mathfrak{X} X , its underlying topological space X \mathfrak{X} X is constructed as follows. Let ( X γ ) γ ∈ Γ (X_\gamma)_{\gamma\in\Gamma} ( X γ ) γ ∈ Γ be a directed diagram of (concentrated) schemes and closed immersions of schemes representing X \mathcal{X} X . Then the topological space X \mathcal{X} X is the inductive limit (the union) of the topological spaces X γ X_\gamma X γ , endowed with the inductive limit topology. So a subset V ⊂ X \mathcal{V} \subset\mathcal{X} V ⊂ X is said to be open if V ∩ X γ \mathcal{V} \cap X_\gamma V ∩ X γ is an open subset of X γ X_\gamma X γ for every γ ∈ Γ \gamma\in\Gamma γ ∈ Γ . Clearly, this definition of the underlying topological space of X \mathcal{X} X does not depend on the choice of a directed diagram of closed immersions of schemes representing X \mathcal{X} X .
Let us say that an open subset V ⊂ X \mathcal{V} \subset\mathcal{X} V ⊂ X is ind-quasi-compact if the open subset V ∩ X γ ⊂ X γ \mathcal{V} \cap X_\gamma\subset X_\gamma V ∩ X γ ⊂ X γ is quasi-compact for every γ ∈ Γ \gamma\in\Gamma γ ∈ Γ .
Lemma 2.13. For any ind-scheme X \mathcal{X} X , there is a natural order-preserving bijection between the poset of all open ind-subschemes in X \mathcal{X} X and the poset of all ind-quasi-compact open subsets in the underlying topological space of X \mathcal{X} X , both ordered by inclusion. To any open ind-subscheme j : U ⟶ X \mathrm{j}: \mathcal{U} \longrightarrow\mathcal{X} j : U ⟶ X , the open subset j ( U ) ⊂ X \mathrm{j}(\mathcal{U}) \subset\mathcal{X} j ( U ) ⊂ X of the underlying topological space of X \mathcal{X} X is assigned.
Proof. The proof is left to the reader. □
To keep things uncomplicated, let k k k be a field of cardinality at least 3. Let R n = k [ x 1 , … , x n ] R_n = k[x_1,\ldots,x_n] R n = k [ x 1 , … , x n ] , n ≥ 0 n \ge0 n ≥ 0 , be the rings of polynomials in n n n variables over k k k . Then we have the natural isomorphisms of k k k -algebras R n + 1 / ( x n + 1 ) ≃ R n R_{n+1}/(x_{n+1}) \simeq R_n R n + 1 / ( x n + 1 ) ≃ R n taking x i x_i x i to x i x_i x i for all 1 ≤ i ≤ n 1 \le i \le n 1 ≤ i ≤ n . This produces a projective system of surjective homomorphisms of commutative rings R 0 ⟵ R 1 ⟵ R 2 ⟵ ⋯ R_0 \longleftarrow R_1 \longleftarrow R_2 \longleftarrow\cdots R 0 ⟵ R 1 ⟵ R 2 ⟵ ⋯ . Put X n = Spec R n X_n = \operatorname{Spec} R_n X n = Spec R n , and let X 0 ⟶ X 1 ⟶ X 2 ⟶ ⋯ X_0 \longrightarrow X_1 \longrightarrow X_2 \longrightarrow\cdots X 0 ⟶ X 1 ⟶ X 2 ⟶ ⋯ be the corresponding inductive system of closed immersions of affine schemes. Then the underlying topological space of the strict ind-affine ℵ 0 \aleph_0 ℵ 0 -ind-scheme X = l i m → n ≥ 0 X n \mathcal{X} = \underset{n \ge0}{\varinjlim}\, X_n X = n ≥ 0 lim X n is not quasi-compact.
Indeed, let a , b ∈ k a,b \in k a , b ∈ k be any two different nonzero elements, a ≠ b a \ne b a = b , a ≠ 0 ≠ b a \ne0 \ne b a = 0 = b . Consider the Cartesian power set F = { a , b } Z ≥ 1 F = \{a,b\}^{\mathbb{Z}_{\ge1}} F = { a , b } Z ≥ 1 . The elements of F F F are arbitrary functions ϕ : Z ≥ 1 ⟶ { a , b } \phi\colon\mathbb{Z}_{\ge1} \longrightarrow\{a,b\} ϕ : Z ≥ 1 ⟶ { a , b } . Given such a function ϕ \phi ϕ , put R n , ϕ = R n [ ( x 1 − ϕ ( 1 ) ) − 1 , … , ( x n − ϕ ( n ) ) − 1 ] R_{n,\phi} = R_n[(x_1-\phi(1))^{-1},\ldots,(x_n-\phi(n))^{-1}] R n , ϕ = R n [( x 1 − ϕ ( 1 ) ) − 1 , … , ( x n − ϕ ( n ) ) − 1 ] . Here x i − ϕ ( i ) ∈ k [ x i ] ⊂ R n x_i-\phi(i) \in k[x_i] \subset R_n x i − ϕ ( i ) ∈ k [ x i ] ⊂ R n (for 1 ≤ i ≤ n 1 \le i \le n 1 ≤ i ≤ n ) is a polynomial of degree 1. So X n , ϕ = Spec R n , ϕ X_{n,\phi} = \operatorname{Spec} R_{n,\phi} X n , ϕ = Spec R n , ϕ is a principal affine open subscheme of the affine scheme X n X_n X n .
Similarly to the above, we have natural isomorphism of k k k -algebras R n + 1 , ϕ / ( x n + 1 ) ≃ R n , ϕ R_{n+1,\phi}/(x_{n+1}) \simeq R_{n,\phi} R n + 1 , ϕ / ( x n + 1 ) ≃ R n , ϕ taking x i x_i x i to x i x_i x i for all 1 ≤ i ≤ n 1 \le i \le n 1 ≤ i ≤ n , and the related projective system of surjective homomorphisms of commutative rings R 0 , ϕ ⟵ R 1 , ϕ ⟵ R 2 , ϕ ⟵ ⋯ R_{0,\phi} \longleftarrow R_{1,\phi} \longleftarrow R_{2,\phi} \longleftarrow\cdots R 0 , ϕ ⟵ R 1 , ϕ ⟵ R 2 , ϕ ⟵ ⋯ . Let X 0 , ϕ ⟶ X 1 , ϕ ⟶ X 2 , ϕ ⟶ ⋯ X_{0,\phi} \longrightarrow X_{1,\phi} \longrightarrow X_{2,\phi} \longrightarrow\cdots X 0 , ϕ ⟶ X 1 , ϕ ⟶ X 2 , ϕ ⟶ ⋯ be the corresponding inductive system of closed immersions of affine schemes. Then X ϕ = l i m → n ≥ 0 X n , ϕ \mathcal{X}_\phi= \underset{n \ge0}{\varinjlim}\, X_{n,\phi} X ϕ = n ≥ 0 lim X n , ϕ is an open ind-subscheme in X \mathcal{X} X .
Let C ⊂ F C \subset F C ⊂ F be any subset such that, for any function ϕ ∈ F \phi\in F ϕ ∈ F and any integer n ≥ 1 n \ge1 n ≥ 1 , there exists a function ψ ∈ C \psi\in C ψ ∈ C for which ϕ ( i ) = ψ ( i ) \phi(i) = \psi(i) ϕ ( i ) = ψ ( i ) for all 1 ≤ i ≤ n 1 \le i \le n 1 ≤ i ≤ n . One can easily make the set C C C countable, but it cannot be made finite. Now X = ⋃ ψ ∈ C X ψ \mathcal{X} = \bigcup_{\psi\in C} \mathcal{X}_\psi X = ⋃ ψ ∈ C X ψ is an infinite open covering of X \mathcal{X} X that has no finite subcovering. Indeed, one has X n = ⋃ ψ ∈ C X n , ψ X_n = \bigcup_{\psi\in C} X_{n,\psi} X n = ⋃ ψ ∈ C X n , ψ for every n ≥ 0 n \ge0 n ≥ 0 , and this open covering of the affine scheme X n X_n X n has no subcovering consisting of fewer than 2 n 2^n 2 n open subsets. Another counterexample of non-quasi-compact strict ind-affine ℵ 0 \aleph_0 ℵ 0 -ind-scheme is provided by the ind-scheme X \mathcal{X} X from Remark 2.1 and Example 2.12.
Ind-affine-finite ind-schemes
Let X \mathcal{X} X be a strict ind-concentrated ind-scheme. We will say that the ind-scheme X \mathcal{X} X is ind-affine-finite if there exists a finite collection of open ind-subschemes U 1 , … , U N ⊂ X \mathcal{U}_1,\ldots,\mathcal{U}_N\subset\mathcal{X} U 1 , … , U N ⊂ X such that the underlying topological space of X \mathcal{X} X is the union of the underlying topological spaces of U α \mathcal{U}_\alpha U α , that is X = ⋃ α = 1 N U α \mathcal{X}=\bigcup_{\alpha=1}^N\mathcal{U}_\alpha X = ⋃ α = 1 N U α .
The latter condition can be also expressed in the language of schemes and ind-schemes as follows. In the context of § 2.4 , consider the property P \mathbf{P} P of a morphism of schemes to be surjective. The class of all surjective morphisms of schemes is closed under pullbacks by [21 Lemma Tag 01S1] . This produces the notions of a surjective and an ind-surjective morphism of ind-schemes.
Given a finite collection of open ind-subschemes U 1 , … , U N ⊂ X \mathcal{U}_1,\ldots,\mathcal{U}_N\subset\mathcal{X} U 1 , … , U N ⊂ X , consider the disjoint/disconnected union of ind-schemes U = ∐ α = 1 N U α \mathcal{U}=\coprod_{\alpha=1}^N\mathcal{U}_\alpha U = ∐ α = 1 N U α . Then we have a natural morphism of ind-schemes U ⟶ X \mathcal{U}\longrightarrow\mathcal{X} U ⟶ X , which is obviously “representable by (concentrated) schemes” (since so are the open immersions U α ⟶ X \mathcal{U}_\alpha\longrightarrow\mathcal{X} U α ⟶ X ). One has X = ⋃ α = 1 N U α \mathcal{X}=\bigcup_{\alpha=1}^N\mathcal{U}_\alpha X = ⋃ α = 1 N U α if and only if the morphism of ind-schemes U ⟶ X \mathcal{U}\longrightarrow\mathcal{X} U ⟶ X is surjective, or equivalently, ind-surjective.
Example 2.15. Let X \mathcal{X} X be an ind-affine-finite ind-scheme and f : Y ⟶ X f:\mathcal{Y}\longrightarrow\mathcal{X} f : Y ⟶ X be a lax-ind-affine morphism of ind-schemes. (In particular, any ind-affine morphism of ind-schemes is lax-ind-affine.) Then the ind-scheme Y \mathcal{Y} Y is also ind-affine-finite.
Specifically, let X = ⋃ α = 1 N U α \mathcal{X}=\bigcup_{\alpha=1}^N\mathcal{U}_\alpha X = ⋃ α = 1 N U α be a finite ind-affine open covering of X \mathcal{X} X . Put V α = U α × X Y \mathcal{V}_\alpha=\mathcal{U}_\alpha\times_{\mathcal{X}}\mathcal{Y} V α = U α × X Y . Following the discussion of the preservation by pullbacks in § 2.4 , the morphisms of ind-schemes V α ⟶ Y \mathcal{V}_\alpha\longrightarrow\mathcal{Y} V α ⟶ Y are open immersions, while the morphisms V α ⟶ U α \mathcal{V}_\alpha\longrightarrow\mathcal{U}_\alpha V α ⟶ U α are lax-ind-affine. According to the discussion of lax-ind-affine morphisms in § 2.4 , it follows that the ind-schemes V α \mathcal{V}_\alpha V α are ind-affine. Finally, one has Y = ⋃ α = 1 N V α \mathcal{Y}=\bigcup_{\alpha=1}^N\mathcal{V}_\alpha Y = ⋃ α = 1 N V α (e. g., since the classes of surjective and ind-surjective morphisms of ind-schemes are closed under pullbacks according to § 2.4 ).
For any two categories K \mathbf{K} K and L \mathbf{L} L , any functor F : K ⟶ L F:\mathbf{K}\longrightarrow\mathbf{L} F : K ⟶ L induces a functor on the categories of ind-objects Ind F : Ind K ⟶ Ind L \operatorname{Ind}F:\operatorname{Ind}\mathbf{K}\longrightarrow\operatorname{Ind}\mathbf{L} Ind F : Ind K ⟶ Ind L defined by the obvious rule ( Ind F ) ( ‘ ‘ l i m → γ ∈ Γ X γ ‘ ‘ ) = ‘ ‘ l i m → γ ∈ Γ F ( X γ ) ‘ ‘ (\operatorname{Ind}F)(``\varinjlim_{\gamma\in\Gamma}X_\gamma``)=``\varinjlim_{\gamma\in\Gamma}F(X_\gamma)`` ( Ind F ) ( ‘‘ lim γ ∈ Γ X γ ‘‘ ) = ‘‘ lim γ ∈ Γ F ( X γ ) ‘‘ . In particular, for any concentrated scheme X X X , the maximal reduced closed subscheme X r e d ⊂ X X_{\mathrm{red}}\subset X X red ⊂ X is a concentrated scheme depending functorially on X X X . Using the construction above, the functor X ⟼ X r e d : C S c h ⟶ C S c h X\longmapsto X_{\mathrm{red}}:\mathrm{CSch}\longrightarrow\mathrm{CSch} X ⟼ X red : CSch ⟶ CSch induces a functor Ind ( C S c h ) ⟶ Ind ( C S c h ) \operatorname{Ind}(\mathrm{CSch})\longrightarrow\operatorname{Ind}(\mathrm{CSch}) Ind ( CSch ) ⟶ Ind ( CSch ) , which we also denote by X ⟼ X r e d \mathcal{X}\longmapsto\mathcal{X}_{\mathrm{red}} X ⟼ X red .
Clearly, the functor X ⟼ X r e d X\longmapsto X_{\mathrm{red}} X ⟼ X red takes closed immersions of schemes to closed immersions of schemes; hence, for any strict ind-scheme X \mathcal{X} X , the ind-scheme X r e d \mathcal{X}_{\mathrm{red}} X red is also strict. Similarly, the functor X ⟼ X r e d \mathcal{X}\longmapsto\mathcal{X}_{\mathrm{red}} X ⟼ X red preserves the ind-affineness, ind-semi-separatedness, and ind-Noetherianity of ind-schemes, and takes ℵ 0 \aleph_0 ℵ 0 -ind-schemes to ℵ 0 \aleph_0 ℵ 0 -ind-schemes.
In the context of the theory of ind-schemes, a (strict ind-concentrated) ind-scheme X \mathcal{X} X is said to be a formal scheme if the ind-scheme X r e d \mathcal{X}_{\mathrm{red}} X red is a concentrated scheme, i. e., X r e d ∈ C S c h ⊂ Ind ( C S c h ) \mathcal{X}_{\mathrm{red}}\in\mathrm{CSch}\subset\operatorname{Ind}(\mathrm{CSch}) X red ∈ CSch ⊂ Ind ( CSch ) . Given a representation X = ‘ ‘ l i m → γ ∈ Γ X γ ‘ ‘ \mathcal{X}=``\varinjlim_{\gamma\in\Gamma}X_\gamma`` X = ‘‘ lim γ ∈ Γ X γ ‘‘ of X \mathcal{X} X by a directed diagram of closed immersions of concentrated schemes ( X γ ) γ ∈ Γ (X_\gamma)_{\gamma\in\Gamma} ( X γ ) γ ∈ Γ , the ind-scheme X \mathfrak{X} X is a formal scheme if and only if there exists an index γ 0 ∈ Γ \gamma_0\in\Gamma γ 0 ∈ Γ for which the morphism ( X γ 0 ) r e d ⟶ X r e d (X_{\gamma_0})_{\mathrm{red}}\longrightarrow\mathfrak{X}_{\mathrm{red}} ( X γ 0 ) red ⟶ X red is an isomorphism. Equivalently, the morphisms ( X γ 0 ) r e d ⟶ ( X γ ) r e d (X_{\gamma_0})_{\mathrm{red}}\longrightarrow(X_\gamma)_{\mathrm{red}} ( X γ 0 ) red ⟶ ( X γ ) red are isomorphisms for all γ ≥ γ 0 \gamma\geq\gamma_0 γ ≥ γ 0 . So an ind-scheme is a formal scheme if and only if it can be represented by a directed diagram of bijective closed immersions of concentrated schemes.
A Noetherian formal scheme is an ind-scheme that is simultaneously an ind-Noetherian ind-scheme and a formal scheme. Similarly, an affine formal scheme is an ind-scheme that is simultaneously an ind-affine ind-scheme and a formal scheme. The definition of a semi-separated formal scheme is also similar. So a formal scheme is Noetherian if and only if it can be represented by a directed diagram of bijective closed immersions of Noetherian schemes, and similarly for affine/semi-separated formal schemes (by Corollary 2.3(b,c,e) and the subsequent paragraph).
Example 2.16. Any Noetherian formal scheme X \mathfrak{X} X is ind-affine-finite. Indeed, let ( X γ ) γ ∈ Γ (X_\gamma)_{\gamma\in\Gamma} ( X γ ) γ ∈ Γ be a directed diagram of bijective closed immersions of Noetherian schemes representing X \mathfrak{X} X . So we have ( X γ ) r e d = X = X r e d (X_\gamma)_{\mathrm{red}}=X=\mathfrak{X}_{\mathrm{red}} ( X γ ) red = X = X red for all γ ∈ Γ \gamma\in\Gamma γ ∈ Γ .
Let X = ⋃ α = 1 N U α X=\bigcup_{\alpha=1}^{N}U_\alpha X = ⋃ α = 1 N U α be a finite affine open covering of the Noetherian scheme X X X . For every index γ ∈ Γ \gamma\in\Gamma γ ∈ Γ , the underlying topological space of X γ X_\gamma X γ is naturally identified with the underlying topological space of X X X ; and U α U_\alpha U α is an open subset of this topological space. So we have a uniquely defined open subscheme U α , γ ⊂ X γ U_{\alpha,\gamma}\subset X_\gamma U α , γ ⊂ X γ corresponding to this open subset. In other words, ( U α , γ ) r e d = U α (U_{\alpha,\gamma})_{\mathrm{red}}=U_\alpha ( U α , γ ) red = U α . The scheme U α , γ U_{\alpha,\gamma} U α , γ is Noetherian, and by [21 Lemma Tag 01YQ] it follows that U α , γ U_{\alpha,\gamma} U α , γ is affine.
It remains to put U α = ‘‘lim" γ ∈ Γ U α , γ \mathcal{U}_\alpha=\underset{\gamma\in\Gamma}{\operatorname{``lim"}}\,U_{\alpha,\gamma} U α = γ ∈ Γ ‘‘lim" U α , γ . Then U α \mathcal{U}_\alpha U α is an ind-affine open ind-subscheme in X \mathfrak{X} X , and X = ⋃ α = 1 N U α \mathfrak{X}=\bigcup_{\alpha=1}^{N}\mathcal{U}_\alpha X = ⋃ α = 1 N U α .
Remark 2.17. A strict ind-quasi-compact ind-(quasi)-separated ℵ 0 \aleph_0 ℵ 0 -ind-scheme need not be ind-affine-finite. Indeed, let X = ⋃ α = 1 N U α \mathfrak{X}=\bigcup_{\alpha=1}^{N}\mathcal{U}_\alpha X = ⋃ α = 1 N U α be an ind-affine-finite strict ind-concentrated ind-scheme, covered by N N N ind-affine open ind-subschemes U α \mathcal{U}_\alpha U α . Let Z ⊂ X Z\subset\mathfrak{X} Z ⊂ X be a closed subscheme. Then the schemes Z × X U α Z\times_{\mathfrak{X}}\mathcal{U}_\alpha Z × X U α are affine as closed subschemes in the ind-affine ind-schemes U α \mathcal{U}_\alpha U α . So Z = ⋃ α = 1 N ( Z × X U α ) Z=\bigcup_{\alpha=1}^{N}(Z\times_{\mathfrak{X}}\mathcal{U}_\alpha) Z = ⋃ α = 1 N ( Z × X U α ) is an affine open covering of Z Z Z by N N N affine open subschemes. Thus the minimal number of covering open affines for closed subschemes Z ⊂ X Z\subset\mathfrak{X} Z ⊂ X is uniformly bounded by one constant N N N common for all the closed subschemes Z ⊂ X Z\subset\mathfrak{X} Z ⊂ X .
For a specific counterexample, let k k k be a field and V = k ( ω ) V=k^{(\omega)} V = k ( ω ) be a k k k -vector space of countably infinite dimension. Then the projectivization P ( V ) \mathbb{P}(V) P ( V ) is a strict ind-quasi-compact ind-separated ℵ 0 \aleph_0 ℵ 0 -ind-scheme such that, for every field extension K / k K/k K / k , the K K K -points of P ( V ) \mathbb{P}(V) P ( V ) correspond bijectively to one-dimensional vector subspaces in the K K K -vector space K ⊗ k V K\otimes_k V K ⊗ k V . Simply put, P ( V ) \mathbb{P}(V) P ( V ) is the ind-scheme represented by the sequence of standard coordinate closed immersions of projective spaces of increasing dimension Spec k = P k 0 ⟶ P k 1 ⟶ P k 2 ⟶ ⋯ \operatorname{Spec}k=\mathbb{P}_k^0\longrightarrow\mathbb{P}_k^1\longrightarrow\mathbb{P}_k^2\longrightarrow\cdots Spec k = P k 0 ⟶ P k 1 ⟶ P k 2 ⟶ ⋯ . We refer to [53 Example 1.3(2)] or [33 Section 11.2] for a more detailed and formal definition of the ind-scheme P ( V ) \mathbb{P}(V) P ( V ) .
Now the n n n -dimensional projective space P k n \mathbb{P}_k^n P k n over k k k cannot be covered by less than n + 1 n+1 n + 1 open affines. Indeed, for any quasi-compact semi-separated scheme Z Z Z covered by n n n open affines and every quasi-coherent sheaf M \mathcal{M} M on Z Z Z one has H i ( Z , M ) = 0 H^{i}(Z,\mathcal{M}) = 0 H i ( Z , M ) = 0 for all i ≥ n i \ge n i ≥ n , since the sheaf cohomology H ∗ ( Z , M ) H^{*}(Z,\mathcal{M}) H ∗ ( Z , M ) can be computed by the finite Čech complex [21 Lemmas Tags 01XD and 01FM] , which vanishes in the cohomological degrees ≥ n \ge n ≥ n . On the other hand, one has H n ( P k n , O ( − n − 1 ) ) = k ≠ 0 H^{n}(\mathbb{P}^{n}_{k},\mathcal{O}(-n-1)) = k \ne0 H n ( P k n , O ( − n − 1 )) = k = 0 [21 Lemma Tag 01XT] . For every integer n ≥ 0 n \ge0 n ≥ 0 , the scheme P k n \mathbb{P}^{n}_{k} P k n is a closed subscheme of P ( V ) \mathbb{P}(\mathcal{V}) P ( V ) . Hence the ind-scheme P ( V ) \mathbb{P}(\mathcal{V}) P ( V ) is not ind-affine-finite.
Reasonable ind-schemes
Let X X X be a concentrated scheme. A quasi-coherent sheaf M \mathcal{M} M on X X X is said to be (locally) finitely generated if, for every affine open subscheme U ⊂ X U \subset X U ⊂ X , the O X ( U ) \mathcal{O}_{X}(U) O X ( U ) -module M ( U ) \mathcal{M}(U) M ( U ) is finitely generated. It suffices to check this condition for affine open subschemes U ⊂ X U \subset X U ⊂ X belonging to any chosen affine open covering of X X X (see, e. g., [46 Lemma 3.1(a)] ).
Simply put, a quasi-coherent sheaf M \mathcal{M} M on X X X is finitely generated if and only if, as a sheaf of O X \mathcal{O}_{X} O X -modules, it is generated by a finite set of local sections.
Let i : Z ⟶ X i: Z \longrightarrow X i : Z ⟶ X be a closed immersion of concentrated schemes, and let I ⊂ O X \mathcal{I} \subset\mathcal{O}_{X} I ⊂ O X be the corresponding quasi-coherent sheaf of ideals on X X X . Following the terminology of [5 Sections 7.11.1 and 7.11.4] and [33 Section 2.1] , we say that the closed immersion morphism i i i is reasonable if the quasi-coherent sheaf I \mathcal{I} I on X X X is finitely generated. If this is the case, Z Z Z is also said to be a reasonable closed subscheme in X X X .
Let X \mathfrak{X} X be a (strict ind-concentrated) ind-scheme. A closed subscheme Z ⊂ X Z \subset\mathfrak{X} Z ⊂ X is said to be reasonable if, for every closed subscheme Y ⊂ X Y \subset\mathfrak{X} Y ⊂ X such that Z ⊂ Y Z \subset Y Z ⊂ Y , the closed subscheme Z Z Z is reasonable in Y Y Y . Given a representation of X \mathfrak{X} X by a directed diagram of closed immersions of concentrated schemes ( X γ ) γ ∈ Γ (X_{\gamma})_{\gamma\in\Gamma} ( X γ ) γ ∈ Γ , a closed subscheme Z ⊂ X Z \subset\mathfrak{X} Z ⊂ X is reasonable if and only if, for every index γ ∈ Γ \gamma\in\Gamma γ ∈ Γ such that Z ⊂ X γ Z \subset X_{\gamma} Z ⊂ X γ , the closed subscheme Z Z Z is reasonable in X γ X_{\gamma} X γ [33 Section 2.1] .
An ind-scheme X \mathfrak{X} X is said to be reasonable if it is a filtered inductive limit of its reasonable closed subschemes. Equivalently, X \mathfrak{X} X is reasonable if and only if it can be represented by a directed diagram ( X γ ) γ ∈ Γ (X_{\gamma})_{\gamma\in\Gamma} ( X γ ) γ ∈ Γ of reasonable closed immersions of concentrated schemes X γ X_{\gamma} X γ .
Example 2.18. On a Noetherian scheme X X X , any quasi-coherent sheaf of ideals is locally finitely generated (and in fact, coherent). Therefore, any strict ind-Noetherian ind-scheme is reasonable. Moreover, any closed subscheme of an ind-Noetherian ind-scheme is reasonable [33 Section 4.2] .
Example 2.19. Let Y ⟶ X \mathfrak{Y} \longrightarrow\mathfrak{X} Y ⟶ X be a morphism of ind-schemes that is “representable by (concentrated) schemes”. Assume that the ind-scheme X \mathfrak{X} X is reasonable. Then the ind-scheme Y \mathfrak{Y} Y is reasonable as well [33 Section 2.1] .
Flat and Very Flat Pro-Quasi-Coherent Pro-Sheaves
Following the conventions of the end of Section 2.1 and Remark 2.9, for the rest of this paper an ind-scheme means an object of Ind ( C S c h ) \operatorname{Ind}(\mathrm{CSch}) Ind ( CSch ) represented by a directed diagram of closed immersions of concentrated schemes.
According to the discussion in [5 Sections 7.11.3–4] , there are two kinds of quasi-coherent sheaf-type objects over an ind-scheme. In the terminology of [33 Chapters 2–3] , they are called the quasi-coherent torsion sheaves and the pro-quasi-coherent pro-sheaves . In this paper, we only deal with pro-quasi-coherent pro-sheaves.
Flat pro-quasi-coherent pro-sheaves
Following the discussion in [5 Section 7.11.3] and [33 Examples 3.1 and 3.8(1)] , the category of arbitrary pro-quasi-coherent pro-sheaves on an ind-scheme is usually not abelian and generally not well-behaved homologically. The category of flat pro-quasi-coherent pro-sheaves is a well-behaved exact category (in the sense of Quillen [7 ] ). For this reason, we restrict ourselves to flat pro-quasi-coherent pro-sheaves in this paper. We suggest the survey paper [7 ] as the reference source on exact categories.
Given a scheme X X X , we denote by X X X –Qcoh the Grothendieck abelian category of quasi-coherent sheaves on X X X . A quasi-coherent sheaf F \mathcal{F} F on X X X is said to be flat if, for every affine open subscheme U ⊂ X U \subset X U ⊂ X , the O X ( U ) \mathcal{O}_X(U) O X ( U ) -module F ( U ) \mathcal{F}(U) F ( U ) is flat. It suffices to check this condition for affine open subschemes U ⊂ X U \subset X U ⊂ X belonging to any chosen affine open covering of X X X . The notation X X X –Qcohf l ⊂ X _{\mathrm{fl}} \subset X fl ⊂ X –Qcoh stands for the full subcategory of flat quasi-coherent sheaves in X X X –Qcoh. The full subcategory X X X –Qcohf l _{\mathrm{fl}} fl is closed under extensions, kernels of epimorphisms, infinite coproducts, and directed colimits in X X X –Qcoh; so X X X –Qcohf l _{\mathrm{fl}} fl inherits an exact category structure from the abelian exact structure of X X X –Qcoh.
For every morphism of concentrated schemes f : Y ⟶ X f : Y \longrightarrow X f : Y ⟶ X , we have a pair of adjoint functors of direct image f ∗ : Y f_* : Y f ∗ : Y –Qcoh ⟶ X \longrightarrow X ⟶ X –Qcoh and inverse image f ∗ : X f^* : X f ∗ : X –Qcoh ⟶ Y \longrightarrow Y ⟶ Y –Qcoh. The inverse image functor f ∗ f^* f ∗ is left adjoint to the direct image functor f ∗ f_* f ∗ . The inverse image functor f ∗ f^* f ∗ always takes flat quasi-coherent sheaves on X X X to flat quasi-coherent sheaves on Y Y Y . The resulting functor f ∗ : X f^* : X f ∗ : X –Qcohf l ⟶ Y _{\mathrm{fl}} \longrightarrow Y fl ⟶ Y –Qcohf l _{\mathrm{fl}} fl is exact (as a functor of exact categories).
Let X \mathfrak{X} X be a (strict ind-concentrated) ind-scheme. A flat pro-quasi-coherent pro-sheaf F \mathcal{F} F on X \mathfrak{X} X is the following set of data:
to every closed subscheme Z ⊂ X Z \subset\mathfrak{X} Z ⊂ X , a flat quasi-coherent sheaf F ( Z ) \mathcal{F}^{(Z)} F ( Z ) on Z Z Z is assigned;
to every pair of closed subschemes Z ⊂ Y Z \subset Y Z ⊂ Y in X \mathfrak{X} X with the closed immersion morphism i Z Y : Z ⟶ Y i_{ZY}: Z \longrightarrow Y i Z Y : Z ⟶ Y , an isomorphism of flat quasi-coherent sheaves i Z Y ∗ F ( Y ) ≃ F ( Z ) i^*_{ZY}\mathcal{F}^{(Y)} \simeq\mathcal{F}^{(Z)} i Z Y ∗ F ( Y ) ≃ F ( Z ) on Z Z Z is assigned.
The following compatibility axiom for triples of closed subschemes is imposed:
for every triple of closed subschemes W ⊂ Z ⊂ Y W \subset Z \subset Y W ⊂ Z ⊂ Y in X \mathfrak{X} X , the triangular diagram i W Y ∗ F ( Y ) ⟶ i W Z ∗ F ( Z ) ⟶ F ( W ) i^*_{WY}\mathcal{F}^{(Y)} \longrightarrow i^*_{WZ}\mathcal{F}^{(Z)} \longrightarrow\mathcal{F}^{(W)} i W Y ∗ F ( Y ) ⟶ i W Z ∗ F ( Z ) ⟶ F ( W ) of isomorphisms in W W W –Qcohf l _{\mathrm{fl}} fl is commutative.
Let X = “ lim ” γ ∈ Γ X γ \mathfrak{X} = “\lim”_{\gamma\in\Gamma} X_\gamma X = “ lim ” γ ∈ Γ X γ be a representation of X \mathfrak{X} X by a directed diagram of closed immersions of concentrated schemes. Then, in order to define a flat pro-quasi-coherent pro-sheaf F \mathcal{F} F on X \mathfrak{X} X , it suffices to specify flat pro-quasi-coherent pro-sheaves F ( γ ) = F ( X γ ) \mathcal{F}^{(\gamma)} = \mathcal{F}(X_\gamma) F ( γ ) = F ( X γ ) on the closed subschemes X γ ⊂ X X_\gamma\subset\mathfrak{X} X γ ⊂ X for all indices γ ∈ Γ \gamma\in\Gamma γ ∈ Γ , together with isomorphisms i γ δ ∗ F ( δ ) ≃ F ( γ ) i_{\gamma\delta}^{*}\mathcal{F}^{(\delta)} \simeq\mathcal{F}^{(\gamma)} i γ δ ∗ F ( δ ) ≃ F ( γ ) for all pair of indices γ ≤ δ ∈ Γ \gamma\leq\delta\in\Gamma γ ≤ δ ∈ Γ . Here the notation is i γ δ = i X γ X δ : X γ ⟶ X δ i_{\gamma\delta}=i_{X_{\gamma}X_{\delta}}: X_{\gamma} \longrightarrow X_{\delta} i γ δ = i X γ X δ : X γ ⟶ X δ . Assuming that the compatibility axiom (iii) is satisfied for all triples of closed subschemes W = X β W=X_{\beta} W = X β , Z = X γ Z=X_{\gamma} Z = X γ , and Y = X δ Y=X_{\delta} Y = X δ with β ≤ γ ≤ δ ∈ Γ \beta\leq\gamma\leq\delta\in\Gamma β ≤ γ ≤ δ ∈ Γ , such a datum of quasi-coherent sheaves F ( γ ) \mathcal{F}^{(\gamma)} F ( γ ) and related isomorphisms i γ δ ∗ F ( δ ) ≃ F ( γ ) i_{\gamma\delta}^{*}\mathcal{F}^{(\delta)} \simeq\mathcal{F}^{(\gamma)} i γ δ ∗ F ( δ ) ≃ F ( γ ) can be extended to a flat pro-quasi-coherent pro-sheaf F \mathcal{F} F on X \mathfrak{X} X in a unique way. The flat quasi-coherent sheaves F ( Z ) \mathcal{F}(Z) F ( Z ) for all other closed subschemes Z ⊂ X Z \subset\mathfrak{X} Z ⊂ X together with the related isomorphisms (ii) can be uniquely recovered in such a way that the compatibility axiom (iii) is satisfied for all W ⊂ Z ⊂ Y ⊂ X W \subset Z \subset Y \subset\mathfrak{X} W ⊂ Z ⊂ Y ⊂ X .
Morphisms of flat pro-quasi-coherent pro-sheaves on X \mathfrak{X} X are defined in the obvious way. We denote the additive category of flat pro-quasi-coherent pro-sheaves on an ind-scheme X \mathfrak{X} X by X \mathfrak{X} X –Proqcf l _{\mathrm{fl}} fl .
A short sequence of flat pro-quasi-coherent pro-sheaves 0 ⟶ F ⟶ G ⟶ H ⟶ 0 0 \longrightarrow\mathcal{F} \longrightarrow\mathcal{G} \longrightarrow\mathcal{H} \longrightarrow0 0 ⟶ F ⟶ G ⟶ H ⟶ 0 on X \mathfrak{X} X is said to be (admissible) exact if the short sequence of flat quasi-coherent sheaves 0 ⟶ F ( Z ) ⟶ G ( Z ) ⟶ H ( Z ) ⟶ 0 0 \longrightarrow\mathcal{F}(Z) \longrightarrow\mathcal{G}(Z) \longrightarrow\mathcal{H}(Z) \longrightarrow0 0 ⟶ F ( Z ) ⟶ G ( Z ) ⟶ H ( Z ) ⟶ 0 is exact in Z Z Z –Qcohf l _{\mathrm{fl}} fl for every closed subscheme Z ⊂ X Z \subset\mathfrak{X} Z ⊂ X . Given a directed diagram of closed immersions of concentrated schemes ( X γ ) γ ∈ Γ (X_{\gamma})_{\gamma\in\Gamma} ( X γ ) γ ∈ Γ representing the ind-scheme X \mathfrak{X} X , it suffices to check the latter exactness condition for the closed subschemes Z = X γ ⊂ X Z=X_{\gamma}\subset\mathfrak{X} Z = X γ ⊂ X , γ ∈ Γ \gamma\in\Gamma γ ∈ Γ .
The class of short exact sequences defined in the previous paragraph defines an exact category structure on the additive category X \mathfrak{X} X –Proqcf l _{\mathrm{fl}} fl of flat pro-quasi-coherent pro-sheaves on X \mathfrak{X} X [33 Proposition 3.5] . The fact that the inverse image functors i Z Y ∗ : Y i_{ZY}^{*}:Y i Z Y ∗ : Y –Qcohf l ⟶ X _{\mathrm{fl}}\longrightarrow X fl ⟶ X –Qcohf l _{\mathrm{fl}} fl are exact plays a crucial role here.
All set-indexed coproducts and directed colimits exist in the additive category X \mathfrak{X} X –Proqcf l _{\mathrm{fl}} fl . Given a infinite family ( F ξ ) ξ ∈ Ξ (\mathcal{F}_{\xi})_{\xi\in\Xi} ( F ξ ) ξ ∈ Ξ of flat pro-quasi-coherent pro-sheaves on X \mathfrak{X} X , indexed by some set Ξ \Xi Ξ , the coproduct G = ∐ ξ ∈ Ξ F ξ \mathcal{G}=\coprod_{\xi\in\Xi}\mathcal{F}_{\xi} G = ∐ ξ ∈ Ξ F ξ is constructed by the obvious rule G ( Z ) = ∐ ξ ∈ Ξ F ξ ( Z ) \mathcal{G}(Z)=\coprod_{\xi\in\Xi}\mathcal{F}_{\xi}^{(Z)} G ( Z ) = ∐ ξ ∈ Ξ F ξ ( Z ) for all closed subschemes Z ⊂ X Z\subset\mathfrak{X} Z ⊂ X . Similarly, given a directed diagram ( F υ ) υ ∈ Υ (\mathcal{F}_{\upsilon})_{\upsilon\in\Upsilon} ( F υ ) υ ∈ Υ of flat pro-quasi-coherent pro-sheaves on X \mathfrak{X} X , indexed by some directed poset Υ \Upsilon Υ , the directed colimit G = lim ⟶ υ ∈ Υ F υ \mathcal{G}=\underset{\upsilon\in\Upsilon}{\lim\longrightarrow}\mathcal{F}_{\upsilon} G = υ ∈ Υ lim ⟶ F υ is constructed by the obvious rule G ( Z ) = lim ⟶ υ ∈ Υ F υ ( Z ) \mathcal{G}(Z)=\underset{\upsilon\in\Upsilon}{\lim\longrightarrow}\mathcal{F}_{\upsilon}^{(Z)} G ( Z ) = υ ∈ Υ lim ⟶ F υ ( Z ) for all closed subschemes Z ⊂ X Z\subset\mathfrak{X} Z ⊂ X [33 Section 3.5] .
In other words, the functors F ⟼ F ( Z ) : X \mathcal{F}\longmapsto\mathcal{F}(Z):\mathfrak{X} F ⟼ F ( Z ) : X –Proqcf l ⟶ Z _{\mathrm{fl}}\longrightarrow Z fl ⟶ Z –Qcohf l _{\mathrm{fl}} fl preserve coproducts and directed colimits for all closed subschemes Z ⊂ X Z\subset\mathfrak{X} Z ⊂ X . The functors of directed colimit are exact in the exact category X \mathfrak{X} X –Proqcf l _{\mathrm{fl}} fl ; in particular, this applies to the infinite coproduct functors.
Example 3.1. Let U = “ lim ⟶ γ ∈ Γ ” U γ \mathcal{U}=\text{“}\underset{\gamma\in\Gamma}{\lim\longrightarrow}\text{”}\,U_{\gamma} U = “ γ ∈ Γ lim ⟶ ” U γ be a strict ind-affine K 0 \mathcal{K}_{0} K 0 -ind-scheme represented by a countable directed diagram of closed immersions of affine schemes. Let R = lim ⟵ γ ∈ Γ O ( U γ ) \mathcal{R}=\underset{\gamma\in\Gamma}{\lim\longleftarrow}\mathcal{O}(U_{\gamma}) R = γ ∈ Γ lim ⟵ O ( U γ ) be the corresponding complete, separated topological commutative ring with a countable base of neighbourhoods of zero consisting of open ideals, as per Lemma 2.7. In this case, the exact category of flat pro-quasi-coherent pro-sheaves U \mathcal{U} U –Proqcf l _{\mathrm{fl}} fl is naturally embedded into an abelian category, which is called the category of R \mathcal{R} R -contramodules and denoted by R \mathcal{R} R –Contra.
We skip the definition and introductory discussion of contramodules over a topological ring, which can be found in the papers [42 Sections 1.1–1.2 and 5] , [44, Sections 6–7], [31 Section 2] , [35 Section 7] , or [39 Section 3] . For any complete, separated topological associative rings with a countable base of neighborhoods of zero consisting of open right ideals, the category of left R \mathcal{R} R -contramodules R \mathcal{R} R –Contra is a locally ℵ 1 \aleph_{1} ℵ 1 -presentable (in the sense of [1 Definition 1.17 and Theorem 1.20] ) abelian category with enough projective objects.
The relevant full exact subcategory in R \mathcal{R} R –Contra is the full subcategory of flat R \mathcal{R} R -contramodules R \mathcal{R} R –Contraf l ⊂ R _{\mathrm{fl}} \subset\mathcal{R} fl ⊂ R –Contra [29 Section E.1] , [42 Sections 6–7] , [31 Section 3] , [35 Section 8] . A relevant description of the category of flat contramodules over a topological associative ring with a countable base of two-sided ideals can be found in [35 Lemma 8.6(b)] (see also [35 Lemma 8.4] for background results providing the description of the exact category structure on R \mathcal{R} R –Contraf l _{\mathrm{fl}} fl ).
In the context of an ind-affine ℵ 0 \aleph_{0} ℵ 0 -ind-scheme U \mathcal{U} U and the corresponding topological ring R = O ( U ) \mathcal{R} = \mathcal{O}(\mathcal{U}) R = O ( U ) , U = Spf R \mathcal{U} = \operatorname{Spf}\mathcal{R} U = Spf R , there is a natural equivalence of exact categories
U – Proqc f l ≃ R – Contra f l . \mathcal{U}\text{–}\operatorname{Proqc}_{\mathrm{fl}} \simeq\mathcal{R}\text{–}\operatorname{Contra}_{\mathrm{fl}}. U – Proqc fl ≃ R – Contra fl .
To a flat pro-quasi-coherent pro-sheaf F \mathcal{F} F on U \mathcal{U} U , the flat R \mathcal{R} R -contramodule F = l i m ← γ ∈ Γ F ( U γ ) ( U γ ) \mathcal{F} = \varprojlim_{\gamma\in\Gamma} \mathcal{F}(U_{\gamma})(U_{\gamma}) F = lim γ ∈ Γ F ( U γ ) ( U γ ) is assigned. To a flat R \mathcal{R} R -contramodule F \mathcal{F} F , the flat pro-quasi-coherent pro-sheaf F \mathcal{F} F with the quasi-coherent sheaf F ( U γ ) \mathcal{F}(U_{\gamma}) F ( U γ ) on U γ U_{\gamma} U γ corresponding to the O ( U γ ) \mathcal{O}(U_{\gamma}) O ( U γ ) -module O ( U γ ) ⊙ R F = F / ( I γ ⋏ F ) \mathcal{O}(U_{\gamma}) \odot_{\mathcal{R}} \mathcal{F} = \mathcal{F}/(I_{\gamma} \curlywedge\mathcal{F}) O ( U γ ) ⊙ R F = F / ( I γ ⋏ F ) is assigned [33 Examples 3.8(1–2)] . Here I γ ⊂ R I_{\gamma} \subset\mathcal{R} I γ ⊂ R is the open kernel ideal of the natural surjective ring homomorphism R ⟶ O ( U γ ) \mathcal{R} \longrightarrow\mathcal{O}(U_{\gamma}) R ⟶ O ( U γ ) ; so O ( U γ ) ≃ R / I γ \mathcal{O}(U_{\gamma}) \simeq\mathcal{R}/I_{\gamma} O ( U γ ) ≃ R / I γ . See the references above for the explanations of the notation ⊙ R \odot_{\mathcal{R}} ⊙ R (for the contratensor product over R \mathcal{R} R ) and I ⋏ F ⊂ F I \curlywedge\mathcal{F} \subset\mathcal{F} I ⋏ F ⊂ F .
It is also worth noticing that the full subcategory of flat contramodules R \mathcal{R} R –Contraf l _{\mathrm{fl}} fl is closed under directed colimits (as well as extensions and kernels of epimorphisms) in the locally presentable abelian category R \mathcal{R} R –Contra. So the directed colimits in U \mathcal{U} U –Proqcf l _{\mathrm{fl}} fl agree with those in R \mathcal{R} R –Contraf l _{\mathrm{fl}} fl and in R \mathcal{R} R –Contra.
Inverse and direct images
Let f : Y ⟶ X f\colon\mathcal{Y} \longrightarrow\mathcal{X} f : Y ⟶ X be a morphism of ind-schemes. For any closed subscheme W ⊂ Y W \subset\mathcal{Y} W ⊂ Y , there exists a closed subscheme Z ⊂ X Z \subset\mathcal{X} Z ⊂ X such that the composition W ⟶ Y ⟶ X W \longrightarrow\mathcal{Y} \longrightarrow\mathcal{X} W ⟶ Y ⟶ X factorizes (always uniquely) as W ⟶ Z ⟶ X W \longrightarrow Z \longrightarrow\mathcal{X} W ⟶ Z ⟶ X . So we have a commutative square diagram of morphisms of schemes and ind-schemes
W → Y f W Z ↓ ↓ f Z → X (2) \begin{CD}
W @>>> \mathcal{Y} \\
@V{f_{WZ}}VV @VV{f}V \\
Z @>>> \mathcal{X} \tag*{(2)}
\end{CD} W f W Z ↓ ⏐ Z Y ↓ ⏐ f X (2)
with the horizontal arrows with tails showing closed immersion morphisms.
Let F \mathcal{F} F be a flat pro-quasi-coherent pro-sheaf on X \mathcal{X} X . Then, with the notation of the diagram (2), the rule ( f ∗ F ) ( W ) = f W Z ∗ ( F ( Z ) ) (f^{*}\mathcal{F})(W) = f_{WZ}^{*}(\mathcal{F}(Z)) ( f ∗ F ) ( W ) = f W Z ∗ ( F ( Z )) defines a flat pro-quasi-coherent pro-sheaf f ∗ F f^{*}\mathcal{F} f ∗ F on Y \mathcal{Y} Y . The resulting functor
f ∗ : X – Proqc f l ⟶ Y – Proqc f l (3) f^{*}\colon\mathcal{X}\text{–}\operatorname{Proqc}_{\mathrm{fl}} \longrightarrow\mathcal{Y}\text{–}\operatorname{Proqc}_{\mathrm{fl}} \tag*{(3)} f ∗ : X – Proqc fl ⟶ Y – Proqc fl (3)
is called the functor of inverse image of flat pro-quasi-coherent pro-sheaves. For any morphism of ind-schemes f f f , the inverse image functor f ∗ f^{*} f ∗ (3) is exact (as a functor of exact categories).
In particular, if i : Z ⟶ X i: Z \longrightarrow\mathcal{X} i : Z ⟶ X is a closed immersion of a concentrated scheme Z Z Z into an ind-scheme X \mathcal{X} X , then for any flat pro-quasi-coherent pro-sheaf F \mathcal{F} F on X \mathcal{X} X one has i ∗ F = F ( Z ) i^{*}\mathcal{F}=\mathcal{F}_{(Z)} i ∗ F = F ( Z ) .
A morphism of schemes f : Y ⟶ X f:Y\longrightarrow X f : Y ⟶ X is said to be flat if, for every pair of affine open subschemes V ⊂ Y V\subset Y V ⊂ Y and U ⊂ X U\subset X U ⊂ X such that f ( V ) ⊂ U f(V)\subset U f ( V ) ⊂ U , the related homomorphism of commutative rings O X ( U ) ⟶ O Y ( V ) \mathcal{O}_{X}(U)\longrightarrow\mathcal{O}_{Y}(V) O X ( U ) ⟶ O Y ( V ) makes O Y ( V ) \mathcal{O}_{Y}(V) O Y ( V ) a flat O X ( U ) \mathcal{O}_{X}(U) O X ( U ) -module. Given two affine open coverings Y = ⋃ α V α Y=\bigcup_{\alpha}V_{\alpha} Y = ⋃ α V α and X = ⋃ α U α X=\bigcup_{\alpha}U_{\alpha} X = ⋃ α U α of the schemes Y Y Y and X X X , indexed by the same set of indices { α } \{\alpha\} { α } and such that f ( V α ) ⊂ U α f(V_{\alpha})\subset U_{\alpha} f ( V α ) ⊂ U α for all α \alpha α , it suffices to check the flatness condition above for the ring homomorphisms O X ( U α ) ⟶ O Y ( V α ) \mathcal{O}_{X}(U_{\alpha})\longrightarrow\mathcal{O}_{Y}(V_{\alpha}) O X ( U α ) ⟶ O Y ( V α ) [21 Lemma Tag 01U5] . The class of all flat morphisms of schemes is closed under pullbacks [21 Lemma Tag 01U9] .
For any flat affine morphism of schemes f : Y ⟶ X f:Y\longrightarrow X f : Y ⟶ X , the direct image functor f ∗ f_{*} f ∗ takes flat quasi-coherent sheaves on Y Y Y to flat quasi-coherent sheaves on X X X . The resulting functor f ∗ : Y –Qcoh f l ⟶ X –Qcoh f l f_{*}:Y\text{--Qcoh}_{\mathrm{fl}}\longrightarrow X\text{--Qcoh}_{\mathrm{fl}} f ∗ : Y –Qcoh fl ⟶ X –Qcoh fl is exact (as a functor of exact categories).
In the context of Section 2.4, take P \mathrm{P} P to be the property of a morphism of schemes to be flat. This produces the notions of a flat and an ind-flat morphism of ind-schemes.
Let f : Y ⟶ X f:\mathcal{Y}\longrightarrow\mathcal{X} f : Y ⟶ X be a flat affine morphism of ind-schemes. Then, for every closed subscheme Z ⊂ X Z\subset\mathcal{X} Z ⊂ X , the ind-scheme W = Z × X Y W=Z\times_{\mathcal{X}}\mathcal{Y} W = Z × X Y is a concentrated scheme, and the induced morphism of schemes f Z : W ⟶ Z f_{Z}:W\longrightarrow Z f Z : W ⟶ Z is flat and affine (while the morphism W ⟶ Y W\longrightarrow\mathcal{Y} W ⟶ Y is a closed immersion). We have a pullback diagram of morphisms of schemes and ind-schemes
W → Y f Z ↓ ↓ f Z → X (4) \begin{CD}
W @>>> \mathcal{Y} \\
@V{f_Z}VV @VV{f}V \\
Z @>>> \mathcal{X} \tag*{(4)}
\end{CD} W f Z ↓ ⏐ Z Y ↓ ⏐ f X (4)
which is a special case of the diagram (2).
Let G \mathcal{G} G be a flat pro-quasi-coherent pro-sheaf on Y \mathcal{Y} Y . Then, with the notation of the pullback diagram (4), the rule ( f ∗ G ) ( Z ) = ( f Z ) ∗ ( G ( W ) ) (f_{*}\mathcal{G})_{(Z)}=(f_{Z})_{*}(\mathcal{G}_{(W)}) ( f ∗ G ) ( Z ) = ( f Z ) ∗ ( G ( W ) ) defines a flat pro-quasi-coherent pro-sheaf f ∗ G f_{*}\mathcal{G} f ∗ G on X \mathcal{X} X . One needs to use the affine base change lemma for quasi-coherent sheaves [21 Lemma Tag 02KG] , [33 Lemma 3.3(a)] in order to construct the isomorphisms (ii) from Section 3.1 above for the flat quasi-coherent sheaves ( f ∗ G ) ( Z ) (f_{*}\mathcal{G})_{(Z)} ( f ∗ G ) ( Z ) defined in this way.
The resulting functor
f ∗ : Y –Proqc f l ⟶ X –Proqc f l (5) f_{*}:\mathcal{Y}\text{--Proqc}_{\mathrm{fl}}\longrightarrow\mathcal{X}\text{--Proqc}_{\mathrm{fl}} \tag*{(5)} f ∗ : Y –Proqc fl ⟶ X –Proqc fl (5)
is called the functor of direct image of flat pro-quasi-coherent pro-sheaves. For any flat affine morphism of ind-schemes f f f , the direct image functor f ∗ f_{*} f ∗ (5) is exact (as a functor of exact categories).
For any flat affine morphism of ind-schemes f f f , the direct image functor f ∗ f_{*} f ∗ (5) is right adjoint to the inverse image functor f ∗ f^{*} f ∗ (3) [33 Sections 3.3 and 3.4] .
Very flat modules and ring homomorphisms
The concept of very flatness was introduced in the present author’s preprint on contraherent cosheaves [29 Sections 1.1, 1.10, and 4.1] . Further significant developments in this context are the papers [49 ] , [43 ] , and [16 ] . For discussions of the complementary concept of contraadjustedness, see [29 Sections 1.1, 2.5, and 4.1] and the papers [49 , 30 , 34 ] . Another reference source is the preprint [38 ] .
All rings and modules in this paper are associative and unital. Given an associative ring R R R , we denote by R R R –Mod the abelian category of left R R R -modules. The full subcategory of flat left R R R -modules is denoted by R R R –Modf l ⊂ R _{\mathrm{fl}} \subset R fl ⊂ R –Mod.
A left R R R -module C C C is called cotorsion (in the sense of Enochs [12 ] ) if Ext R 1 ( F , C ) = 0 \operatorname{Ext}^{1}_{R}(F,C)=0 Ext R 1 ( F , C ) = 0 for all flat left R R R -modules C C C , or equivalently, Ext R n ( F , C ) = 0 \operatorname{Ext}^{n}_{R}(F,C)=0 Ext R n ( F , C ) = 0 for all flat left R R R -modules C C C and all n ≥ 1 n \ge1 n ≥ 1 . The class of cotorsion R R R -modules is closed under extensions, cokernels of injective morphisms, and infinite direct products in R R R –Mod. We denote the full subcategory of cotorsion left R R R -modules by R R R –Modc o t ⊂ R ^{\mathrm{cot}} \subset R cot ⊂ R –Mod. The full subcategories R R R –Modf l _{\mathrm{fl}} fl and R R R –Modc o t ^{\mathrm{cot}} cot inherit exact category structures from the abelian exact structure of R R R –Mod.
Let R R R be a commutative ring. Given an element s ∈ R s \in R s ∈ R , we denote by R [ s − 1 ] R[s^{-1}] R [ s − 1 ] the ring obtained from R R R by formally inverting the element s s s . In other words, R [ s − 1 ] R[s^{-1}] R [ s − 1 ] is the ring of regular functions on the principal affine open subscheme in Spec R \operatorname{Spec} R Spec R corresponding to s s s , or the localization of R R R at the multiplicative subset { 1 , s , s 2 , s 3 , … } \{1,s,s^{2},s^{3},\ldots\} { 1 , s , s 2 , s 3 , … } . The R R R -module R [ s − 1 ] R[s^{-1}] R [ s − 1 ] is flat of projective dimension at most 1 [30 proof of Lemma 2.1] , [38 Remark 5.1] .
An R R R -module C C C is said to be contraadjusted if Ext R 1 ( R [ s − 1 ] , C ) = 0 \operatorname{Ext}^{1}_{R}(R[s^{-1}],C)=0 Ext R 1 ( R [ s − 1 ] , C ) = 0 for all s ∈ R s \in R s ∈ R . The class of contraadjusted R R R -modules is closed under extensions and infinite direct products in R R R –Mod. Any quotient R R R -module of a contraadjusted R R R -module is contraadjusted. All cotorsion R R R -modules are contraadjusted. We denote the full subcategory of contraadjusted R R R -modules by R R R –Modc t a ⊂ R ^{\mathrm{cta}} \subset R cta ⊂ R –Mod.
An R R R -module F F F is said to be very flat if Ext R 1 ( F , C ) = 0 \operatorname{Ext}^{1}_{R}(F,C)=0 Ext R 1 ( F , C ) = 0 for all contraadjusted R R R -modules C C C . The class of very flat R R R -modules is closed under extensions, kernels of surjective morphisms, and infinite direct sums in R R R –Mod. All very flat R R R -modules have projective dimensions at most 1 [29 Section 1.1] . All very flat R R R -modules are flat. We denote the full subcategory of very flat R R R -modules by R R R –Modv f l ⊂ R _{\mathrm{vfl}} \subset R vfl ⊂ R –Modf l ⊂ R _{\mathrm{fl}} \subset R fl ⊂ R –Mod. The full subcategories R R R –Modc t a ^{\mathrm{cta}} cta and R R R –Modv f l _{\mathrm{vfl}} vfl inherit exact category structures from the abelian exact structure of R R R –Mod.
The following characterization of very flat R R R -modules [29 Corollary 1.1.4] , based on the results of the paper [11 ] , may be more intuitive than the abstract homological algebra definition above. An R R R -module G G G is very flat if and only if it is a direct summand of an R R R -module F F F admitting an increasing filtration ( F i ⊂ F ) i ∈ α (F_i \subset F)_{i \in\alpha} ( F i ⊂ F ) i ∈ α indexed by some ordinal α \alpha α , with the successive quotient R R R -modules F i + 1 / F i F_{i+1}/F_i F i + 1 / F i of the form F i + 1 / F i ≅ R [ s i − 1 ] F_{i+1}/F_i \cong R[s_i^{-1}] F i + 1 / F i ≅ R [ s i − 1 ] for some elements s i ∈ R s_i \in R s i ∈ R , 0 ≤ i < α 0 \le i < \alpha 0 ≤ i < α . Here the notation presumes that the filtration is given by a smooth chain of R R R -submodules with F 0 = 0 F_0=0 F 0 = 0 , F i ⊂ F j F_i \subset F_j F i ⊂ F j for all 0 ≤ i ≤ j ≤ α 0 \le i \le j \le\alpha 0 ≤ i ≤ j ≤ α , F j = ⋃ i < j F i F_j=\bigcup_{i<j}F_i F j = ⋃ i < j F i for all limit ordinals j ≤ α j \le\alpha j ≤ α , and F α = F F_\alpha=F F α = F .
For any affine scheme U U U with an affine open subscheme V ⊂ U V \subset U V ⊂ U , the ring O ( V ) \mathcal{O}(V) O ( V ) is very flat as an O ( U ) \mathcal{O}(U) O ( U ) -module [29 Lemma 1.2.4] . Much more generally, for any homomorphism of commutative rings R ⟶ S R \longrightarrow S R ⟶ S such that S S S is a finitely presented R R R -algebra and a flat R R R -module, the R R R -module S S S is actually very flat [43 Main Theorem 1.1] .
A homomorphism of commutative rings R ⟶ S R \longrightarrow S R ⟶ S is said to be very flat if, for every element s ∈ S s \in S s ∈ S , the ring S [ s − 1 ] S[s^{-1}] S [ s − 1 ] is very flat as an R R R -module [43 Section 9] , [38 Section 6.1] . If this is the case, the ring S S S is also said to be a very flat commutative R R R -algebra . This condition first appears in [29 Lemma 1.2.3] . Requiring very flatness of the R R R -module S S S itself is not sufficient here [43 Example 9.7] , [38 Section 6.1] .
Very flat sheaves and morphisms of schemes
A quasi-coherent sheaf F \mathcal{F} F on a scheme X X X is said to be very flat if, for every affine open subscheme U ⊂ X U \subset X U ⊂ X , the O X ( U ) \mathcal{O}_X(U) O X ( U ) -module F ( U ) \mathcal{F}(U) F ( U ) is very flat. It suffices to check this condition for affine open subschemes U U U belonging to any chosen affine open covering of the scheme X X X [29 Lemma 1.2.6(a) and Section 1.10] , [34 Example 2.5] .
We denote the full subcategory of very flat quasi-coherent sheaves by X X X –Q c o h v f l ⊂ X \mathrm{Qcoh}_{\mathrm{vfl}} \subset X Qcoh vfl ⊂ X –Q c o h f l ⊂ X \mathrm{Qcoh}_{\mathrm{fl}} \subset X Qcoh fl ⊂ X –Q c o h \mathrm{Qcoh} Qcoh . The full subcategory X X X –Q c o h v f l \mathrm{Qcoh}_{\mathrm{vfl}} Qcoh vfl is closed under extensions, kernels of epimorphisms, and infinite coproducts in the abelian category X X X –Q c o h \mathrm{Qcoh} Qcoh (and in its full subcategory X X X –Q c o h f l \mathrm{Qcoh}_{\mathrm{fl}} Qcoh fl ); so X X X –Q c o h f l \mathrm{Qcoh}_{\mathrm{fl}} Qcoh fl inherits an exact category structure from the abelian exact structure of X X X –Q c o h \mathrm{Qcoh} Qcoh .
For any morphism of schemes f : Y ⟶ X f : Y \longrightarrow X f : Y ⟶ X , the inverse image functor f ∗ : X f^* : X f ∗ : X –Q c o h ⟶ Y \mathrm{Qcoh} \longrightarrow Y Qcoh ⟶ Y –Q c o h \mathrm{Qcoh} Qcoh takes very flat quasi-coherent sheaves on X X X to very flat quasi-coherent sheaves on Y Y Y [29 Lemma 1.2.2(b) and Section 1.10] .
A morphism of schemes f : Y ⟶ X f : Y \longrightarrow X f : Y ⟶ X is said to be very flat if, for every pair of affine open subschemes V ⊂ Y V \subset Y V ⊂ Y and U ⊂ X U \subset X U ⊂ X such that f ( V ) ⊂ U f(V) \subset U f ( V ) ⊂ U , the O X ( U ) \mathcal{O}_X(U) O X ( U ) -algebra O Y ( V ) \mathcal{O}_Y(V) O Y ( V ) is very flat. Requiring this for all pairs of affine open subschemes V ⊂ Y V \subset Y V ⊂ Y and U ⊂ X U \subset X U ⊂ X as above is equivalent to requiring the O X ( U ) \mathcal{O}_X(U) O X ( U ) -module O Y ( V ) \mathcal{O}_Y(V) O Y ( V ) to be very flat for all such pairs of affine open subschemes.
Let X = ⋃ α U α X = \bigcup_{\alpha} U_{\alpha} X = ⋃ α U α be an open covering of X X X . Then a morphism of schemes f : Y ⟶ X f : Y \longrightarrow X f : Y ⟶ X is very flat if and only if, for every index α \alpha α and every affine open subscheme V ⊂ Y V \subset Y V ⊂ Y such that f ( V ) ⊂ U α f(V) \subset U_{\alpha} f ( V ) ⊂ U α , the O X ( U α ) \mathcal{O}_X(U_{\alpha}) O X ( U α ) -module O Y ( V ) \mathcal{O}_Y(V) O Y ( V ) is very flat [29 Corollary 1.2.5(b), Lemma 1.2.6(a), and Section 1.10] .
Let X = ⋃ α U α X = \bigcup_{\alpha} U_{\alpha} X = ⋃ α U α and Y = ⋃ α V α Y = \bigcup_{\alpha} V_{\alpha} Y = ⋃ α V α be affine open coverings of the schemes X X X and Y Y Y , indexed by the same set of indices α \alpha α , such that one has f ( V α ) ⊂ U α f(V_{\alpha}) \subset U_{\alpha} f ( V α ) ⊂ U α for every α \alpha α . Then the morphism of schemes f f f is very flat if and only if, for every index α \alpha α , the O X ( U α ) \mathcal{O}_X(U_{\alpha}) O X ( U α ) -algebra O Y ( V α ) \mathcal{O}_Y(V_{\alpha}) O Y ( V α ) is very flat [29 Lemma 1.8.6 and Section 1.10] .
For any very flat affine morphism of schemes f : Y ⟶ X f : Y \longrightarrow X f : Y ⟶ X , the direct image functor f ∗ : Y f_* : Y f ∗ : Y –Q c o h ⟶ X \mathrm{Qcoh} \longrightarrow X Qcoh ⟶ X –Q c o h \mathrm{Qcoh} Qcoh takes very flat quasi-coherent sheaves on Y Y Y to very flat quasi-coherent sheaves on X X X [29 Lemma 1.2.3(b) and Section 1.10] .
Very flat pro-quasi-coherent pro-sheaves
Let X \mathfrak{X} X be an ind-scheme. A flat pro-quasi-coherent pro-sheaf F \mathcal{F} F on X \mathfrak{X} X is said to be very flat if, for every closed subscheme Z ⊂ X Z \subset\mathfrak{X} Z ⊂ X , the flat quasi-coherent sheaf F ( Z ) \mathcal{F}(Z) F ( Z ) on Z Z Z is very flat.
Given a representation of X \mathfrak{X} X by a directed diagram of closed immersions of concentrated schemes ( X γ ) γ ∈ Γ (X_{\gamma})_{\gamma\in\Gamma} ( X γ ) γ ∈ Γ , it suffices to check the very flatness condition above for the closed subschemes Z = X γ ⊂ X Z = X_{\gamma} \subset\mathfrak{X} Z = X γ ⊂ X , γ ∈ Γ \gamma\in\Gamma γ ∈ Γ . This follows from the stability of the class of very flat quasi-coherent sheaves under inverse images with respect to morphisms of schemes.
We denote the full subcategory of very flat pro-quasi-coherent pro-sheaves by X \mathscr{X} X –Proqcv f l ⊂ X _{\mathrm{vfl}} \subset\mathscr{X} vfl ⊂ X –Proqcf l _{\mathrm{fl}} fl . The full subcategory of very flat pro-quasi-coherent pro-sheaves X \mathscr{X} X –Proqcv f l _{\mathrm{vfl}} vfl is closed under extensions, kernels of admissible epimorphisms, and infinite coproducts in the exact category of flat pro-quasi-coherent pro-sheaves X \mathscr{X} X –Proqcf l _{\mathrm{fl}} fl ; so the full subcategory X \mathscr{X} X –Proqcv f l _{\mathrm{vfl}} vfl inherits an exact category structure from X \mathscr{X} X –Proqcf l _{\mathrm{fl}} fl .
For any morphism of ind-schemes f : Y ⟶ X f\colon\mathscr{Y} \longrightarrow\mathscr{X} f : Y ⟶ X , the inverse image functor f ∗ f^* f ∗ (3) takes very flat pro-quasi-coherent pro-sheaves on X \mathscr{X} X to very flat pro-quasi-coherent pro-sheaves on Y \mathscr{Y} Y . So we have an exact inverse image functor
f ∗ : X –Proqc v f l ⟶ Y –Proqc v f l . (6) f^*\colon\mathscr{X}\text{--Proqc}_{\mathrm{vfl}} \longrightarrow\mathscr{Y}\text{--Proqc}_{\mathrm{vfl}}. \tag*{(6)} f ∗ : X –Proqc vfl ⟶ Y –Proqc vfl . (6)
Our next aim is to define the notion of a very flat morphism of ind-schemes. The property of a morphism of schemes to be very flat is not known to be stable under pullbacks; so the theory of Section 2.4 is not applicable. However, the very flatness of morphisms of schemes is preserved by the pullbacks with respect to locally closed immersions of schemes [29 Lemma 2.10.1] .
Let f : Y ⟶ X f\colon\mathscr{Y} \longrightarrow\mathscr{X} f : Y ⟶ X be a morphism of ind-schemes that is “representable by (concentrated) schemes”. So, for every closed subscheme Z ⊂ X Z \subset\mathscr{X} Z ⊂ X , the ind-scheme W = Z × X Y W=Z\times_{\mathscr{X}}\mathscr{Y} W = Z × X Y is a concentrated scheme (and a closed subscheme in Y \mathscr{Y} Y ). We will say that the morphism of ind-schemes f f f is very flat if, for every closed subscheme Z ⊂ X Z\subset\mathscr{X} Z ⊂ X , the morphism of schemes W = Z × X Y ⟶ Z W=Z\times_{\mathscr{X}}\mathscr{Y}\longrightarrow Z W = Z × X Y ⟶ Z is very flat.
Once again, given a representation of X \mathscr{X} X by a directed diagram of closed immersions of concentrated schemes ( X γ ) γ ∈ Γ (X_\gamma)_{\gamma\in\Gamma} ( X γ ) γ ∈ Γ , it suffices to check the very flatness condition above for the closed subschemes Z = X γ ⊂ X Z=X_\gamma\subset\mathscr{X} Z = X γ ⊂ X , γ ∈ Γ \gamma\in\Gamma γ ∈ Γ . This follows from the stability of the class of very flat morphisms of schemes under the pullbacks by closed immersions.
In particular, all open immersions of ind-schemes are very flat morphisms (since all open immersions of schemes are very flat morphisms).
For any very flat affine morphism of ind-schemes f : Y ⟶ X f\colon\mathscr{Y}\longrightarrow\mathscr{X} f : Y ⟶ X , the direct image functor f ∗ f_* f ∗ (5) takes very flat pro-quasi-coherent pro-sheaves on Y \mathscr{Y} Y to very flat pro-quasi-coherent pro-sheaves on X \mathscr{X} X . So we have an exact direct image functor
f ∗ : Y –Proqc v f l ⟶ X –Proqc v f l . (7) f_*\colon\mathscr{Y}\text{--Proqc}_{\mathrm{vfl}} \longrightarrow\mathscr{X}\text{--Proqc}_{\mathrm{vfl}}. \tag*{(7)} f ∗ : Y –Proqc vfl ⟶ X –Proqc vfl . (7)
Example 3.2. We continue the discussion of ind-affine ℵ 0 \aleph_0 ℵ 0 -ind-schemes started in Lemma 2.7 and Example 3.1, and keep the notation introduced there.
Let R \mathscr{R} R be a complete, separated topological commutative ring with a countable base of neighborhoods of zero consisting of open ideals. Then an R \mathscr{R} R -contramodule F \mathscr{F} F is said to be very flat if, for every open ideal I ⊂ R \mathscr{I}\subset\mathscr{R} I ⊂ R , the R / I \mathscr{R}/\mathscr{I} R / I -module R / I ⊙ R F ≃ F ( I \ F ) \mathscr{R}/\mathscr{I}\odot_{\mathscr{R}}\mathscr{F}\simeq\mathscr{F}(\mathscr{I}\backslash\mathscr{F}) R / I ⊙ R F ≃ F ( I \ F ) is very flat [29 Section E.4], [42, Example 7.12(3)] (cf. [41 Section 2.12] for a special case). It suffices to check this condition for open ideals I \mathscr{I} I belonging to any chosen base of neighborhoods of zero in R \mathscr{R} R consisting of open ideals.
All very flat contramodules are flat (since all very flat modules over commutative rings are flat). The full subcategory of very flat R \mathscr{R} R -contramodules R \mathscr{R} R –Contrav f l _{\mathrm{vfl}} vfl is closed under extensions and kernels of epimorphisms in R \mathscr{R} R –Contra (or, which is the same, closed under extensions and kernels of admissible epimorphisms in R \mathcal{R} R –Contrafl), so it inherits an exact category structure.
Let U = Spf R \mathcal{U} = \operatorname{Spf}\mathcal{R} U = Spf R be the strict ind-affine ℵ 0 \aleph_{0} ℵ 0 -ind-scheme corresponding to R \mathcal{R} R . Then the equivalence of exact categories from Example 3.1 restricts to an equivalence of exact subcategories
U – Proqcvfl ≃ R – Contravfl . \mathcal{U}\text{–}\operatorname{Proqcvfl} \simeq\mathcal{R}\text{–}\operatorname{Contravfl}. U – Proqcvfl ≃ R – Contravfl .
Accessibility and Pseudolimits
Background on accessible categories
Let κ \kappa κ be a regular cardinal. We refer to the book [1 Theorem 1.5, Definition 1.13, and Remark 1.21] and the paper [2 ] for a discussion of the definitions of, the difference between, and the essential equivalence of the κ \kappa κ -directed and κ \kappa κ -filtered colimits in categories. Let us only mention, to avoid any confusion, that a poset Γ \Gamma Γ is said to be κ \kappa κ -directed if any subposet of cardinality smaller than κ \kappa κ has an upper bound in Γ \Gamma Γ .
Let K \mathsf{K} K be a category with κ \kappa κ -directed colimits. An object K ∈ K K \in\mathsf{K} K ∈ K is said to be κ \kappa κ -presentable if the functor Mor K ( K , − ) : K ⟶ S e t s \operatorname{Mor}_{\mathsf{K}}(K,-): \mathsf{K} \longrightarrow\mathsf{Sets} Mor K ( K , − ) : K ⟶ Sets preserves κ \kappa κ -directed (or equivalently, κ \kappa κ -filtered) colimits.
The category K \mathsf{K} K is said to be κ \kappa κ -accessible if there exists a set (rather than a proper class) of objects S ⊂ K S \subset\mathsf{K} S ⊂ K such that all the objects of S S S are κ \kappa κ -presentable in K \mathsf{K} K and every object of K \mathsf{K} K is a κ \kappa κ -directed colimit of objects from S S S [1 Definition 2.1] . If this is the case, then the κ \kappa κ -presentable objects of K \mathsf{K} K are precisely all the retracts of the objects from S S S .
A category is said to be accessible if it is κ \kappa κ -accessible for some regular cardinal κ \kappa κ [1 Remark 2.2(1)] . The ℵ 0 \aleph_{0} ℵ 0 -accessible categories are called finitely accessible , and the ℵ 1 \aleph_{1} ℵ 1 -accessible categories can be similarly called countably accessible . The same terminology applies to objects: the ℵ 0 \aleph_{0} ℵ 0 -presentable objects are called finitely presentable , and the ℵ 1 \aleph_{1} ℵ 1 -presentable objects are likewise called countably presentable .
In a κ \kappa κ -accessible category K \mathsf{K} K , every object K ∈ K K \in\mathsf{K} K ∈ K has a presentability rank [1 Remark 2.2(3)] . This means that for every K ∈ K K \in\mathsf{K} K ∈ K there exists a regular cardinal ρ ≥ κ \rho\geq\kappa ρ ≥ κ such that the object K K K is ρ \rho ρ -presentable.
A κ \kappa κ -accessible category where all colimits exist is said to be locally κ \kappa κ -presentable [1 Definition 1.17 and Theorem 1.20] . A category is said to be locally presentable if it is locally κ \kappa κ -presentable for some regular cardinal κ \kappa κ . The locally ℵ 0 \aleph_{0} ℵ 0 -presentable categories are called locally finitely presentable [1 Definition 1.9 and Theorem 1.11] .
Background of pseudolimits
Accessibility of pseudolimits of accessible categories is a result from the book [24 Theorem 5.1.6] . More precise bounds on the accessibility ranks of the pseudolimits are provided by results of the unpublished old preprint [54 Theorem 3.8, Corollary 3.9, and Remark 3.11(II)] , rediscovered in the recent paper [36 ] . An introductory discussion can be found in [1 Section 2.H] .
In this section we do not discuss any weighted pseudolimits/bilimits generally, but rather concentrate on the concrete constructions of the products, inserters , and equifiers , as well as the pseudopullbacks and what we call the isomorphers (otherwise known as iso-inserters ) of categories.
Proposition 4.1. Let κ \kappa κ be a regular cardinal, Ξ \Xi Ξ be a set of cardinality smaller than κ \kappa κ , and ( K ξ ) ξ ∈ Ξ (K_{\xi})_{\xi\in\Xi} ( K ξ ) ξ ∈ Ξ be a Ξ \Xi Ξ -indexed family of κ \kappa κ -accessible categories. Then the Cartesian product category K = ∏ ξ ∈ Ξ K ξ \mathsf{K} = \prod_{\xi\in\Xi} K_{\xi} K = ∏ ξ ∈ Ξ K ξ is κ \kappa κ -accessible. The κ \kappa κ -presentable objects of K \mathsf{K} K are precisely all the families of objects ( S ξ ∈ K ξ ) ξ ∈ Ξ (S_{\xi} \in K_{\xi})_{\xi\in\Xi} ( S ξ ∈ K ξ ) ξ ∈ Ξ such that the object S ξ S_{\xi} S ξ is κ \kappa κ -presentable in K ξ K_{\xi} K ξ for every index ξ ∈ Ξ \xi\in\Xi ξ ∈ Ξ .
Proof. This is a corrected version of [1 Proposition 2.67] , as per [36 Proposition 2.1] . □
Let λ \lambda λ be an ordinal and K \mathsf{K} K be a category. A λ \lambda λ -indexed chain (of objects and morphisms) in K \mathsf{K} K is a functor λ ⟶ K \lambda\longrightarrow\mathsf{K} λ ⟶ K . Here we are using the observations that any ordinal can be viewed as a poset, and any poset Γ \Gamma Γ can be viewed as a category with a morphism γ ′ ⟶ γ ′ ′ \gamma^{\prime} \longrightarrow\gamma^{\prime\prime} γ ′ ⟶ γ ′′ for every pair of elements γ ′ ≤ γ ′ ′ ∈ Γ \gamma^{\prime} \leq\gamma^{\prime\prime} \in\Gamma γ ′ ≤ γ ′′ ∈ Γ ; so an ordinal λ \lambda λ can be viewed as a category. In particular, this applies to any cardinal λ \lambda λ .
Let K \mathsf{K} K and L \mathsf{L} L be two categories, and let F , G : K ⇉ L F, G \colon\mathsf{K} \rightrightarrows\mathsf{L} F , G : K ⇉ L be two functors from K \mathsf{K} K to L \mathsf{L} L . In this context, the inserter category D \mathsf{D} D is defined as the category of pairs ( K , θ ) (K,\theta) ( K , θ ) , where K ∈ K K \in\mathsf{K} K ∈ K is an object and θ : F ( K ) ⟶ G ( K ) \theta\colon F(K) \longrightarrow G(K) θ : F ( K ) ⟶ G ( K ) is a morphism in L \mathsf{L} L . Morphisms in D \mathsf{D} D are defined in the obvious way.
Proposition 4.2. Let κ \kappa κ be a regular cardinal and λ \lambda λ be an infinite cardinal smaller than κ \kappa κ . Let K \mathsf{K} K and L \mathsf{L} L be two κ \kappa κ -accessible categories where colimits of all λ \lambda λ -indexed chains exist. Let F , G : K ⇉ L F, G \colon\mathsf{K} \rightrightarrows\mathsf{L} F , G : K ⇉ L be a parallel pair of functors preserving κ \kappa κ -directed colimits and colimits of λ \lambda λ -indexed chains, and taking κ \kappa κ -presentable objects to κ \kappa κ -presentable objects. Then the inserter category D \mathsf{D} D is κ \kappa κ -accessible. An object ( K , θ ) ∈ D (K,\theta) \in\mathsf{D} ( K , θ ) ∈ D is κ \kappa κ -presentable if and only if the object K K K is κ \kappa κ -presentable in K \mathsf{K} K .
Proof. This is [54 Theorem 3.8, Corollary 3.9, and Remark 3.11(II)] or [36 Theorem 4.1] . □
Let D \mathsf{D} D and M \mathsf{M} M be two categories, let F , G : D ⇉ M F, G \colon\mathsf{D} \rightrightarrows\mathsf{M} F , G : D ⇉ M be two functors, and let ϕ , ψ : F ⇒ G \phi, \psi\colon F \Rightarrow G ϕ , ψ : F ⇒ G be two natural transformations. In this context, the equifier category E \mathsf{E} E is the full subcategory E ⊂ D \mathsf{E} \subset\mathsf{D} E ⊂ D consisting of all the objects D ∈ D D \in\mathsf{D} D ∈ D for which the two morphisms ϕ D \phi_D ϕ D and ψ D : F ( D ) ⇒ G ( D ) \psi_D \colon F(D) \Rightarrow G(D) ψ D : F ( D ) ⇒ G ( D ) are equal to each other.
Proposition 4.3. Let κ \kappa κ be a regular cardinal and λ \lambda λ be an infinite cardinal smaller than κ \kappa κ . Let D \mathsf{D} D and M \mathsf{M} M be two κ \kappa κ -accessible categories where colimits of all λ \lambda λ -indexed chains exist. Let F , G : D ⇉ M F, G \colon\mathsf{D} \rightrightarrows\mathsf{M} F , G : D ⇉ M be a parallel pair of functors preserving κ \kappa κ -directed colimits and colimits of λ \lambda λ -indexed chains, and taking κ \kappa κ -presentable objects to κ \kappa κ -presentable objects. Let ϕ , ψ : F ⇒ G \phi, \psi\colon F \Rightarrow G ϕ , ψ : F ⇒ G be two natural transformations. Then the equifier category E \mathsf{E} E is κ \kappa κ -accessible. An object of E \mathsf{E} E is κ \kappa κ -presentable if and only if it is κ \kappa κ -presentable in D \mathsf{D} D .
Proof. This is [54 Theorem 3.8, Corollary 3.9, and Remark 3.11(II)] or [36 Theorem 3.1] . □
Let K K K , L L L , and M M M be three categories, and let F : K ⟶ M F: K \longrightarrow M F : K ⟶ M and G : L ⟶ M G: L \longrightarrow M G : L ⟶ M be two functors. In this context, the pseudopullback category P P P is defined as the category of triples ( K , L , θ ) (K,L,\theta) ( K , L , θ ) , where K ∈ K K \in K K ∈ K and L ∈ L L \in L L ∈ L are two objects and θ : F ( K ) ≅ G ( L ) \theta:F(K)\cong G(L) θ : F ( K ) ≅ G ( L ) is an isomorphism in M M M . Morphisms in P P P are defined in the obvious way.
Corollary 4.4. Let κ \kappa κ be a regular cardinal and λ \lambda λ be an infinite cardinal smaller than κ \kappa κ . Let K K K , L L L , and M M M be three κ \kappa κ -accessible categories where colimits of all λ \lambda λ -indexed chains exist. Let F : K ⟶ M F: K \longrightarrow M F : K ⟶ M and G : L ⟶ M G: L \longrightarrow M G : L ⟶ M be two functors preserving κ \kappa κ -directed colimits and colimits of λ \lambda λ -indexed chains, and taking κ \kappa κ -presentable objects to κ \kappa κ -presentable objects. Then the pseudopullback category P P P is κ \kappa κ -accessible. An object ( K , L , θ ) ∈ P (K,L,\theta) \in P ( K , L , θ ) ∈ P is κ \kappa κ -presentable if and only if the object K K K is κ \kappa κ -presentable in K K K and the object L L L is κ \kappa κ -presentable in L L L .
Proof. This is [54 Remark 3.2(I), Theorem 3.8, Corollary 3.9, and Remark 3.11(II)] , or [52 Pseudopullback Theorem 2.2] (based on [8 Proposition 3.1] ), or [36 Corollary 5.1] . This corollary can be easily deduced from Propositions 4.1, 4.2, and 4.3. □
Let K K K and L L L be two categories, and let F , G : K ⇉ L F,G: K \rightrightarrows L F , G : K ⇉ L be two functors from K K K to L L L . In this context, the isomorpher (or iso-inserter ) category D D D is defined as the category of pairs ( K , θ ) (K,\theta) ( K , θ ) , where K ∈ K K \in K K ∈ K is an object and θ : F ( K ) ≅ G ( K ) \theta:F(K)\cong G(K) θ : F ( K ) ≅ G ( K ) is an isomorphism in L L L . Morphisms in D D D are defined in the obvious way.
Corollary 4.5. Let κ \kappa κ be a regular cardinal and λ \lambda λ be an infinite cardinal smaller than κ \kappa κ . Let K K K and L L L be two κ \kappa κ -accessible categories where colimits of all λ \lambda λ -indexed chains exist. Let F , G : K ⇉ L F,G: K \rightrightarrows L F , G : K ⇉ L be a parallel pair of functors preserving κ \kappa κ -directed colimits and colimits of λ \lambda λ -indexed chains, and taking κ \kappa κ -presentable objects to κ \kappa κ -presentable objects. Then the isomorpher category D D D is κ \kappa κ -accessible. An object ( K , θ ) ∈ D (K,\theta) \in D ( K , θ ) ∈ D is κ \kappa κ -presentable if and only if the object K K K is κ \kappa κ -presentable in K K K .
Proof. This is [54 Remark 3.2(I), Theorem 3.8, Corollary 3.9, and Remark 3.11(II)] or [36 Remark 5.2] . This corollary is essentially an equivalent restatement of the previous one. □
Given an additive category K K K , we denote by Com ( K ) \operatorname{Com}(K) Com ( K ) the additive category of (un-bounded) complexes in K K K . An additive category is said to be κ \kappa κ -accessible if it is κ \kappa κ -accessible as an abstract category.
Proposition 4.6. Let κ \kappa κ be a regular cardinal and λ \lambda λ be an infinite cardinal smaller than κ \kappa κ . Let K K K be a κ \kappa κ -accessible additive category where colimits of all λ \lambda λ -indexed chains exist. Then the additive category Com ( K ) \operatorname{Com}(K) Com ( K ) of complexes in K K K is κ \kappa κ -accessible. A complex K ∙ K^\bullet K ∙ in K K K is κ \kappa κ -presentable as an object of Com ( K ) \operatorname{Com}(K) Com ( K ) if and only if all its terms K n K^n K n , n ∈ Z n \in\mathbb{Z} n ∈ Z , are κ \kappa κ -presentable in K K K .
Proof. This is a special case of [36 Theorem 6.2] ; cf. [36 proof of Corollary 10.4] . □
Lemma 4.7. Let κ \kappa κ be a regular cardinal and K K K be a locally κ \kappa κ -presentable additive category. Then the additive category Com ( K ) \operatorname{Com}(K) Com ( K ) of complexes in K K K is locally κ \kappa κ -presentable. In the case of an uncountable regular cardinal κ ≥ ℵ 1 \kappa\geq\aleph_1 κ ≥ ℵ 1 , a complex K ∙ K^\bullet K ∙ in K K K is
κ \kappa κ -presentable as an object of Com ( K ) \operatorname{Com}(\mathbf{K}) Com ( K ) if and only if all its terms K n K^n K n , n ∈ Z n \in\mathbb{Z} n ∈ Z , are κ \kappa κ -presentable in K \mathbf{K} K . In the case of the countable cardinal κ = ℵ 0 \kappa= \aleph_0 κ = ℵ 0 , a complex K ∙ K^\bullet K ∙ in K \mathbf{K} K is finitely presentable as an object of Com ( K ) \operatorname{Com}(\mathbf{K}) Com ( K ) if and only if K ∙ K^\bullet K ∙ is bounded and all its terms K n K^n K n , n ∈ Z n \in\mathbb{Z} n ∈ Z , are finitely presentable in K \mathbf{K} K .
Proof. The case of κ ≥ ℵ 1 \kappa\geq\aleph_1 κ ≥ ℵ 1 is covered by Proposition 4.6 (take λ = ℵ 0 \lambda= \aleph_0 λ = ℵ 0 ). In the case of κ = ℵ 0 \kappa= \aleph_0 κ = ℵ 0 , the first assertion of the lemma is covered by [45 Lemma 6.3] , and the second assertion can be discerned from the same argument using [50 Lemma A.3(2)] . Another relevant reference is [20 Theorem 1.2] . □ \square □
Background on accessibility over schemes
Let R R R be an associative ring. Then the abelian category R – Mod R\text{--}\operatorname{Mod} R – Mod is locally finitely presentable, hence also locally κ \kappa κ -presentable for all regular cardinals κ \kappa κ . An R R R -module is κ \kappa κ -presentable if and only if it is (isomorphic to) the cokernel of a morphism of two free R R R -modules with fewer than κ \kappa κ generators.
Let X X X be a scheme. A quasi-coherent sheaf M \mathcal{M} M on X X X is said to be locally κ \kappa κ -presentable if, for every affine open subscheme U ⊂ X U \subset X U ⊂ X , the O X ( U ) \mathcal{O}_X(U) O X ( U ) -module M ( U ) \mathcal{M}(U) M ( U ) is κ \kappa κ -presentable. It suffices to check this condition for affine open subschemes U U U belonging to any chosen affine open covering of the scheme X X X [46 Lemma 3.1] .
For any morphism of schemes f : Y ⟶ X f:Y \longrightarrow X f : Y ⟶ X , the inverse image functor f ∗ : X – Qcoh ⟶ Y – Qcoh f^*:X\text{--}\operatorname{Qcoh}\longrightarrow Y\text{--}\operatorname{Qcoh} f ∗ : X – Qcoh ⟶ Y – Qcoh takes locally κ \kappa κ -presentable quasi-coherent sheaves on X X X to locally κ \kappa κ -presentable quasi-coherent sheaves on Y Y Y .
Let X X X be a concentrated scheme. Then the abelian category X – Qcoh X\text{--}\operatorname{Qcoh} X – Qcoh is locally finitely presentable [18 0.5.2.5 and Corollaire I.6.9.12] , [21 Definition Tag 01BN and Lemma Tag 01PJ or 01PK] , hence also locally κ \kappa κ -presentable for all regular cardinals κ \kappa κ . An quasi-coherent sheaf on X X X is κ \kappa κ -presentable if and only if it is locally κ \kappa κ -presentable in the sense of the previous paragraph (see also [46 Remark 4.6] ).
Proposition 4.8. Let X X X be a concentrated scheme and κ ≥ ℵ 1 \kappa\geq\aleph_1 κ ≥ ℵ 1 be an uncountable regular cardinal. Then the category X – Qcoh f l X\text{--}\operatorname{Qcoh}_{\mathrm{fl}} X – Qcoh fl of flat quasi-coherent sheaves on X X X is κ \kappa κ -accessible. A flat quasi-coherent sheaf on X X X is κ \kappa κ -presentable as an object of X – Qcoh f l X\text{--}\operatorname{Qcoh}_{\mathrm{fl}} X – Qcoh fl if and only if it is locally κ \kappa κ -presentable (or equivalently, κ \kappa κ -presentable as an object of X – Qcoh X\text{--}\operatorname{Qcoh} X – Qcoh ).
Proof. This is essentially [46 Theorem 3.5] (see also [46 Theorem 4.5] ). □ \square □
For an exact category E \mathbf{E} E , we denote by Ac ( E ) ⊂ Com ( E ) \operatorname{Ac}(\mathbf{E}) \subset\operatorname{Com}(\mathbf{E}) Ac ( E ) ⊂ Com ( E ) the full subcategory of acyclic complexes in E \mathbf{E} E . In particular, the notation Ac ( X – Qcoh f l ) \operatorname{Ac}(X\text{--}\operatorname{Qcoh}_{\mathrm{fl}}) Ac ( X – Qcoh fl ) stands for the category of all acyclic complexes in the exact category X – Qcoh f l X\text{--}\operatorname{Qcoh}_{\mathrm{fl}} X – Qcoh fl of flat quasi-coherent sheaves on X X X . Notice that a complex in X – Qcoh f l X\text{--}\operatorname{Qcoh}_{\mathrm{fl}} X – Qcoh fl that is acyclic in X – Qcoh X\text{--}\operatorname{Qcoh} X – Qcoh need not be acyclic in X – Qcoh f l X\text{--}\operatorname{Qcoh}_{\mathrm{fl}} X – Qcoh fl , that is Ac ( X – Qcoh f l ) ≠ Com ( X – Qcoh f l ) ∩ Ac ( \operatorname{Ac}(X\text{--}\operatorname{Qcoh}_{\mathrm{fl}}) \neq\operatorname{Com}(X\text{--}\operatorname{Qcoh}_{\mathrm{fl}}) \cap\operatorname{Ac}(\text{} Ac ( X – Qcoh fl ) = Com ( X – Qcoh fl ) ∩ Ac ( X− − Qcoh ) --\operatorname{Qcoh}) − − Qcoh ) . The complexes acyclic in X – Qcoh f l X\text{--}\operatorname{Qcoh}_{\mathrm{fl}} X – Qcoh fl are otherwise known as pure acyclic complexes of flat quasi-coherent sheaves.
Proposition 4.9. Let X X X be a concentrated scheme and κ ≥ ℵ 1 \kappa\geq\aleph_1 κ ≥ ℵ 1 be an uncountable regular cardinal. Then the category Com ( X – Qcoh f l ) \operatorname{Com}(X\text{--}\operatorname{Qcoh}_{\mathrm{fl}}) Com ( X – Qcoh fl ) of complexes of flat quasi-coherent sheaves on X X X is κ \kappa κ -accessible. A complex F ∙ \mathcal{F}^\bullet F ∙ of flat quasi-coherent sheaves on X X X is
κ \kappa κ -presentable as an object of Com(X X X –Qcohf l _{\mathrm{fl}} fl ) if and only if all its terms F n \mathcal{F}^{n} F n , n ∈ Z n \in\mathbb{Z} n ∈ Z , are locally κ \kappa κ -presentable quasi-coherent sheaves.
Proof. This is essentially [46 Theorem 5.1] . Alternatively, the assertion is provable by comparing Propositions 4.6 and 4.8. □
Proposition 4.10. Let X X X be a concentrated scheme and κ ≥ ℵ 1 \kappa\ge\aleph_{1} κ ≥ ℵ 1 be an uncountable regular cardinal. Then the category Ac(X X X –Qcohf l _{\mathrm{fl}} fl ) of pure acyclic complexes of flat quasi-coherent sheaves on X X X is κ \kappa κ -accessible. A pure acyclic complex F ∙ \mathcal{F}^{\bullet} F ∙ of flat quasi-coherent sheaves on X X X is κ \kappa κ -presentable as an object of Ac(X X X –Qcohf l _{\mathrm{fl}} fl ) if and only if all its terms F n \mathcal{F}^{n} F n , n ∈ Z n \in\mathbb{Z} n ∈ Z , are locally κ \kappa κ -presentable quasi-coherent sheaves.
Proof. This is essentially [46 Theorem 5.2] . □
Accessibility of flat pro-quasi-coherent pro-sheaves
Let μ \mu μ be an infinite cardinal. An (ind-concentrated) ind-scheme X \mathcal{X} X is said to be an μ \mu μ -ind-scheme if X \mathcal{X} X can be represented by a directed diagram of concentrated schemes ( X γ ) γ ∈ Γ (X_{\gamma})_{\gamma\in\Gamma} ( X γ ) γ ∈ Γ indexed by a poset Γ \Gamma Γ of cardinality smaller than or equal to μ \mu μ .
The obvious version of Lemma 2.4 is valid for directed (sub)posets of cardinality ≤ μ \le\mu ≤ μ in place of countable directed (sub)posets. The proof is the same, based on the fact that for any directed poset Λ \Lambda Λ and any subset Λ ′ ′ ⊂ Λ \Lambda^{\prime\prime} \subset\Lambda Λ ′′ ⊂ Λ of cardinality ≤ μ \le\mu ≤ μ there is a directed subposet Λ ′ ⊂ Λ \Lambda^{\prime} \subset\Lambda Λ ′ ⊂ Λ of cardinality ≤ μ \le\mu ≤ μ such that for every λ ′ ′ ∈ Λ ′ ′ \lambda^{\prime\prime} \in\Lambda^{\prime\prime} λ ′′ ∈ Λ ′′ there exists λ ′ ∈ Λ ′ \lambda^{\prime} \in\Lambda^{\prime} λ ′ ∈ Λ ′ with λ ′ ′ ≤ λ ′ \lambda^{\prime\prime} \le\lambda^{\prime} λ ′′ ≤ λ ′ in Λ \Lambda Λ . In this sense, the notion of a strict μ \mu μ -ind-scheme is well-behaved.
The following theorem is the main result of Section 4.
Theorem 4.11. Let μ \mu μ be an infinite cardinal, κ ≥ ℵ 1 \kappa\ge\aleph_{1} κ ≥ ℵ 1 , κ > μ \kappa> \mu κ > μ be a regular cardinal, and X \mathcal{X} X be an (ind-concentrated) μ \mu μ -ind-scheme. Then the category X \mathcal{X} X –Proqcf l _{\mathrm{fl}} fl of flat pro-quasi-coherent pro-sheaves on X \mathcal{X} X is κ \kappa κ -accessible. A flat pro-quasi-coherent pro-sheaf F \mathcal{F} F on X \mathcal{X} X is κ \kappa κ -presentable as an object of X \mathcal{X} X –Proqcf l _{\mathrm{fl}} fl if and only if, for every closed subscheme Z ⊂ X Z \subset\mathcal{X} Z ⊂ X , the quasi-coherent sheaf F ( Z ) \mathcal{F}^{(Z)} F ( Z ) on Z Z Z is locally κ \kappa κ -presentable (i. e., κ \kappa κ -presentable as an object of X \mathcal{X} X –Qcohf l _{\mathrm{fl}} fl , or equivalently, as an object of X \mathcal{X} X –Qcoh).
Proof. This is a far-reaching partial generalization of the discussion of accessibility of flat contramodules in [42 Corollary 7.5] and [35 Theorem 10.1] ; cf. Example 3.1 above. Let X = “ lim ⟶ γ ∈ Γ ” X γ \mathcal{X} = “\underset{\gamma\in\Gamma}{\lim\limits_{\longrightarrow}}” X_{\gamma} X = “ γ ∈ Γ ⟶ lim ” X γ be a representation of X \mathcal{X} X by a directed diagram of concentrated schemes, indexed by a directed poset Γ \Gamma Γ of cardinality at most μ \mu μ . Consider the Cartesian products of categories K = ∏ γ ∈ Γ X γ \mathrm{K} = \prod_{\gamma\in\Gamma} X_{\gamma} K = ∏ γ ∈ Γ X γ –Qcohf l _{\mathrm{fl}} fl and L = ∏ γ < δ ∈ Γ X γ \mathrm{L} = \prod_{\gamma< \delta\in\Gamma} X_{\gamma} L = ∏ γ < δ ∈ Γ X γ –Qcohf l _{\mathrm{fl}} fl . Here, in the construction of the category L \mathrm{L} L , the Cartesian product is taken over the set of all pairs of indices ( γ , δ ) ∈ Γ 2 (\gamma, \delta) \in\Gamma^{2} ( γ , δ ) ∈ Γ 2 such that γ < δ \gamma< \delta γ < δ . Propositions 4.1 and 4.8 tell us that the categories K \mathrm{K} K and L \mathrm{L} L are κ \kappa κ -accessible, and provide a description of the full subcategories of κ \kappa κ -presentable objects in K \mathrm{K} K and L \mathrm{L} L .
Consider the following pair of functors F , G : K ⇒ L F, G \colon\mathrm{K} \Rightarrow\mathrm{L} F , G : K ⇒ L . To a collection of flat quasi-coherent sheaves ( F γ ) γ ∈ Γ ∈ K (\mathcal{F}_{\gamma})_{\gamma\in\Gamma} \in\mathrm{K} ( F γ ) γ ∈ Γ ∈ K , the functor F F F assigns the collection of flat quasi-coherent sheaves ( H γ δ ) γ < δ ∈ Γ ∈ L (\mathcal{H}_{\gamma\delta})_{\gamma< \delta\in\Gamma} \in\mathrm{L} ( H γ δ ) γ < δ ∈ Γ ∈ L with H γ δ = F γ \mathcal{H}_{\gamma\delta} = \mathcal{F}_{\gamma} H γ δ = F γ for all γ < δ ∈ Γ \gamma< \delta\in\Gamma γ < δ ∈ Γ . To the same collection of flat quasi-coherent sheaves ( F γ ) γ ∈ Γ ∈ K (\mathcal{F}_{\gamma})_{\gamma\in\Gamma} \in\mathrm{K} ( F γ ) γ ∈ Γ ∈ K , the functor G G G assigns the collection of flat quasi-coherent sheaves ( G γ δ ) γ < δ ∈ Γ ∈ L (\mathcal{G}_{\gamma\delta})_{\gamma< \delta\in\Gamma} \in\mathrm{L} ( G γ δ ) γ < δ ∈ Γ ∈ L with G γ δ = i γ δ ∗ F δ \mathcal{G}_{\gamma\delta} = i_{\gamma\delta}^{*}\mathcal{F}_{\delta} G γ δ = i γ δ ∗ F δ for all γ < δ ∈ Γ \gamma< \delta\in\Gamma γ < δ ∈ Γ . Here i γ δ : X γ ⟶ X δ i_{\gamma\delta}: X_{\gamma} \longrightarrow X_{\delta} i γ δ : X γ ⟶ X δ is the notation for the transition morphism in the direct system of schemes ( X γ ) γ ∈ Γ (X_{\gamma})_{\gamma\in\Gamma} ( X γ ) γ ∈ Γ .
Notice that all directed colimits exist in the categories K \mathrm{K} K and L \mathrm{L} L , and all of them are preserved by the functors F F F and G G G . Furthermore, the functors F F F and G G G take κ \kappa κ -presentable objects of K \mathrm{K} K to κ \kappa κ -presentable objects of L \mathrm{L} L . Take λ \lambda λ to be the countable cardinal, λ = ℵ 0 \lambda= \aleph_{0} λ = ℵ 0 . Then the assumptions of Corollary 4.5 are satisfied for the pair of functors F , G : K ⇒ L F, G: \mathrm{K} \Rightarrow\mathrm{L} F , G : K ⇒ L . Thus the isomorpher category D \mathrm{D} D is κ \kappa κ -accessible, and we also obtain a description of κ \kappa κ -presentable objects in D \mathrm{D} D .
The isomorpher category D \mathrm{D} D is a rough version of the category of flat pro-quasi-coherent pro-sheaves X - Proqc f l \mathcal{X}\text{-}\operatorname{Proqc}_{\mathrm{fl}} X - Proqc fl as defined in § 3.1 , with an unimportant difference that only the closed subschemes Z = X γ ⊂ X Z = X_{\gamma} \subset\mathcal{X} Z = X γ ⊂ X , γ ∈ Γ \gamma\in\Gamma γ ∈ Γ , are involved, and an important difference that the triple compatibility condition (iii) is not imposed. In order to impose the triple compatibility condition (iii) from § 3.1 , we use the equifier construction.
Consider the Cartesian product of categories M = ∏ β < γ < δ ∈ Γ X β - Qcoh f l \mathrm{M} = \prod_{\beta< \gamma< \delta\in\Gamma} X_{\beta}\text{-}\operatorname{Qcoh}_{\mathrm{fl}} M = ∏ β < γ < δ ∈ Γ X β - Qcoh fl , taken over the set of all triples of indices ( β , γ , δ ) ∈ Γ 3 (\beta, \gamma, \delta) \in\Gamma^{3} ( β , γ , δ ) ∈ Γ 3 such that β < γ < δ \beta< \gamma< \delta β < γ < δ . Propositions 4.1 and 4.8 tell us that the category M \mathrm{M} M is κ \kappa κ -accessible, and provide a description of the full subcategory of κ \kappa κ -presentable objects in M \mathrm{M} M .
Consider the following pair of functors F , G : D ⇒ M F, G: \mathrm{D} \Rightarrow\mathrm{M} F , G : D ⇒ M . An object of D \mathrm{D} D is a collection of flat quasi-coherent sheaves ( F γ ) γ ∈ Γ ∈ K (\mathcal{F}_{\gamma})_{\gamma\in\Gamma} \in\mathrm{K} ( F γ ) γ ∈ Γ ∈ K together with a collection of isomorphisms θ γ δ : i γ δ ∗ F δ ≃ F γ \theta_{\gamma\delta}: i_{\gamma\delta}^{*}\mathcal{F}_{\delta} \simeq\mathcal{F}_{\gamma} θ γ δ : i γ δ ∗ F δ ≃ F γ defined for all pairs of indices γ < δ ∈ Γ \gamma< \delta\in\Gamma γ < δ ∈ Γ . To such a datum, the functor F F F assigns the collection of flat quasi-coherent sheaves ( H β γ δ ) β < γ < δ ∈ Γ ∈ M (\mathcal{H}_{\beta\gamma\delta})_{\beta< \gamma< \delta\in\Gamma} \in\mathrm{M} ( H β γ δ ) β < γ < δ ∈ Γ ∈ M with H β γ δ = F β \mathcal{H}_{\beta\gamma\delta} = \mathcal{F}_{\beta} H β γ δ = F β for all β < γ < δ ∈ Γ \beta< \gamma< \delta\in\Gamma β < γ < δ ∈ Γ . To the same datum, the functor G G G assigns the collection of flat quasi-coherent sheaves ( G β γ δ ) β < γ < δ ∈ Γ ∈ M (\mathcal{G}_{\beta\gamma\delta})_{\beta< \gamma< \delta\in\Gamma} \in\mathrm{M} ( G β γ δ ) β < γ < δ ∈ Γ ∈ M with G β γ δ = i β δ ∗ F δ \mathcal{G}_{\beta\gamma\delta} = i_{\beta\delta}^{*}\mathcal{F}_{\delta} G β γ δ = i β δ ∗ F δ for all β < γ < δ ∈ Γ \beta< \gamma< \delta\in\Gamma β < γ < δ ∈ Γ .
Consider the following pair of natural transformations ϕ , ψ : F ⇒ G \phi, \psi: F \Rightarrow G ϕ , ψ : F ⇒ G . Let D = ( ( F γ ) γ ∈ Γ , ( θ γ δ ) γ < δ ∈ Γ ) D = ((\mathcal{F}_{\gamma})_{\gamma\in\Gamma}, (\theta_{\gamma\delta})_{\gamma< \delta\in\Gamma}) D = (( F γ ) γ ∈ Γ , ( θ γ δ ) γ < δ ∈ Γ ) be an object of D \mathrm{D} D . For any triple of indices β < γ < δ ∈ Γ \beta< \gamma< \delta\in\Gamma β < γ < δ ∈ Γ , the corresponding component of the morphism ϕ D : F ( D ) ⟶ G ( D ) \phi_{D}: F(D) \longrightarrow G(D) ϕ D : F ( D ) ⟶ G ( D ) is the isomorphism θ β δ − 1 : F β ⟶ i β δ ∗ F δ \theta_{\beta\delta}^{-1}: \mathcal{F}_{\beta} \longrightarrow i_{\beta\delta}^{*}\mathcal{F}_{\delta} θ β δ − 1 : F β ⟶ i β δ ∗ F δ . For the same triple of indices β < γ < δ ∈ Γ \beta< \gamma< \delta\in\Gamma β < γ < δ ∈ Γ , the corresponding component of the morphism ψ D : F ( D ) ⟶ G ( D ) \psi_{D}: F(D) \longrightarrow G(D) ψ D : F ( D ) ⟶ G ( D ) is the isomorphism θ β γ − 1 ∘ i β γ ∗ ( θ γ δ − 1 ) : F β ⟶ i β δ ∗ F δ \theta_{\beta\gamma}^{-1} \circ i_{\beta\gamma}^{*}(\theta_{\gamma\delta}^{-1}): \mathcal{F}_{\beta} \longrightarrow i_{\beta\delta}^{*}\mathcal{F}_{\delta} θ β γ − 1 ∘ i β γ ∗ ( θ γ δ − 1 ) : F β ⟶ i β δ ∗ F δ .
Notice that all directed colimits exist in the categories D \mathrm{D} D and M \mathrm{M} M , and all of them are preserved by the functors F F F and G G G . Furthermore, the functors F F F and G G G take κ \kappa κ -presentable objects of D \mathrm{D} D to κ \kappa κ -presentable objects of M \mathrm{M} M . Once again, we take λ \lambda λ to be the countable cardinal, λ = ℵ 0 \lambda= \aleph_{0} λ = ℵ 0 . Then the assumptions of Proposition 4.3 are satisfied for the pair of functors F , G : D ⇒ M F, G: \mathrm{D} \Rightarrow\mathrm{M} F , G : D ⇒ M . Thus the equifier category E \mathrm{E} E is κ \kappa κ -accessible, and we also obtain a description of κ \kappa κ -presentable objects in E \mathrm{E} E .
Comparing the discussion in § 3.1 with the constructions above, one can immediately see that the equifier category E \mathrm{E} E is naturally equivalent to X - Proqc f l \mathcal{X}\text{-}\operatorname{Proqc}_{\mathrm{fl}} X - Proqc fl .
Accessibility for complexes
In this section we extend the result of Theorem 4.11 to complexes of flat pro-quasi-coherent pro-sheaves. The results that we obtain are a far-reaching partial generalization of the results about accessibility of various classes of complexes of flat contramodules from [35 Proposition 10.2 and Corollary 11.4] ; cf. Example 3.1 above.
Theorem 4.12. Let μ \mu μ be an infinite cardinal, κ ≥ ℵ 1 \kappa\ge\aleph_{1} κ ≥ ℵ 1 , κ > μ \kappa> \mu κ > μ be a regular cardinal, and X \mathfrak{X} X be an (ind-concentrated) μ \mu μ -ind-scheme. Then the category Com ( X – Proqc f l ) \operatorname{Com}(\mathfrak{X}\text{--}\operatorname{Proqc}_{\mathrm{fl}}) Com ( X – Proqc fl ) of complexes in the additive category X – Proqc f l \mathfrak{X}\text{--}\operatorname{Proqc}_{\mathrm{fl}} X – Proqc fl is κ \kappa κ -accessible. A complex of flat pro-quasi-coherent pro-sheaves F ∙ \mathcal{F}^{\bullet} F ∙ on X \mathfrak{X} X is κ \kappa κ -presentable as an object of Com ( X – Proqc f l ) \operatorname{Com}(\mathfrak{X}\text{--}\operatorname{Proqc}_{\mathrm{fl}}) Com ( X – Proqc fl ) if and only if all its terms F n \mathcal{F}^{n} F n , n ∈ Z n \in\mathbb{Z} n ∈ Z , are κ \kappa κ -presentable as objects of X – Proqc f l \mathfrak{X}\text{--}\operatorname{Proqc}_{\mathrm{fl}} X – Proqc fl . Equivalently, a complex F ∙ \mathcal{F}^{\bullet} F ∙ is κ \kappa κ -presentable in Com ( X – Proqc f l ) \operatorname{Com}(\mathfrak{X}\text{--}\operatorname{Proqc}_{\mathrm{fl}}) Com ( X – Proqc fl ) if and only if, for every closed subscheme Z ⊂ X Z \subset\mathfrak{X} Z ⊂ X , the complex of flat quasi-coherent sheaves F ∙ ( Z ) \mathcal{F}^{\bullet}(Z) F ∙ ( Z ) on Z Z Z is κ \kappa κ -presentable as an object of Com ( Z – Qcoh f l ) \operatorname{Com}(Z\text{--}\operatorname{Qcoh}_{\mathrm{fl}}) Com ( Z – Qcoh fl ) .
Proof. The proof is similar to that of Theorem 4.11 and using Proposition 4.9. One just needs to perform all the constructions from the proof of Theorem 4.11 working with complexes of (flat quasi-coherent) sheaves on the schemes X γ X_{\gamma} X γ instead of just sheaves. Basically, the only difference is that, in the context of the proof of the present theorem, one needs to consider the Cartesian product categories K = ∏ γ ∈ Γ Com ( X γ – Qcoh f l ) K = \prod_{\gamma\in\Gamma} \operatorname{Com}(X_{\gamma}\text{--}\operatorname{Qcoh}_{\mathrm{fl}}) K = ∏ γ ∈ Γ Com ( X γ – Qcoh fl ) , L = ∏ γ < δ ∈ Γ Com ( X γ – Qcoh f l ) L = \prod_{\gamma<\delta\in\Gamma} \operatorname{Com}(X_{\gamma}\text{--}\operatorname{Qcoh}_{\mathrm{fl}}) L = ∏ γ < δ ∈ Γ Com ( X γ – Qcoh fl ) , and M = ∏ β < γ < δ ∈ Γ Com ( X β – Qcoh f l ) M = \prod_{\beta<\gamma<\delta\in\Gamma} \operatorname{Com}(X_{\beta}\text{--}\operatorname{Qcoh}_{\mathrm{fl}}) M = ∏ β < γ < δ ∈ Γ Com ( X β – Qcoh fl ) . Alternatively, the assertions of the theorem can be deduced from Theorem 4.11 by applying Proposition 4.6. □
Theorem 4.13. Let μ \mu μ be an infinite cardinal, κ ≥ ℵ 1 \kappa\ge\aleph_{1} κ ≥ ℵ 1 , κ > μ \kappa> \mu κ > μ be a regular cardinal, and X \mathfrak{X} X be an (ind-concentrated) μ \mu μ -ind-scheme. Then the category Ac ( X – Proqc f l ) \operatorname{Ac}(\mathfrak{X}\text{--}\operatorname{Proqc}_{\mathrm{fl}}) Ac ( X – Proqc fl ) of acyclic complexes in the exact category X – Proqc f l \mathfrak{X}\text{--}\operatorname{Proqc}_{\mathrm{fl}} X – Proqc fl is κ \kappa κ -accessible. A (pure) acyclic complex of flat pro-quasi-coherent pro-sheaves F ∙ \mathcal{F}^{\bullet} F ∙ on X \mathfrak{X} X is κ \kappa κ -presentable as an object of Ac ( X – Proqc f l ) \operatorname{Ac}(\mathfrak{X}\text{--}\operatorname{Proqc}_{\mathrm{fl}}) Ac ( X – Proqc fl ) if and only if all its terms F n \mathcal{F}^{n} F n , n ∈ Z n \in\mathbb{Z} n ∈ Z , are κ \kappa κ -presentable as objects of X – Proqc f l \mathfrak{X}\text{--}\operatorname{Proqc}_{\mathrm{fl}} X – Proqc fl . Equivalently, a complex F ∙ \mathcal{F}^{\bullet} F ∙ is κ \kappa κ -presentable in Ac ( X – Proqc f l ) \operatorname{Ac}(\mathfrak{X}\text{--}\operatorname{Proqc}_{\mathrm{fl}}) Ac ( X – Proqc fl ) if and only if, for every closed subscheme Z ⊂ X Z \subset\mathfrak{X} Z ⊂ X , the pure acyclic complex of flat quasi-coherent sheaves F ∙ ( Z ) \mathcal{F}^{\bullet}(Z) F ∙ ( Z ) on $Z is κ \kappa κ -presentable as an object of Ac ( Z – Qcoh f l ) \operatorname{Ac}(Z\text{--}\operatorname{Qcoh}_{\mathrm{fl}}) Ac ( Z – Qcoh fl ) .
Proof. Similar to the proofs of Theorems 4.11 and 4.12, and using Proposition 4.10. The only difference is that, in the context of the proof of the present theorem, one needs to consider the Cartesian product categories K = ∏ γ ∈ Γ Ac ( X γ – Qcoh f l ) K = \prod_{\gamma\in\Gamma} \operatorname{Ac}(X_{\gamma}\text{--}\operatorname{Qcoh}_{\mathrm{fl}}) K = ∏ γ ∈ Γ Ac ( X γ – Qcoh fl ) , L = ∏ γ < δ ∈ Γ Ac ( X γ – Qcoh f l ) L = \prod_{\gamma<\delta\in\Gamma} \operatorname{Ac}(X_{\gamma}\text{--}\operatorname{Qcoh}_{\mathrm{fl}}) L = ∏ γ < δ ∈ Γ Ac ( X γ – Qcoh fl ) , and M = ∏ β < γ < δ ∈ Γ Ac ( X β – Qcoh f l ) M = \prod_{\beta<\gamma<\delta\in\Gamma} \operatorname{Ac}(X_{\beta}\text{--}\operatorname{Qcoh}_{\mathrm{fl}}) M = ∏ β < γ < δ ∈ Γ Ac ( X β – Qcoh fl ) . □
Remark 4.14. Similarly to (and rather simpler than) Theorem 4.13, one can prove that, under the same assumptions, the category of (admissible) short exact sequences in X – Proqc f l \mathfrak{X}\text{--}\operatorname{Proqc}_{\mathrm{fl}} X – Proqc fl is κ \kappa κ -accessible, and its κ \kappa κ -presentable objects are the short exact sequences of κ \kappa κ -presentable objects of X – Proqc f l \mathfrak{X}\text{--}\operatorname{Proqc}_{\mathrm{fl}} X – Proqc fl . (See [36 Corollary 10.13] or [37 Proposition 4.6] for a similar result for flat modules.) In the terminology of the paper [37 ] , this assertion can be restated by saying that X – Proqc f l \mathfrak{X}\text{--}\operatorname{Proqc}_{\mathrm{fl}} X – Proqc fl is a locally κ \kappa κ -coherent exact category (see [37 proof of Lemma 1.2] ).
Deconstructibility and the Hill Lemma
Background on deconstructibility
We suggest the papers [48 Sections 1–2] , [42 Section 4] , [46 Section 7] , [40 Section 1] , the lecture notes [51 Section 3] , and the book [29 Section B.2] as reference sources on deconstructibility in exact categories.
Let K be a category and λ \lambda λ be an ordinal. A λ \lambda λ -indexed chain in K is a functor λ ⟶ K \lambda\longrightarrow\mathrm{K} λ ⟶ K , in the sense explained in § 4.2 . In other words, a λ \lambda λ -indexed chain is a directed diagram ( K i ⟶ K j ) 0 ≤ i < j < λ (K_i \longrightarrow K_j)_{0\leq i<j<\lambda} ( K i ⟶ K j ) 0 ≤ i < j < λ of objects and morphisms in K indexed by the ordered set λ \lambda λ . A λ \lambda λ -indexed chain in K is said to be smooth if K j = lim i < j K i K_j=\lim\limits_{i<j}K_i K j = i < j lim K i in K for all limit ordinals 0 < j < λ 0<j<\lambda 0 < j < λ .
Let E be an exact category. A λ \lambda λ -indexed filtration ( F i ) 0 ≤ i ≤ λ (F_i)_{0\leq i\leq\lambda} ( F i ) 0 ≤ i ≤ λ (on the object F = F λ F=F_\lambda F = F λ ) in E is a smooth ( λ + 1 ) (\lambda+1) ( λ + 1 ) -indexed chain in E such that F 0 = F F_0=F F 0 = F and the transition morphisms F i ⟶ F i + 1 F_i\longrightarrow F_{i+1} F i ⟶ F i + 1 are admissible monomorphisms for all 0 ≤ i < λ 0\leq i<\lambda 0 ≤ i < λ .
If this is the case, denote by S i ∈ E S_i\in\mathrm{E} S i ∈ E , 0 ≤ i < λ 0\leq i<\lambda 0 ≤ i < λ , the cokernels of the admissible monomorphisms F i ⟶ F i + 1 F_i\longrightarrow F_{i+1} F i ⟶ F i + 1 . Then the object F = F λ F=F_\lambda F = F λ is said to be filtered by the objects S i S_i S i in E. In a different terminology, the object F F F is said to be a transfinitely iterated extension (in the sense of the directed colimit) of the objects { S i } 0 ≤ i < λ \{S_i\}_{0\leq i<\lambda} { S i } 0 ≤ i < λ .
Let S ⊂ E S\subset\mathrm{E} S ⊂ E be a class of objects. The class of all objects (that can be) filtered by (objects isomorphic to) objects from S S S in E is denoted by Fil ( S ) ⊂ E \operatorname{Fil}(S)\subset\mathrm{E} Fil ( S ) ⊂ E . A class of objects F ⊂ E F\subset\mathrm{E} F ⊂ E is said to be deconstructible if there exists a set (rather than a proper class) of objects S ⊂ E S\subset\mathrm{E} S ⊂ E such that F = Fil ( S ) F=\operatorname{Fil}(S) F = Fil ( S ) .
Background on the Hill lemma in Grothendieck categories
The material of this section is an extraction from the paper [50 ] .
Let κ \kappa κ be a regular cardinal. Given a category K with κ \kappa κ -directed colimits, an object K ∈ K K\in\mathrm{K} K ∈ K is said to be κ \kappa κ -generated if the functor Mor K ( K , − ) : K ⟶ S e t s \operatorname{Mor}_{\mathrm{K}}(K,-):\mathrm{K}\longrightarrow\mathrm{Sets} Mor K ( K , − ) : K ⟶ Sets preserves the colimits of all the κ \kappa κ -directed diagrams of monomorphisms in K. In a locally κ \kappa κ -presentable abelian category K, an object is κ \kappa κ -generated if and only if it is an epimorphic image of a κ \kappa κ -presentable object [1 Proposition 1.69] , [50 Lemma A.3] .
For any Grothendieck category K, there exists a regular cardinal κ \kappa κ such that the category K is κ \kappa κ -presentable [50 Lemma A.1] . If this is the case, then the classes of ρ \rho ρ -generated and ρ \rho ρ -presentable objects are closed under extensions in K for every regular cardinal ρ ≥ κ \rho\geq\kappa ρ ≥ κ [50 Lemma A.4] .
As the class of ρ \rho ρ -presentable objects is preserved by colimits of ρ \rho ρ -small diagrams (in any category with ρ \rho ρ -directed colimits) [1 Proposition 1.16] , it follows that the class of ρ \rho ρ -presentable objects in K is closed under transfinitely iterated extensions indexed by ordinals λ < ρ \lambda<\rho λ < ρ [50 Corollary A.5] .
A complete lattice L L L is a poset where every subset has a least upper bound (the join) and a greatest lower bound (the meet). A complete sublattice in a lattice L L L is a subset of L L L closed under the joins and the meets of all subsets. A homomorphism of complete lattices is a map preserving all the joins and the meets.
Given a set Λ \Lambda Λ , we denote by P ( Λ ) = ( P ( Λ ) , ∪ , ∩ ) \mathcal{P}(\Lambda)=(\mathcal{P}(\Lambda),\cup,\cap) P ( Λ ) = ( P ( Λ ) , ∪ , ∩ ) the complete lattice of all subsets of Λ \Lambda Λ (ordered by inclusion). Given a complete, cocomplete abelian category K and an object M ∈ K M \in\mathcal{K} M ∈ K , we denote by Sub ( M ) = ( Sub ( M ) , Σ , ∩ ) \operatorname{Sub}(M) = (\operatorname{Sub}(M), \Sigma, \cap) Sub ( M ) = ( Sub ( M ) , Σ , ∩ ) the complete lattice of all subobjects of M M M (ordered by inclusion).
Theorem 5.1. Let κ \kappa κ be a regular cardinal and K \mathcal{K} K be a locally κ \kappa κ -presentable Grothendieck abelian category. Let λ \lambda λ be an ordinal and M ∈ K M \in\mathcal{K} M ∈ K be an object endowed with a λ \lambda λ -indexed filtration ( M i ⊂ M ) 0 ≤ i ≤ λ (M_i \subset M)_{0 \leq i \leq\lambda} ( M i ⊂ M ) 0 ≤ i ≤ λ such that the successive quotient objects S i = M i + 1 / M i S_i = M_{i+1}/M_i S i = M i + 1 / M i are κ \kappa κ -presentable in K \mathcal{K} K for all 0 ≤ i < λ 0 \leq i < \lambda 0 ≤ i < λ . Then there exists a complete sublattice L ⊂ P ( λ ) L \subset P(\lambda) L ⊂ P ( λ ) and a map ℓ : L ⟶ Sub ( M ) \ell: L \longrightarrow\operatorname{Sub}(M) ℓ : L ⟶ Sub ( M ) with the following properties:
(H1) For every ordinal 0 ≤ α ≤ λ 0 \leq\alpha\leq\lambda 0 ≤ α ≤ λ viewed as a subset α ⊂ λ \alpha\subset\lambda α ⊂ λ , one has α ∈ L \alpha\in L α ∈ L and ℓ ( α ) = M α \ell(\alpha) = M_\alpha ℓ ( α ) = M α .
(H2) The map ℓ \ell ℓ is a homomorphism of complete lattices. So ℓ \ell ℓ takes the unions of arbitrary families of subsets of λ \lambda λ belonging to L L L to the sums of subobjects in M M M , and the intersections of arbitrary families of subsets of λ \lambda λ belonging to L L L to the intersections of subobjects in M M M .
(H3) For any two subsets A , B ⊂ λ A, B \subset\lambda A , B ⊂ λ such that A , B ∈ L A, B \in L A , B ∈ L and A ⊂ B A \subset B A ⊂ B , the object N = ℓ ( B ) / ℓ ( A ) ∈ K N = \ell(B)/\ell(A) \in\mathcal{K} N = ℓ ( B ) / ℓ ( A ) ∈ K has a filtration ( N j ) 0 ≤ j ≤ ν (N_j)_{0 \leq j \leq\nu} ( N j ) 0 ≤ j ≤ ν indexed by some ordinal ν \nu ν endowed with a bijection b : B ∖ A ⟶ ν b: B \setminus A \longrightarrow\nu b : B ∖ A ⟶ ν such that there is an isomorphism M i + 1 / M i ≅ N b ( i ) + 1 / N b ( i ) M_{i+1}/M_i \cong N_{b(i)+1}/N_{b(i)} M i + 1 / M i ≅ N b ( i ) + 1 / N b ( i ) in K \mathcal{K} K for every i ∈ B ∖ A i \in B \setminus A i ∈ B ∖ A .
(H4) For any κ \kappa κ -generated subobject K ⊂ M K \subset M K ⊂ M , there exists a subset C ⊂ λ C \subset\lambda C ⊂ λ , C ∈ L C \in L C ∈ L , with the cardinality of C C C smaller than κ \kappa κ , such that K ⊂ ℓ ( C ) ⊂ M K \subset\ell(C) \subset M K ⊂ ℓ ( C ) ⊂ M .
Proof. This is the statement of the Hill lemma in Grothendieck categories [50 Theorem 2.1] . □
Given a class of objects F \mathcal{F} F in an additive category K \mathcal{K} K , we denote by F ⊕ ⊂ K \mathcal{F}^{\oplus} \subset\mathcal{K} F ⊕ ⊂ K the class of all direct summands of objects from F \mathcal{F} F in K \mathcal{K} K .
Let κ \kappa κ be a regular cardinal and K \mathcal{K} K be a locally κ \kappa κ -presentable Grothendieck category. A class of objects F ⊂ K \mathcal{F} \subset\mathcal{K} F ⊂ K is said to be κ \kappa κ -deconstructible if F = Fil ( S ) ⊂ K \mathcal{F} = \operatorname{Fil}(\mathcal{S}) \subset\mathcal{K} F = Fil ( S ) ⊂ K for some set of κ \kappa κ -presentable objects S \mathcal{S} S in K \mathcal{K} K .
Proposition 5.2. Let κ ≥ ℵ 1 \kappa\geq\aleph_1 κ ≥ ℵ 1 be an uncountable regular cardinal and K \mathcal{K} K be a locally κ \kappa κ -presentable Grothendieck category. Then, for any κ \kappa κ -deconstructible class of objects F ⊂ K \mathcal{F} \subset\mathcal{K} F ⊂ K , the class of objects F ⊕ ⊂ K \mathcal{F}^{\oplus} \subset\mathcal{K} F ⊕ ⊂ K is also κ \kappa κ -deconstructible.
Proof. This is [50 Proposition 2.9(1)] . □
Given an exact category K \mathcal{K} K and a full subcategory E ⊂ K \mathcal{E} \subset\mathcal{K} E ⊂ K closed under extensions in K \mathcal{K} K , we always endow E \mathcal{E} E with the inherited exact category structure. We also endow the category of complexes Com ( K ) \operatorname{Com}(\mathcal{K}) Com ( K ) with the termwise exact structure; so a short sequence of complexes in K \mathcal{K} K is (admissible) exact if and only if it is exact in every degree. Then the termwise exact structure on Com ( E ) \operatorname{Com}(\mathcal{E}) Com ( E ) is inherited from the termwise exact structure from Com ( K ) \operatorname{Com}(\mathcal{K}) Com ( K ) .
For an abelian category K \mathcal{K} K endowed with the abelian exact structure, the termwise exact structure on Com ( K ) \operatorname{Com}(\mathcal{K}) Com ( K ) coincides with the abelian exact structure.
Proposition 5.3. Let κ ≥ ℵ 1 \kappa\ge\aleph_{1} κ ≥ ℵ 1 be an uncountable regular cardinal and K \mathbf{K} K be a locally κ \kappa κ -presentable Grothendieck category. Let F ⊂ K \mathbf{F} \subset\mathbf{K} F ⊂ K be a κ \kappa κ -deconstructible class of objects in K \mathbf{K} K (viewed as a full subcategory in K \mathbf{K} K ). Then the class of objects Com ( F ) \operatorname{Com}(\mathbf{F}) Com ( F ) is κ \kappa κ -deconstructible in the Grothendieck category Com ( K ) \operatorname{Com}(\mathbf{K}) Com ( K ) .
Proof. This is [50 Proposition 4.3] . It only needs to be mentioned that the Grothendieck category Com ( K ) \operatorname{Com}(\mathbf{K}) Com ( K ) is locally κ \kappa κ -presentable, and the κ \kappa κ -presentable objects of Com ( K ) \operatorname{Com}(\mathbf{K}) Com ( K ) are precisely all the complexes of κ \kappa κ -presentable objects in K \mathbf{K} K , by Lemma 4.7 above. □
Proposition 5.4. Let κ ≥ ℵ 1 \kappa\ge\aleph_{1} κ ≥ ℵ 1 be an uncountable regular cardinal and K \mathbf{K} K be a locally κ \kappa κ -presentable Grothendieck category. Let F ⊂ K \mathbf{F} \subset\mathbf{K} F ⊂ K be a κ \kappa κ -deconstructible class of objects in K \mathbf{K} K (viewed as a full subcategory in K \mathbf{K} K with the inherited exact structure). Then the class of objects Ac ( F ) \operatorname{Ac}(\mathbf{F}) Ac ( F ) is κ \kappa κ -deconstructible in the Grothendieck category Com ( K ) \operatorname{Com}(\mathbf{K}) Com ( K ) .
Proof. This is [50 Proposition 4.4] . □
Background on deconstructibility over schemes
Let κ \kappa κ be a cardinal and X X X be a scheme with an affine open covering X = ⋃ α U α X = \bigcup_{\alpha} U_{\alpha} X = ⋃ α U α . We will say that the affine open covering X = ⋃ α U α X = \bigcup_{\alpha} U_{\alpha} X = ⋃ α U α can be extended to a κ \kappa κ -small affine open covering of the intersections if there exists an affine open covering X = ⋃ ξ ∈ Ξ U ξ X = \bigcup_{\xi\in\Xi} U_{\xi} X = ⋃ ξ ∈ Ξ U ξ of the scheme X X X such that
the set of indices { α } \{\alpha\} { α } is a subset of Ξ \Xi Ξ ;
for any two indices υ ′ , υ ′ ′ ∈ Ξ \upsilon', \upsilon'' \in\Xi υ ′ , υ ′′ ∈ Ξ , the intersection U υ ′ ∩ U υ ′ ′ ⊂ X U_{\upsilon'} \cap U_{\upsilon''} \subset X U υ ′ ∩ U υ ′′ ⊂ X is the union of those open subschemes U ξ U_{\xi} U ξ , ξ ∈ Ξ \xi\in\Xi ξ ∈ Ξ , that are contained in U υ ′ ∩ U υ ′ ′ U_{\upsilon'} \cap U_{\upsilon''} U υ ′ ∩ U υ ′′ ;
the cardinality of the set of indices Ξ \Xi Ξ is smaller than κ \kappa κ .
One can easily see that any finite affine open covering of a concentrated scheme can be extended to a countable (i.e., ℵ 1 \aleph_{1} ℵ 1 -small) affine open covering of the intersections. In fact, the same applies to any countable affine open covering of a countably quasi-compact countably quasi-separated scheme in the sense of [46 Section 4] . Moreover, given a regular cardinal κ \kappa κ , a scheme X X X admits an affine open covering that can be extended to a κ \kappa κ -small affine open covering of the intersections if and only if the scheme X X X is κ \kappa κ -quasi-compact and κ \kappa κ -quasi-separated in the sense of [46 Remark 4.6] .
Proposition 5.5. Let κ ≥ ℵ 1 \kappa\ge\aleph_{1} κ ≥ ℵ 1 be an uncountable regular cardinal and X X X be a scheme with an affine open covering X = ⋃ α U α X = \bigcup_{\alpha} U_{\alpha} X = ⋃ α U α that can be extended to a κ \kappa κ -small affine open covering of the intersections. Suppose that we are given, for every index α \alpha α , a κ \kappa κ -deconstructible class of modules F α ⊂ O X ( U α ) -Mod \mathbf{F}_{\alpha} \subset\mathcal{O}_{X}(U_{\alpha})\text{-Mod} F α ⊂ O X ( U α ) -Mod . Denote by F X ⊂ X -Qcoh \mathbf{F}_{X} \subset X\text{-Qcoh} F X ⊂ X -Qcoh the class of all quasi-coherent sheaves F \mathcal{F} F on X X X such that F ( U α ) ∈ F α \mathcal{F}(U_{\alpha}) \in\mathbf{F}_{\alpha} F ( U α ) ∈ F α for every index α \alpha α . Then the class F X \mathbf{F}_{X} F X is κ \kappa κ -deconstructible in X -Qcoh X\text{-Qcoh} X -Qcoh .
Proof. This is essentially [15 Notation 3.12 and Theorem 3.13] . To arrive at our formulation of this result from [15 ] , one needs to use the facts that the Grothendieck category X -Qcoh X\text{-Qcoh} X -Qcoh is locally κ \kappa κ -presentable, and a quasi-coherent sheaf on X X X is κ \kappa κ -presentable as an object of X -Qcoh X\text{-Qcoh} X -Qcoh if and only if it is locally κ \kappa κ -presentable; this is essentially [46 Remark 4.6] . For another formulation of a comparable assertion, see [29 Lemma B.3.5] . □
Corollary 5.6. In the context and notation of Proposition 5.5, consider the Grothendieck category Com ( X – Qcoh ) \operatorname{Com}(X\text{--}\operatorname{Qcoh}) Com ( X – Qcoh ) of complexes of quasi-coherent sheaves on X X X and the class Com ( F X ) ⊂ Com ( X – Qcoh ) \operatorname{Com}(\mathcal{F}_{X}) \subset\operatorname{Com}(X\text{--}\operatorname{Qcoh}) Com ( F X ) ⊂ Com ( X – Qcoh ) of all complexes with the terms belonging to F X \mathcal{F}_{X} F X . Then the class Com ( F X ) \operatorname{Com}(\mathcal{F}_{X}) Com ( F X ) is κ \kappa κ -deconstructible in Com ( X – Qcoh ) \operatorname{Com}(X\text{--}\operatorname{Qcoh}) Com ( X – Qcoh ) .
Proof. Compare Propositions 5.3 and 5.5. □
Corollary 5.7. In the context and notation of Proposition 5.5, consider the Grothendieck category Com ( X – Qcoh ) \operatorname{Com}(X\text{--}\operatorname{Qcoh}) Com ( X – Qcoh ) and the class Ac ( F X ) ⊂ Com ( X – Qcoh ) \operatorname{Ac}(\mathcal{F}_{X}) \subset\operatorname{Com}(X\text{--}\operatorname{Qcoh}) Ac ( F X ) ⊂ Com ( X – Qcoh ) of all acyclic complexes of quasi-coherent sheaves on X X X with the terms belonging to F X \mathcal{F}_{X} F X and the sheaves of cocycles also belonging to F X \mathcal{F}_{X} F X . Then the class Ac ( F X ) \operatorname{Ac}(\mathcal{F}_{X}) Ac ( F X ) is κ \kappa κ -deconstructible in Com ( X – Qcoh ) \operatorname{Com}(X\text{--}\operatorname{Qcoh}) Com ( X – Qcoh ) .
Proof. Compare Propositions 5.4 and 5.5. □
Lemma 5.8. (a) Let κ ≥ ℵ 1 \kappa\ge\aleph_{1} κ ≥ ℵ 1 be an uncountable regular cardinal and R R R be an associative ring of cardinality smaller than κ \kappa κ . Then the class of all flat left R R R -modules R – Mod f l R\text{--}\operatorname{Mod}_{\mathrm{fl}} R – Mod fl is κ \kappa κ -deconstructible in R – Mod R\text{--}\operatorname{Mod} R – Mod }.
(b) Let R R R be a commutative ring. Then the class of all very flat R R R -modules R – Mod v f l R\text{--}\operatorname{Mod}_{\mathrm{vfl}} R – Mod vfl is ℵ 1 \aleph_{1} ℵ 1 -deconstructible in R – Mod R\text{--}\operatorname{Mod} R – Mod .
Proof. Part (a) is [6 Lemma 1 and proof of Proposition 2] . Part (b) is [49 Lemma 2.2] (the assertion follows from [29 Corollary 1.1.4] , see the discussion in Section 3.3 above, by virtue of Proposition 5.2). □
Corollary 5.9. Let κ ≥ ℵ 1 \kappa\ge\aleph_{1} κ ≥ ℵ 1 be an uncountable regular cardinal and X X X be a scheme with an affine open covering X = ⋃ α U α X = \bigcup_{\alpha} U_{\alpha} X = ⋃ α U α that can be extended to a κ \kappa κ -small affine open covering of the intersections. In this setting:
(a) Assume that the cardinalities of all the rings O X ( U α ) \mathcal{O}_{X}(U_{\alpha}) O X ( U α ) are smaller than κ \kappa κ . Then the class X – Qcoh f l X\text{--}\operatorname{Qcoh}_{\mathrm{fl}} X – Qcoh fl of all flat quasi-coherent sheaves on X X X is κ \kappa κ -deconstructible in X – Qcoh X\text{--}\operatorname{Qcoh} X – Qcoh }.
(b) The class X – Qcoh v f l X\text{--}\operatorname{Qcoh}_{\mathrm{vfl}} X – Qcoh vfl of all very flat quasi-coherent sheaves on X X X is κ \kappa κ -deconstructible in X – Qcoh X\text{--}\operatorname{Qcoh} X – Qcoh .
Proof. This was mentioned, in some form, in [29 proof of Lemma 5.1.1] . To prove the corollary, compare Proposition 5.5 with Lemma 5.8. □
Corollary 5.10. Let κ ≥ ℵ 1 \kappa\ge\aleph_{1} κ ≥ ℵ 1 be an uncountable regular cardinal and X X X be a scheme with an affine open covering X = ⋃ α U α X = \bigcup_{\alpha} U_{\alpha} X = ⋃ α U α that can be extended to a κ \kappa κ -small affine open covering of the intersections. In this setting:
(a) Assume that the cardinalities of all the rings O X ( U α ) \mathcal{O}_{X}(U_{\alpha}) O X ( U α ) are smaller than κ \kappa κ . Then the class Com ( X – Qcoh f l ) \operatorname{Com}(X\text{--}\operatorname{Qcoh}_{\mathrm{fl}}) Com ( X – Qcoh fl ) of all complexes of flat quasi-coherent sheaves on X X X is κ \kappa κ -deconstructible in Com ( X – Qcoh ) \operatorname{Com}(X\text{--}\operatorname{Qcoh}) Com ( X – Qcoh ) .
(b) The class Com ( X – Qcoh v f l ) \operatorname{Com}(X\text{--}\operatorname{Qcoh}_{\mathrm{vfl}}) Com ( X – Qcoh vfl ) of all complexes of very flat quasi-coherent sheaves on X X X is κ \kappa κ -deconstructible in Com ( X – Qcoh ) \operatorname{Com}(X\text{--}\operatorname{Qcoh}) Com ( X – Qcoh ) .
Proof. This was mentioned, in some form, in [29 proof of Lemma 5.1.9] . To prove the corollary, compare Corollary 5.6 with Lemma 5.8.
Corollary 5.11. Let κ ≥ ℵ 1 \kappa\ge\aleph_{1} κ ≥ ℵ 1 be an uncountable regular cardinal and X X X be a scheme with an affine open covering X = ⋃ α U α X = \bigcup_{\alpha} U_{\alpha} X = ⋃ α U α that can be extended to a κ \kappa κ -small affine open covering of the intersections. In this setting:
(a) Assume that the cardinalities of all the rings O X ( U α ) \mathcal{O}_{X}(U_{\alpha}) O X ( U α ) are smaller than κ \kappa κ . Then the class Ac ( X – Q c o h f l ) \operatorname{Ac}(X\text{--}\mathrm{Qcoh}_{\mathrm{fl}}) Ac ( X – Qcoh fl ) of all acyclic complexes of flat quasi-coherent sheaves on X X X with flat quasi-coherent sheaves of cocycles is κ \kappa κ -deconstructible in Com ( X – Q c o h ) \operatorname{Com}(X\text{--}\mathrm{Qcoh}) Com ( X – Qcoh ) .
(b) The class Ac ( X – Q c o h v f l ) \operatorname{Ac}(X\text{--}\mathrm{Qcoh}_{\mathrm{vfl}}) Ac ( X – Qcoh vfl ) of all acyclic complexes of very flat quasi-coherent sheaves on X X X with very flat quasi-coherent sheaves of cocycles is κ \kappa κ -deconstructible in Com ( X – Q c o h ) \operatorname{Com}(X\text{--}\mathrm{Qcoh}) Com ( X – Qcoh ) .
Proof. This was also mentioned, in some form, in [29 proof of Lemma 5.1.1] . To prove the corollary, compare Corollary 5.7 with Lemma 5.8. □ \square □
Deconstructing flat and very flat pro-quasi-coherent pro-sheaves
We start with a simple category-theoretic lemma. Given a regular cardinal κ \kappa κ , a cocomplete abelian category K \mathcal{K} K is said to be locally κ \kappa κ -generated if K \mathcal{K} K has a set of generators consisting of κ \kappa κ -generated objects. For example, any locally κ \kappa κ -presentable abelian category is locally κ \kappa κ -generated.
In any locally κ \kappa κ -generated cocomplete abelian category, the functors of κ \kappa κ -directed colimits are exact. This is a version of [46 Corollary 9.6] for an arbitrary regular cardinal κ \kappa κ in place of the countable cardinal, provable in the same way. Alternatively, instead of using this fact, one could simply assume the category L \mathcal{L} L in part (b) of the next lemma to be Grothendieck, which is sufficient for our purposes.
Lemma 5.12. (a) Let K \mathcal{K} K be locally κ \kappa κ -generated cocomplete abelian category. Let N ⊂ M ∈ K N \subset M \in\mathcal{K} N ⊂ M ∈ K be an object with a subobject such that the quotient object M / N M/N M / N is κ \kappa κ -generated. Then there exists a κ \kappa κ -generated subobject K ⊂ M K \subset M K ⊂ M such that M = N + K M = N+K M = N + K .
(b) Let K \mathcal{K} K and L \mathcal{L} L be locally κ \kappa κ -generated cocomplete abelian categories and F : K ⟶ L F:\mathcal{K}\longrightarrow\mathcal{L} F : K ⟶ L be a functor preserving epimorphisms and κ \kappa κ -directed colimits. Let N ⊂ M ∈ K N \subset M \in\mathcal{K} N ⊂ M ∈ K be an object with a subobject, and let H ⊂ F ( M / N ) ∈ L H \subset F(M/N) \in\mathcal{L} H ⊂ F ( M / N ) ∈ L be a κ \kappa κ -generated subobject in F ( M / N ) F(M/N) F ( M / N ) . Then there exists a κ \kappa κ -generated subobject K ⊂ M K \subset M K ⊂ M such that the image of the induced morphism F ( K ) ⟶ F ( M / N ) F(K)\longrightarrow F(M/N) F ( K ) ⟶ F ( M / N ) contains H H H .
Proof. Let us prove part (b). By assumption, there exists a set Ξ \Xi Ξ and a Ξ \Xi Ξ -indexed family of κ \kappa κ -generated objects S ξ ∈ K S_{\xi} \in\mathcal{K} S ξ ∈ K together with a morphism f ξ : S ξ ⟶ M f_{\xi}:S_{\xi}\longrightarrow M f ξ : S ξ ⟶ M for each ξ ∈ Ξ \xi\in\Xi ξ ∈ Ξ such that the induced morphism f : ∐ ξ ∈ Ξ S ξ ⟶ M f:\coprod_{\xi\in\Xi}S_{\xi}\longrightarrow M f : ∐ ξ ∈ Ξ S ξ ⟶ M is an epimorphism in K \mathcal{K} K . Denote by Υ \Upsilon Υ the poset of all subsets of Ξ \Xi Ξ of cardinality smaller than κ \kappa κ , ordered by inclusion. Put K ν = ∐ ξ ∈ ν S ξ K_{\nu}=\coprod_{\xi\in\nu}S_{\xi} K ν = ∐ ξ ∈ ν S ξ for every ν ∈ Υ \nu\in\Upsilon ν ∈ Υ . Then Υ \Upsilon Υ is κ \kappa κ -directed poset, K ν ∈ K K_{\nu}\in\mathcal{K} K ν ∈ K are κ \kappa κ -generated objects, and ∐ ξ ∈ Ξ S ξ = lim ⟶ ν ∈ Υ K ν \coprod_{\xi\in\Xi}S_{\xi}=\lim\limits_{\longrightarrow_{\nu\in\Upsilon}}K_{\nu} ∐ ξ ∈ Ξ S ξ = ⟶ ν ∈ Υ lim K ν .
So we have an epimorphism f : lim ⟶ ν ∈ Υ K ν ⟶ M f:\lim\limits_{\longrightarrow_{\nu\in\Upsilon}}K_{\nu}\longrightarrow M f : ⟶ ν ∈ Υ lim K ν ⟶ M in K \mathcal{K} K , and it follows that F ( f ) : lim ⟶ ν ∈ Υ F ( K ν ) ⟶ F ( M ) F(f):\lim\limits_{\longrightarrow_{\nu\in\Upsilon}}F(K_{\nu})\longrightarrow F(M) F ( f ) : ⟶ ν ∈ Υ lim F ( K ν ) ⟶ F ( M ) is an epimorphism in L \mathcal{L} L . The morphism F ( M ) ⟶ F ( M / N ) F(M)\longrightarrow F(M/N) F ( M ) ⟶ F ( M / N ) is an epimorphism in L \mathcal{L} L , too. Denote by L ν ⊂ F ( M / N ) L_{\nu}\subset F(M/N) L ν ⊂ F ( M / N ) the image of the morphism F ( K ν ) ⟶ F ( M / N ) F(K_{\nu})\longrightarrow F(M/N) F ( K ν ) ⟶ F ( M / N ) . Then the epimorphism lim ⟶ ν ∈ Υ F ( K ν ) ⟶ F ( M / N ) \lim\limits_{\longrightarrow_{\nu\in\Upsilon}}F(K_{\nu})\longrightarrow F(M/N) ⟶ ν ∈ Υ lim F ( K ν ) ⟶ F ( M / N ) factorizes as lim ⟶ ν ∈ Υ F ( K ν ) ⟶ lim ⟶ ν ∈ Υ L ν ⟶ F ( M / N ) \lim\limits_{\longrightarrow_{\nu\in\Upsilon}}F(K_{\nu})\longrightarrow\lim\limits_{\longrightarrow_{\nu\in\Upsilon}}L_{\nu}\longrightarrow F(M/N) ⟶ ν ∈ Υ lim F ( K ν ) ⟶ ⟶ ν ∈ Υ lim L ν ⟶ F ( M / N ) . By a suitable version of [46 Proposition 9.5] , the morphism lim υ ∈ Υ L υ ⟶ F ( M / N ) \lim_{\upsilon\in\Upsilon} L_{\upsilon}\longrightarrow F(M/N) lim υ ∈ Υ L υ ⟶ F ( M / N ) is a monomorphism. Thus the morphism lim υ ∈ Υ L υ ⟶ F ( M / N ) \lim_{\upsilon\in\Upsilon} L_{\upsilon}\longrightarrow F(M/N) lim υ ∈ Υ L υ ⟶ F ( M / N ) is actually an isomorphism. Consider the intersections of subobjects L υ ∩ H L_{\upsilon}\cap H L υ ∩ H in F ( M / N ) F(M/N) F ( M / N ) . By suitable versions of [46 Proposition 9.5] and [26 Proposition III.1.2] , it follows that the morphism lim υ ∈ Υ ( L υ ∩ H ) ⟶ H \lim_{\upsilon\in\Upsilon}(L_{\upsilon}\cap H)\longrightarrow H lim υ ∈ Υ ( L υ ∩ H ) ⟶ H is an isomorphism. Since the object H H H is κ \kappa κ -generated and the poset Υ \Upsilon Υ is κ \kappa κ -directed, we can conclude that there exists υ ∈ Υ \upsilon\in\Upsilon υ ∈ Υ such that H = L υ ∩ H H=L_{\upsilon}\cap H H = L υ ∩ H . In other words, we have H ⊂ L υ H\subset L_{\upsilon} H ⊂ L υ . It remains to denote by K K K the image of the morphism K υ ⟶ M K_{\upsilon}\longrightarrow M K υ ⟶ M .
Part (a) is a special case of part (b) for the categories K = L \mathsf{K}=\mathsf{L} K = L , the identity functor F = Id F=\operatorname{Id} F = Id , and the subobject H = M / N H=M/N H = M / N . □
The definition of a μ \mu μ -ind-scheme for an infinite cardinal μ \mu μ was given in Section 4.4. Given an ind-scheme \mathcal{X}=\mathop{\mathrm{``lim"}\limits_{\gamma\in\Gamma}}X_{\gamma} and a flat pro-quasi-coherent pro-sheaf F \mathcal{F} F on X \mathcal{X} X , we denote by F ( γ ) = F ( X γ ) ∈ X γ – Qcoh f l \mathcal{F}^{(\gamma)}=\mathcal{F}(X_{\gamma})\in X_{\gamma}\text{--}\operatorname{Qcoh}_{\mathrm{fl}} F ( γ ) = F ( X γ ) ∈ X γ – Qcoh fl the inverse images of F \mathcal{F} F to X γ X_{\gamma} X γ .
Let us extend our terminology of “κ \kappa κ -deconstructible classes” to κ \kappa κ -accessible exact categories. Let κ \kappa κ be a regular cardinal and E \mathbf{E} E be an exact category whose underlying (additive) category E \mathbf{E} E is κ \kappa κ -accessible. Then we will say that a class of objects F ⊂ E \mathbf{F}\subset\mathbf{E} F ⊂ E is κ \kappa κ -deconstructible if F = Fil ( S ) \mathbf{F}=\operatorname{Fil}(\mathbf{S}) F = Fil ( S ) , where S \mathbf{S} S is a set of objects in E \mathbf{E} E and all the objects from S \mathbf{S} S are κ \kappa κ -presentable in E \mathbf{E} E .
The following theorem is the first main result of Section 5.
Theorem 5.13. Let κ ≥ ℵ 1 \kappa\geq\aleph_{1} κ ≥ ℵ 1 be an uncountable regular cardinal and \mathcal{X}=\mathop{\mathrm{``lim"}\limits_{\gamma\in\Gamma}}X_{\gamma} be an ind-scheme represented by a directed diagram of concentrated schemes indexed by a directed poset Γ \Gamma Γ of cardinality smaller than κ \kappa κ . Suppose that we are given, for every index γ ∈ Γ \gamma\in\Gamma γ ∈ Γ , a class of flat quasi-coherent sheaves F γ ⊂ X γ – Qcoh f l \mathbf{F}_{\gamma}\subset X_{\gamma}\text{--}\operatorname{Qcoh}_{\mathrm{fl}} F γ ⊂ X γ – Qcoh fl on X γ X_{\gamma} X γ such that the class F γ \mathbf{F}_{\gamma} F γ is κ \kappa κ -deconstructible in X γ – Qcoh X_{\gamma}\text{--}\operatorname{Qcoh} X γ – Qcoh (or equivalently, in X γ – Qcoh f l X_{\gamma}\text{--}\operatorname{Qcoh}_{\mathrm{fl}} X γ – Qcoh fl ). Denote by F X ⊂ X – Proqc f l \mathbf{F}_{\mathcal{X}}\subset\mathcal{X}\text{--}\operatorname{Proqc}_{\mathrm{fl}} F X ⊂ X – Proqc fl the class of all flat pro-quasi-coherent pro-sheaves F \mathcal{F} F on X \mathcal{X} X such that F ( γ ) ∈ F γ \mathcal{F}^{(\gamma)}\in\mathbf{F}_{\gamma} F ( γ ) ∈ F γ for every index γ ∈ Γ \gamma\in\Gamma γ ∈ Γ . Then the class F X \mathbf{F}_{\mathcal{X}} F X is κ \kappa κ -deconstructible in X – Proqc f l \mathcal{X}\text{--}\operatorname{Proqc}_{\mathrm{fl}} X – Proqc fl .
Proof. This is a far-reaching partial generalization of the discussion of deconstructibility of classes of contramodules in [42 Corollary 7.6 and Examples 7.12(2–3)] ; cf. Examples 3.1 and 3.2 above. We refer to Theorem 4.11 for the underlying assertion that the category X – Proqc f l \mathcal{X}\text{--}\operatorname{Proqc}_{\mathrm{fl}} X – Proqc fl is κ \kappa κ -accessible, and for the description of the κ \kappa κ -presentable objects of X – Proqc f l \mathcal{X}\text{--}\operatorname{Proqc}_{\mathrm{fl}} X – Proqc fl . The proof of the theorem is based on the Hill lemma in the Grothendieck categories X γ – Qcoh X_{\gamma}\text{--}\operatorname{Qcoh} X γ – Qcoh (Theorem 5.1).
To begin with, recall our usual notation i γ δ : X γ ⟶ X δ i_{\gamma\delta}:X_{\gamma}\longrightarrow X_{\delta} i γ δ : X γ ⟶ X δ , γ < δ ∈ Γ \gamma<\delta\in\Gamma γ < δ ∈ Γ , for the transition morphisms in the directed diagram of schemes ( X γ ) γ ∈ Γ (X_{\gamma})_{\gamma\in\Gamma} ( X γ ) γ ∈ Γ . Denote by S γ ⊂ X γ – Qcoh f l \mathbf{S}_{\gamma}\subset X_{\gamma}\text{--}\operatorname{Qcoh}_{\mathrm{fl}} S γ ⊂ X γ – Qcoh fl the class (up to isomorphism, a set) of all κ \kappa κ -presentable objects in the class F γ ⊂ X γ – Qcoh f l \mathbf{F}_{\gamma}\subset X_{\gamma}\text{--}\operatorname{Qcoh}_{\mathrm{fl}} F γ ⊂ X γ – Qcoh fl . Denote by S X \mathbf{S}_{\mathcal{X}} S X the class (up to isomorphism, a set) of all flat pro-quasi-coherent pro-sheaves S \mathcal{S} S on X \mathcal{X} X such that S ( γ ) ∈ S γ \mathcal{S}^{(\gamma)}\in\mathbf{S}_{\gamma} S ( γ ) ∈ S γ for all γ ∈ Γ \gamma\in\Gamma γ ∈ Γ . We claim that F X = Fil ( S X ) \mathbf{F}_{\mathcal{X}}=\operatorname{Fil}(\mathbf{S}_{\mathcal{X}}) F X = Fil ( S X ) . The inclusion Fil ( S X ) ⊂ F X \operatorname{Fil}(\mathbf{S}_{\mathcal{X}})\subset\mathbf{F}_{\mathcal{X}} Fil ( S X ) ⊂ F X is obvious; our task is to prove that F X ⊂ Fil ( S X ) \mathbf{F}_{\mathcal{X}}\subset\operatorname{Fil}(\mathbf{S}_{\mathcal{X}}) F X ⊂ Fil ( S X ) . More specifically, let F ∈ F X \mathcal{F} \in\mathrm{F}_{\mathcal{X}} F ∈ F X be a flat pro-quasi-coherent pro-sheaf and λ \lambda λ be an ordinal such that, for every γ ∈ Γ \gamma\in\Gamma γ ∈ Γ , we have a λ \lambda λ -indexed filtration ( G i ( γ ) F ( γ ) ) 0 ≤ i ≤ λ (G_i^{(\gamma)}\mathcal{F}^{(\gamma)})_{0 \leq i \leq\lambda} ( G i ( γ ) F ( γ ) ) 0 ≤ i ≤ λ on F ( γ ) \mathcal{F}^{(\gamma)} F ( γ ) with the successive quotients G i + 1 ( γ ) F ( γ ) / G i ( γ ) F ( γ ) G_{i+1}^{(\gamma)}\mathcal{F}^{(\gamma)}/G_i^{(\gamma)}\mathcal{F}^{(\gamma)} G i + 1 ( γ ) F ( γ ) / G i ( γ ) F ( γ ) , 0 ≤ i < λ 0 \leq i < \lambda 0 ≤ i < λ , belonging to S γ \mathcal{S}_\gamma S γ . We will construct a filtration ( G α ( X ) F ) 0 ≤ α ≤ λ (G_\alpha^{(\mathcal{X})}\mathcal{F})_{0 \leq\alpha\leq\lambda} ( G α ( X ) F ) 0 ≤ α ≤ λ on the object F \mathcal{F} F of the exact category X \mathcal{X} X –Proqc f l \operatorname{Proqc}_{\mathrm{fl}} Proqc fl with the successive quotients G α + 1 ( X ) F / G α ( X ) F G_{\alpha+1}^{(\mathcal{X})}\mathcal{F}/G_\alpha^{(\mathcal{X})}\mathcal{F} G α + 1 ( X ) F / G α ( X ) F , 0 ≤ α < λ 0 \leq\alpha< \lambda 0 ≤ α < λ , belonging to S X \mathcal{S}_{\mathcal{X}} S X .
For this purpose, for each index γ ∈ Γ \gamma\in\Gamma γ ∈ Γ , we consider the complete lattice L γ ⊂ P ( λ ) L_\gamma\subset\mathcal{P}(\lambda) L γ ⊂ P ( λ ) and the homomorphism of complete lattices ℓ γ : L γ ⟶ Sub ( F ( γ ) ) \ell_\gamma:L_\gamma\longrightarrow\operatorname{Sub}(\mathcal{F}^{(\gamma)}) ℓ γ : L γ ⟶ Sub ( F ( γ ) ) provided by Theorem 5.1 for the filtration ( G i ( γ ) F ( γ ) ) 0 ≤ i ≤ λ (G_i^{(\gamma)}\mathcal{F}^{(\gamma)})_{0 \leq i \leq\lambda} ( G i ( γ ) F ( γ ) ) 0 ≤ i ≤ λ on the object F ( γ ) \mathcal{F}^{(\gamma)} F ( γ ) of the Grothendieck category X γ \mathcal{X}_\gamma X γ –Q c o h \mathrm{Qcoh} Qcoh . Recall that the category X γ \mathcal{X}_\gamma X γ –$\mathrm{Qcoh} is locally finitely presentable (hence locally κ \kappa κ -presentable) by [18 0.5.2.5 and Corollaire I.6.9.12] or [21 Definition Tag 01BN and Lemma Tag 01PJ or 01PK] , since X γ \mathcal{X}_\gamma X γ is a concentrated scheme by assumption. Even more specifically, for every ordinal $0 \leq \alpha \leq \lambda$ and every index γ ∈ Γ \gamma\in\Gamma γ ∈ Γ , we will construct a subset D α γ ⊂ λ D_\alpha^\gamma\subset\lambda D α γ ⊂ λ , D α γ ∈ L γ D_\alpha^\gamma\in L_\gamma D α γ ∈ L γ , for which we will have ( G α ( X ) F ) ( γ ) = ℓ γ ( D α γ ) ⊂ F ( γ ) (G_\alpha^{(\mathcal{X})}\mathcal{F})^{(\gamma)}=\ell_\gamma(D_\alpha^\gamma) \subset\mathcal{F}^{(\gamma)} ( G α ( X ) F ) ( γ ) = ℓ γ ( D α γ ) ⊂ F ( γ ) and D α γ ⊂ D β γ D_\alpha^\gamma\subset D_\beta^\gamma D α γ ⊂ D β γ for all 0 ≤ α ≤ β ≤ λ 0 \leq\alpha\leq\beta\leq\lambda 0 ≤ α ≤ β ≤ λ .
We proceed by transfinite induction on the ordinals 0 ≤ α ≤ λ 0 \leq\alpha\leq\lambda 0 ≤ α ≤ λ . For α = 0 \alpha=0 α = 0 , put D 0 γ = ∅ D_0^\gamma=\varnothing D 0 γ = ∅ for all γ ∈ Γ \gamma\in\Gamma γ ∈ Γ ; so G 0 ( X ) F = 0 G_0^{(\mathcal{X})}\mathcal{F}=0 G 0 ( X ) F = 0 . For a limit ordinal 0 < β ≤ λ 0<\beta\leq\lambda 0 < β ≤ λ , put D β γ = ⋃ α < β D α γ D_\beta^\gamma=\bigcup_{\alpha<\beta}D_\alpha^\gamma D β γ = ⋃ α < β D α γ for all γ ∈ Γ \gamma\in\Gamma γ ∈ Γ ; so G β ( X ) F = l i m → α < β G α ( X ) F G_\beta^{(\mathcal{X})}\mathcal{F}=\varinjlim_{\alpha<\beta}G_\alpha^{(\mathcal{X})}\mathcal{F} G β ( X ) F = lim α < β G α ( X ) F , as required. A nontrivial construction is needed in the case of a successor ordinal β = α + 1 ≤ λ \beta=\alpha+1 \leq\lambda β = α + 1 ≤ λ .
For all ordinals 0 ≤ α ≤ λ 0 \leq\alpha\leq\lambda 0 ≤ α ≤ λ and all indices γ ∈ Γ \gamma\in\Gamma γ ∈ Γ , we will have α ⊂ D α γ \alpha\subset D_\alpha^\gamma α ⊂ D α γ (where α \alpha α is viewed as a subset of λ \lambda λ ). Now, in order to pass from α \alpha α to β = α + 1 \beta=\alpha+1 β = α + 1 , we start with setting E 0 γ = D α γ ∪ { β } E_0^\gamma=D_\alpha^\gamma\cup\{\beta\} E 0 γ = D α γ ∪ { β } for all γ ∈ Γ \gamma\in\Gamma γ ∈ Γ . So the complement E 0 γ ∖ D α γ E_0^\gamma\setminus D_\alpha^\gamma E 0 γ ∖ D α γ is either empty or a singleton set.
Let μ \mu μ be the cardinality of Γ \Gamma Γ (or the countable cardinality if Γ \Gamma Γ is finite). So we have μ < κ \mu<\kappa μ < κ . Pick a map of sets ψ : μ ⟶ Γ \psi:\mu\longrightarrow\Gamma ψ : μ ⟶ Γ such that, for every γ ∈ Γ \gamma\in\Gamma γ ∈ Γ , the subset ψ − 1 ( γ ) ⊂ μ \psi^{-1}(\gamma)\subset\mu ψ − 1 ( γ ) ⊂ μ is cofinal in μ \mu μ . For example, it suffices to make ψ \psi ψ the first coordinate of a surjective map ( ψ , φ ) : μ ⟶ Γ × μ (\psi,\varphi):\mu\longrightarrow\Gamma\times\mu ( ψ , φ ) : μ ⟶ Γ × μ .
Proceeding by secondary transfinite induction on ordinals 1 ≤ j ≤ μ 1 \leq j \leq\mu 1 ≤ j ≤ μ , for every index γ ∈ Γ \gamma\in\Gamma γ ∈ Γ we will construct a subset E j γ ⊂ λ E_j^\gamma\subset\lambda E j γ ⊂ λ , E j γ ∈ L γ E_j^\gamma\in L_\gamma E j γ ∈ L γ . The following properties will be preserved throughout the construction of the sets E j γ E_j^\gamma E j γ . For all 0 ≤ j ≤ k ≤ μ 0 \leq j \leq k \leq\mu 0 ≤ j ≤ k ≤ μ and all γ ∈ Γ \gamma\in\Gamma γ ∈ Γ , we will have D α γ ⊂ E j γ ⊂ E k γ D_\alpha^\gamma\subset E_j^\gamma\subset E_k^\gamma D α γ ⊂ E j γ ⊂ E k γ , and the cardinality of the complement E j γ ∖ D α γ E_j^\gamma\setminus D_\alpha^\gamma E j γ ∖ D α γ will be smaller than κ \kappa κ .
For a limit ordinal 0 < k ≤ μ 0<k\leq\mu 0 < k ≤ μ , put E k γ = ⋃ j < k E j γ E_k^\gamma=\bigcup_{j<k}E_j^\gamma E k γ = ⋃ j < k E j γ for every γ ∈ Γ \gamma\in\Gamma γ ∈ Γ . Once again, a substantial construction is needed in the case of a successor ordinal k = j + 1 < μ k=j+1<\mu k = j + 1 < μ .
Put H j γ = ℓ γ ( E j γ ) ⊂ F ( γ ) \mathcal{H}_j^\gamma=\ell_\gamma(E_j^\gamma)\subset\mathcal{F}^{(\gamma)} H j γ = ℓ γ ( E j γ ) ⊂ F ( γ ) for every γ ∈ Γ \gamma\in\Gamma γ ∈ Γ , and γ j = ψ ( j ) \gamma_j=\psi(j) γ j = ψ ( j ) . By assertion (H3) of Theorem 5.1, the quasi-coherent sheaves H j γ \mathcal{H}_j^\gamma H j γ and F ( γ ) / H j γ \mathcal{F}^{(\gamma)}/\mathcal{H}_j^\gamma F ( γ ) / H j γ on X γ \mathcal{X}_\gamma X γ are flat for all γ ∈ Γ ( s i n c e s o a r e t h e q u a s i − c o h e r e n t s h e a v e s \gamma\in\Gamma(since so are the quasi-coherent sheaves γ ∈ Γ ( s in ceso a r e t h e q u a s i − co h er e n t s h e a v es G_{i+1}^{(\gamma)}\mathcal{F}^{(\gamma)}/G_i^{(\gamma)}\mathcal{F}^{(\gamma)}f o r a l l $ 0 ≤ i < λ for all \$0\leq i<\lambda f or a l l $0 ≤ i < λ ).
Furthermore, we have ( G α ( X ) F ) ( γ ) ⊂ H j γ (G_\alpha^{(\mathcal{X})}\mathcal{F})^{(\gamma)}\subset\mathcal{H}_j^\gamma ( G α ( X ) F ) ( γ ) ⊂ H j γ (since D α γ ⊂ E j γ D_\alpha^\gamma\subset E_j^\gamma D α γ ⊂ E j γ ). By virtue of the same assertion (H3), we also know that the quasi-coherent sheaf H j γ / ( ( G α ( X ) F ) ( γ ) ) \mathcal{H}_j^\gamma/((G_\alpha^{(\mathcal{X})}\mathcal{F})^{(\gamma)}) H j γ / (( G α ( X ) F ) ( γ ) ) on X γ \mathcal{X}_\gamma X γ is (locally) κ \kappa κ -presentable, since the cardinality of E j γ ∖ D α γ E_j^\gamma\setminus D_\alpha^\gamma E j γ ∖ D α γ is smaller than κ \kappa κ . Given an index γ ∈ Γ \gamma\in\Gamma γ ∈ Γ , we consider four cases. Case I: γ = γ j \gamma=\gamma_j γ = γ j . In this case, we put E k γ j = E j γ j E_k^{\gamma_j}=E_j^{\gamma_j} E k γ j = E j γ j . Case II: the elements γ \gamma γ and γ j \gamma_j γ j are incomparable in Γ \Gamma Γ (γ ≯ γ j \gamma\not> \gamma_j γ > γ j and γ ≮ γ j \gamma\not< \gamma_j γ < γ j ). In this case, we also put E k γ = E j γ E_k^\gamma= E_j^\gamma E k γ = E j γ .
Case III: γ < γ j \gamma< \gamma_j γ < γ j . So we have the transition morphism of schemes i γ , γ j : X γ ⟶ X γ j i_{\gamma,\gamma_j}: X_\gamma\longrightarrow X_{\gamma_j} i γ , γ j : X γ ⟶ X γ j . Applying the inverse image functor i γ , γ j ∗ : X γ –Qcoh ⟶ X γ j –Qcoh i_{\gamma,\gamma_j}^*: X_\gamma\text{--Qcoh} \longrightarrow X_{\gamma_j}\text{--Qcoh} i γ , γ j ∗ : X γ –Qcoh ⟶ X γ j –Qcoh , we obtain a quasi-coherent subsheaf i γ , γ j ∗ H j γ j ⊂ F ( γ ) i_{\gamma,\gamma_j}^*\mathcal{H}_j^{\gamma_j} \subset\mathcal{F}^{(\gamma)} i γ , γ j ∗ H j γ j ⊂ F ( γ ) .
Furthermore, G α ( X ) F G_\alpha^{(\mathcal{X})}\mathcal{F} G α ( X ) F is a flat pro-quasi-coherent pro-sheaf on X \mathcal{X} X (and an admissible subobject of F \mathcal{F} F ) by the assumption of induction on α \alpha α . So we have i γ , γ j ∗ ( ( G α ( X ) F ) ( γ j ) ) = ( G α ( X ) F ) ( γ ) i_{\gamma,\gamma_j}^*((G_\alpha^{(\mathcal{X})}\mathcal{F})^{(\gamma_j)}) = (G_\alpha^{(\mathcal{X})}\mathcal{F})^{(\gamma)} i γ , γ j ∗ (( G α ( X ) F ) ( γ j ) ) = ( G α ( X ) F ) ( γ ) . It follows that
( i γ , γ j ∗ H j γ j ) / ( G α ( X ) F ) ( γ ) ≃ i γ , γ j ∗ ( H j γ j / ( ( G α ( X ) F ) ( γ j ) ) ) \left(i_{\gamma,\gamma_j}^*\mathcal{H}_j^{\gamma_j}\right)/(G_\alpha^{(\mathcal{X})}\mathcal{F})^{(\gamma)} \simeq i_{\gamma,\gamma_j}^*\left(\mathcal{H}_j^{\gamma_j}/((G_\alpha^{(\mathcal{X})}\mathcal{F})^{(\gamma_j)})\right) ( i γ , γ j ∗ H j γ j ) / ( G α ( X ) F ) ( γ ) ≃ i γ , γ j ∗ ( H j γ j / (( G α ( X ) F ) ( γ j ) ) )
is a (locally) κ \kappa κ -presentable quasi-coherent sheaf on X γ X_\gamma X γ . In particular, the latter quasi-coherent quotient sheaf is κ \kappa κ -generated as an object of X γ X_\gamma X γ --Qcoh.
By Lemma 5.12(a), it follows that there exists a κ \kappa κ -generated quasi-coherent subsheaf K ⊂ i γ , γ j ∗ H j γ j \mathcal{K} \subset i_{\gamma,\gamma_j}^*\mathcal{H}_j^{\gamma_j} K ⊂ i γ , γ j ∗ H j γ j such that i γ , γ j ∗ H j γ j = ( G α ( X ) F ) ( γ ) + K i_{\gamma,\gamma_j}^*\mathcal{H}_j^{\gamma_j} = (G_\alpha^{(\mathcal{X})}\mathcal{F})^{(\gamma)} + \mathcal{K} i γ , γ j ∗ H j γ j = ( G α ( X ) F ) ( γ ) + K . Applying statement (H4) of Theorem 5.1 to the κ \kappa κ -generated subobject K ⊂ F ( γ ) \mathcal{K} \subset\mathcal{F}^{(\gamma)} K ⊂ F ( γ ) , we obtain a subset C ⊂ λ C \subset\lambda C ⊂ λ , C ∈ L γ C \in L_\gamma C ∈ L γ , with the cardinality of C C C smaller than κ \kappa κ , such that K ⊂ ℓ γ ( C ) ⊂ F ( γ ) \mathcal{K} \subset\ell_\gamma(C) \subset\mathcal{F}^{(\gamma)} K ⊂ ℓ γ ( C ) ⊂ F ( γ ) .
Now we put E k γ = E j γ ∪ C E_k^\gamma= E_j^\gamma\cup C E k γ = E j γ ∪ C . Setting H k γ = ℓ γ ( E k γ ) ⊂ F ( γ ) \mathcal{H}_k^\gamma= \ell_\gamma(E_k^\gamma) \subset\mathcal{F}^{(\gamma)} H k γ = ℓ γ ( E k γ ) ⊂ F ( γ ) , we have i γ , γ j ∗ H j γ j ⊂ H k γ i_{\gamma,\gamma_j}^*\mathcal{H}_j^{\gamma_j} \subset\mathcal{H}_k^\gamma i γ , γ j ∗ H j γ j ⊂ H k γ .
Case IV: γ > γ j \gamma> \gamma_j γ > γ j . So we have the transition morphism of schemes i γ j , γ : X γ j ⟶ X γ i_{\gamma_j,\gamma}: X_{\gamma_j} \longrightarrow X_\gamma i γ j , γ : X γ j ⟶ X γ . Once again, we have i γ j , γ ∗ ( ( G α ( X ) F ) ( γ ) ) = ( G α ( X ) F ) ( γ j ) i_{\gamma_j,\gamma}^*((G_\alpha^{(\mathcal{X})}\mathcal{F})^{(\gamma)}) = (G_\alpha^{(\mathcal{X})}\mathcal{F})^{(\gamma_j)} i γ j , γ ∗ (( G α ( X ) F ) ( γ ) ) = ( G α ( X ) F ) ( γ j ) , hence i γ j , γ ∗ ( F ( γ ) / ( G α ( X ) F ) ( γ ) ) = F ( γ j ) / ( G α ( X ) F ) ( γ j ) i_{\gamma_j,\gamma}^*(\mathcal{F}^{(\gamma)}/(G_\alpha^{(\mathcal{X})}\mathcal{F})^{(\gamma)}) = \mathcal{F}^{(\gamma_j)}/(G_\alpha^{(\mathcal{X})}\mathcal{F})^{(\gamma_j)} i γ j , γ ∗ ( F ( γ ) / ( G α ( X ) F ) ( γ ) ) = F ( γ j ) / ( G α ( X ) F ) ( γ j ) . The quasi-coherent sheaf H j γ j / ( ( G α ( X ) F ) ( γ j ) ) \mathcal{H}_j^{\gamma_j}/((G_\alpha^{(\mathcal{X})}\mathcal{F})^{(\gamma_j)}) H j γ j / (( G α ( X ) F ) ( γ j ) ) on X γ j X_{\gamma_j} X γ j is (locally) κ \kappa κ -presentable, hence κ \kappa κ -generated as an object of X γ j X_{\gamma_j} X γ j --Qcoh.
By Lemma 5.12(b), it follows that there exists a κ \kappa κ -generated quasi-coherent subsheaf K ⊂ F ( γ ) \mathcal{K} \subset\mathcal{F}^{(\gamma)} K ⊂ F ( γ ) for which the image of the induced morphism i γ j , γ ∗ K ⟶ F ( γ j ) / ( ( G α ( X ) F ) ( γ j ) ) i_{\gamma_j,\gamma}^*\mathcal{K} \longrightarrow\mathcal{F}^{(\gamma_j)}/((G_\alpha^{(\mathcal{X})}\mathcal{F})^{(\gamma_j)}) i γ j , γ ∗ K ⟶ F ( γ j ) / (( G α ( X ) F ) ( γ j ) ) in X γ j X_{\gamma_j} X γ j --Qcoh contains H j γ j / ( ( G α ( X ) F ) ( γ j ) ) \mathcal{H}_j^{\gamma_j}/((G_\alpha^{(\mathcal{X})}\mathcal{F})^{(\gamma_j)}) H j γ j / (( G α ( X ) F ) ( γ j ) ) . Once again, we apply statement (H4) of Theorem 5.1 to the κ \kappa κ -generated subobject K ⊂ F ( γ ) \mathcal{K} \subset\mathcal{F}^{(\gamma)} K ⊂ F ( γ ) and obtain a subset C ⊂ λ C \subset\lambda C ⊂ λ , C ∈ L γ C \in L_\gamma C ∈ L γ , with the cardinality of C C C smaller than κ \kappa κ , such that K ⊂ ℓ γ ( C ) ⊂ F ( γ ) \mathcal{K} \subset\ell_\gamma(C) \subset\mathcal{F}^{(\gamma)} K ⊂ ℓ γ ( C ) ⊂ F ( γ ) .
Once again, we put E k γ = E j γ ∪ C E_k^\gamma= E_j^\gamma\cup C E k γ = E j γ ∪ C . Setting H k γ = ℓ γ ( E k γ ) ⊂ F ( γ ) \mathcal{H}_k^\gamma= \ell_\gamma(E_k^\gamma) \subset\mathcal{F}^{(\gamma)} H k γ = ℓ γ ( E k γ ) ⊂ F ( γ ) , we have H j γ j ⊂ i γ j , γ ∗ H k γ ⊂ F ( γ j ) \mathcal{H}_j^{\gamma_j} \subset i_{\gamma_j,\gamma}^*\mathcal{H}_k^\gamma\subset\mathcal{F}^{(\gamma_j)} H j γ j ⊂ i γ j , γ ∗ H k γ ⊂ F ( γ j ) .
The construction of the subsets E j γ ⊂ λ E_j^\gamma\subset\lambda E j γ ⊂ λ , E j γ ∈ L γ E_j^\gamma\in L_\gamma E j γ ∈ L γ , for all ordinals 0 ≤ j ≤ μ 0 \leq j \leq\mu 0 ≤ j ≤ μ and all indices γ ∈ Γ \gamma\in\Gamma γ ∈ Γ , is finished. The cardinality of the complement E μ γ ∖ D α γ E_\mu^\gamma\setminus D_\alpha^\gamma E μ γ ∖ D α γ always remains smaller than κ \kappa κ , since the union of at most μ \mu μ sets of cardinality smaller than κ \kappa κ has cardinality smaller than κ \kappa κ .
Now we put F γ = ℓ γ ( E μ γ ) ⊂ F ( γ ) \mathcal{F}^\gamma= \ell_\gamma(E_\mu^\gamma) \subset\mathcal{F}^{(\gamma)} F γ = ℓ γ ( E μ γ ) ⊂ F ( γ ) for all γ ∈ Γ \gamma\in\Gamma γ ∈ Γ . So we have F γ = l i m → j < μ H j γ \mathcal{F}^\gamma= \varinjlim_{j<\mu}\mathcal{H}_j^\gamma F γ = lim j < μ H j γ . Let us show that, for every pair of indices δ < ϵ ∈ Γ \delta< \epsilon\in\Gamma δ < ϵ ∈ Γ , we have i δ ϵ ∗ F ϵ = F δ ⊂ F ( δ ) i_{\delta\epsilon}^*\mathcal{F}^\epsilon= \mathcal{F}^\delta\subset\mathcal{F}^{(\delta)} i δ ϵ ∗ F ϵ = F δ ⊂ F ( δ ) .
Indeed, in order to prove the inclusion i δ ϵ ∗ F ϵ ⊂ F δ i_{\delta\epsilon}^*\mathcal{F}^\epsilon\subset\mathcal{F}^\delta i δ ϵ ∗ F ϵ ⊂ F δ , consider the cofinal set of ordinals 0 ≤ j < μ 0 \leq j < \mu 0 ≤ j < μ for which ψ ( j ) = ϵ \psi(j) = \epsilon ψ ( j ) = ϵ . So we have γ j = ψ ( j ) = ϵ \gamma_j = \psi(j) = \epsilon γ j = ψ ( j ) = ϵ ; put k = j + 1 k = j + 1 k = j + 1 and γ = δ \gamma= \delta γ = δ . According to Case III above, we have i δ ϵ ∗ H j ϵ ⊂ H k δ i_{\delta\epsilon}^*\mathcal{H}_j^\epsilon\subset\mathcal{H}_k^\delta i δ ϵ ∗ H j ϵ ⊂ H k δ . Passing to the directed colimit over the ordinals 0 ≤ j < μ 0 \leq j < \mu 0 ≤ j < μ with ψ ( j ) = ϵ \psi(j) = \epsilon ψ ( j ) = ϵ , we conclude that i δ ϵ ∗ F ϵ ⊂ F δ i_{\delta\epsilon}^*\mathcal{F}^\epsilon\subset\mathcal{F}^\delta i δ ϵ ∗ F ϵ ⊂ F δ .
In order to prove the converse inclusion F δ ⊂ i δ ε ∗ F ε \mathcal{F}^{\delta} \subset i^{*}_{\delta\varepsilon}\mathcal{F}^{\varepsilon} F δ ⊂ i δ ε ∗ F ε , consider the cofinal set of ordinals 0 ≤ j < μ 0 \le j < \mu 0 ≤ j < μ for which ψ ( j ) = δ \psi(j)=\delta ψ ( j ) = δ . So we have γ j = ψ ( j ) = δ \gamma_j=\psi(j)=\delta γ j = ψ ( j ) = δ ; put k = j + 1 k=j+1 k = j + 1 and γ = ε \gamma=\varepsilon γ = ε . According to Case IV above, we have H j δ ⊂ i δ ε ∗ H k ε \mathcal{H}^{\delta}_{j} \subset i^{*}_{\delta\varepsilon}\mathcal{H}^{\varepsilon}_{k} H j δ ⊂ i δ ε ∗ H k ε . Passing to the directed colimit over the ordinals 0 ≤ j < μ 0 \le j < \mu 0 ≤ j < μ with ψ ( j ) = δ \psi(j)=\delta ψ ( j ) = δ , we conclude that F δ ⊂ i δ ε ∗ F ε \mathcal{F}^{\delta} \subset i^{*}_{\delta\varepsilon}\mathcal{F}^{\varepsilon} F δ ⊂ i δ ε ∗ F ε .
It remains to put D β γ = E μ γ D^{\gamma}_{\beta}=E^{\gamma}_{\mu} D β γ = E μ γ for all γ ∈ Γ \gamma\in\Gamma γ ∈ Γ . According to the argument above, the quasi-coherent subsheaves F γ = ℓ γ ( D β γ ) ⊂ F ( γ ) \mathcal{F}^{\gamma}=\ell_{\gamma}(D^{\gamma}_{\beta}) \subset\mathcal{F}(\gamma) F γ = ℓ γ ( D β γ ) ⊂ F ( γ ) , taken together for all indices γ ∈ Γ \gamma\in\Gamma γ ∈ Γ , form an admissible subobject G β ( x ) F ⊂ F G^{(x)}_{\beta}\mathcal{F} \subset\mathcal{F} G β ( x ) F ⊂ F in the exact category X \mathsf{X} X –Proqc f l \operatorname{Proqc}_{\mathrm{fl}} Proqc fl . So we have ( G β ( x ) F ) ( γ ) = F γ (G^{(x)}_{\beta}\mathcal{F})(\gamma)=\mathcal{F}^{\gamma} ( G β ( x ) F ) ( γ ) = F γ for all γ ∈ Γ \gamma\in\Gamma γ ∈ Γ .
The admissible quotient object G β ( x ) F / G α ( x ) F G^{(x)}_{\beta}\mathcal{F}/G^{(x)}_{\alpha}\mathcal{F} G β ( x ) F / G α ( x ) F belongs to S X \mathcal{S}_{\mathsf{X}} S X , since the cardinalities of the sets D β γ ∖ D α γ D^{\gamma}_{\beta}\setminus D^{\gamma}_{\alpha} D β γ ∖ D α γ are smaller than κ \kappa κ for all γ ∈ Γ \gamma\in\Gamma γ ∈ Γ . Finally, we have α ⊂ D α γ ⊂ λ \alpha\subset D^{\gamma}_{\alpha} \subset\lambda α ⊂ D α γ ⊂ λ for all ordinals 0 ≤ α ≤ λ 0 \le\alpha\le\lambda 0 ≤ α ≤ λ and indices γ ∈ Γ \gamma\in\Gamma γ ∈ Γ by construction; so D λ γ = λ D^{\gamma}_{\lambda}=\lambda D λ γ = λ , hence ℓ γ ( D λ γ ) = F ( γ ) \ell_{\gamma}(D^{\gamma}_{\lambda})=\mathcal{F}(\gamma) ℓ γ ( D λ γ ) = F ( γ ) and G λ ( x ) F = F G^{(x)}_{\lambda}\mathcal{F}=\mathcal{F} G λ ( x ) F = F .
Corollary 5.14. Let κ ≥ ℵ 1 \kappa\ge\aleph_{1} κ ≥ ℵ 1 be an uncountable regular cardinal and X = “ l i m → γ ∈ Γ X γ ” \mathsf{X}=\text{“}\varinjlim_{\gamma\in\Gamma}X_{\gamma}\text{”} X = “ lim γ ∈ Γ X γ ” be an ind-scheme represented by a directed diagram of concentrated schemes indexed by a directed poset Γ \Gamma Γ of cardinality smaller than κ \kappa κ . In this setting:
(a) Assume further that, for every index γ ∈ Γ \gamma\in\Gamma γ ∈ Γ and every affine open subscheme U ⊂ X γ U \subset X_{\gamma} U ⊂ X γ , the cardinality of the ring O X γ ( U ) \mathcal{O}_{X_{\gamma}}(U) O X γ ( U ) is smaller than κ \kappa κ . Then the exact category of flat pro-quasi-coherent pro-sheaves X \mathsf{X} X –Proqc f l \operatorname{Proqc}_{\mathrm{fl}} Proqc fl is κ \kappa κ -deconstructible in itself (i. e., as a class of objects in X \mathsf{X} X –Proqc f l \operatorname{Proqc}_{\mathrm{fl}} Proqc fl ).
(b) The class of all very flat pro-quasi-coherent pro-sheaves X \mathsf{X} X –Proqc v f l \operatorname{Proqc}_{\mathrm{vfl}} Proqc vfl is κ \kappa κ -deconstructible in the exact category X \mathsf{X} X –Proqc f l \operatorname{Proqc}_{\mathrm{fl}} Proqc fl .
Proof. Compare Corollary 5.9 with Theorem 5.13.
Deconstructibility for complexes
The following theorem is the second main result of Section 5.
Theorem 5.15. Let κ ≥ ℵ 1 \kappa\ge\aleph_{1} κ ≥ ℵ 1 be an uncountable regular cardinal and X = “ l i m → γ ∈ Γ X γ ” \mathsf{X}=\text{“}\varinjlim_{\gamma\in\Gamma}X_{\gamma}\text{”} X = “ lim γ ∈ Γ X γ ” be an ind-scheme represented by a directed diagram of concentrated schemes indexed by a directed poset Γ \Gamma Γ of cardinality smaller than κ \kappa κ . Suppose that we are given, for every index γ ∈ Γ \gamma\in\Gamma γ ∈ Γ , a class of complexes of flat quasi-coherent sheaves F γ ⊂ Com ( X γ – Qcoh f l ) \mathcal{F}_{\gamma} \subset\operatorname{Com}(X_{\gamma}\text{–}\operatorname{Qcoh}_{\mathrm{fl}}) F γ ⊂ Com ( X γ – Qcoh fl ) on X γ X_{\gamma} X γ such that the class F γ \mathcal{F}_{\gamma} F γ is κ \kappa κ -deconstructible in Com ( X γ – Qcoh ) \operatorname{Com}(X_{\gamma}\text{–}\operatorname{Qcoh}) Com ( X γ – Qcoh ) , or equivalently, in Com ( X γ – Qcoh f l ) \operatorname{Com}(X_{\gamma}\text{–}\operatorname{Qcoh}_{\mathrm{fl}}) Com ( X γ – Qcoh fl ) . Denote by F X ⊂ Com ( X – Proqc f l ) \mathcal{F}_{\mathsf{X}} \subset\operatorname{Com}(\mathsf{X}\text{–}\operatorname{Proqc}_{\mathrm{fl}}) F X ⊂ Com ( X – Proqc fl ) the class of all complexes of flat pro-quasi-coherent pro-sheaves F ∙ \mathcal{F}_{\bullet} F ∙ on X \mathsf{X} X such that F ∙ ( γ ) ∈ F γ \mathcal{F}_{\bullet}(\gamma) \in\mathcal{F}_{\gamma} F ∙ ( γ ) ∈ F γ for every index γ ∈ Γ \gamma\in\Gamma γ ∈ Γ . Then the class F X \mathcal{F}_{\mathsf{X}} F X is κ \kappa κ -deconstructible in Com ( X – Proqc f l ) \operatorname{Com}(\mathsf{X}\text{–}\operatorname{Proqc}_{\mathrm{fl}}) Com ( X – Proqc fl ) .
Proof. We refer to Theorem 4.12 for the underlying assertion that the category Com ( X – Proqc f l ) \operatorname{Com}(\mathsf{X}\text{–}\operatorname{Proqc}_{\mathrm{fl}}) Com ( X – Proqc fl ) is κ \kappa κ -accessible, and for the description of the κ \kappa κ -presentable objects of Com ( X – Proqc f l ) \operatorname{Com}(\mathsf{X}\text{–}\operatorname{Proqc}_{\mathrm{fl}}) Com ( X – Proqc fl ) . The proof of the present theorem is completely similar to that of Theorem 5.13. One just needs to replace all the mentions of the Grothendieck categories X γ X_{\gamma} X γ –Qcoh \operatorname{Qcoh} Qcoh with the Grothendieck categories Com ( X γ \operatorname{Com}(X_{\gamma} Com ( X γ –Qcoh ) \operatorname{Qcoh}) Qcoh ) , and all the mentions of the exact subcategories X γ X_{\gamma} X γ –Qcoh f l ⊂ X γ \operatorname{Qcoh}_{\mathrm{fl}} \subset X_{\gamma} Qcoh fl ⊂ X γ –Qcoh \operatorname{Qcoh} Qcoh with the exact subcategories Com ( X γ \operatorname{Com}(X_{\gamma} Com ( X γ –Qcoh f l ) ⊂ Com ( X γ \operatorname{Qcoh}_{\mathrm{fl}}) \subset\operatorname{Com}(X_{\gamma} Qcoh fl ) ⊂ Com ( X γ –Qcoh ) \operatorname{Qcoh}) Qcoh ) . The Grothendieck category Com ( X γ \operatorname{Com}(X_{\gamma} Com ( X γ –Qcoh ) \operatorname{Qcoh}) Qcoh ) is actually locally finitely presentable by Lemma 4.7 (since X γ X_\gamma X γ is a concentrated scheme by assumption), hence also locally κ \kappa κ -presentable. □
Corollary 5.16. Let κ ≥ ℵ 1 \kappa\ge\aleph_1 κ ≥ ℵ 1 be an uncountable regular cardinal and X = l i m → γ ∈ Γ X γ \mathcal{X} = \varinjlim_{\gamma\in\Gamma} X_\gamma X = lim γ ∈ Γ X γ be an ind-scheme represented by a directed diagram of concentrated schemes indexed by a directed poset Γ \Gamma Γ of cardinality smaller than κ \kappa κ . In this setting:
(a) Assume further that, for every index γ ∈ Γ \gamma\in\Gamma γ ∈ Γ and every affine open subscheme U ⊂ X γ U \subset X_\gamma U ⊂ X γ , the cardinality of the ring O X γ ( U ) \mathcal{O}_{X_\gamma}(U) O X γ ( U ) is smaller than κ \kappa κ . Then the exact category of complexes of flat pro-quasi-coherent pro-sheaves Com ( X – Proqc f l ) \operatorname{Com}(\mathcal{X}\text{--}\operatorname{Proqc}_{\mathrm{fl}}) Com ( X – Proqc fl ) is κ \kappa κ -deconstructible in itself, i. e., as a class of objects in Com ( X – Proqc f l ) \operatorname{Com}(\mathcal{X}\text{--}\operatorname{Proqc}_{\mathrm{fl}}) Com ( X – Proqc fl ) .
(b) The class of all complexes of very flat pro-quasi-coherent pro-sheaves Com ( X – Proqc v f l ) \operatorname{Com}(\mathcal{X}\text{--}\operatorname{Proqc}_{\mathrm{vfl}}) Com ( X – Proqc vfl ) is κ \kappa κ -deconstructible in the exact category Com ( X – Proqc f l ) \operatorname{Com}(\mathcal{X}\text{--}\operatorname{Proqc}_{\mathrm{fl}}) Com ( X – Proqc fl ) .
Proof. Compare Corollary 5.10 with Theorem 5.15. □
Corollary 5.17. Let κ ≥ ℵ 1 \kappa\ge\aleph_1 κ ≥ ℵ 1 be an uncountable regular cardinal and X = l i m → γ ∈ Γ X γ \mathcal{X} = \varinjlim_{\gamma\in\Gamma} X_\gamma X = lim γ ∈ Γ X γ be an ind-scheme represented by a directed diagram of concentrated schemes indexed by a directed poset Γ \Gamma Γ of cardinality smaller than κ \kappa κ . In this setting:
(a) Assume further that, for every index γ ∈ Γ \gamma\in\Gamma γ ∈ Γ and every affine open subscheme U ⊂ X γ U \subset X_\gamma U ⊂ X γ , the cardinality of the ring O X γ ( U ) \mathcal{O}_{X_\gamma}(U) O X γ ( U ) is smaller than κ \kappa κ . Then the class Ac ( X – Proqc f l ) ⊂ Com ( X – Proqc f l ) \operatorname{Ac}(\mathcal{X}\text{--}\operatorname{Proqc}_{\mathrm{fl}}) \subset\operatorname{Com}(\mathcal{X}\text{--}\operatorname{Proqc}_{\mathrm{fl}}) Ac ( X – Proqc fl ) ⊂ Com ( X – Proqc fl ) of all acyclic complexes in the exact category X – Proqc f l \mathcal{X}\text{--}\operatorname{Proqc}_{\mathrm{fl}} X – Proqc fl of flat pro-quasi-coherent pro-sheaves on X \mathcal{X} X is κ \kappa κ -deconstructible in the exact category Com ( X – Proqc f l ) \operatorname{Com}(\mathcal{X}\text{--}\operatorname{Proqc}_{\mathrm{fl}}) Com ( X – Proqc fl ) .
(b) The class Ac ( X – Proqc v f l ) ⊂ Com ( X – Proqc f l ) \operatorname{Ac}(\mathcal{X}\text{--}\operatorname{Proqc}_{\mathrm{vfl}}) \subset\operatorname{Com}(\mathcal{X}\text{--}\operatorname{Proqc}_{\mathrm{fl}}) Ac ( X – Proqc vfl ) ⊂ Com ( X – Proqc fl ) of all acyclic complexes in the exact category X – Proqc v f l \mathcal{X}\text{--}\operatorname{Proqc}_{\mathrm{vfl}} X – Proqc vfl of very flat pro-quasi-coherent pro-sheaves on X \mathcal{X} X is κ \kappa κ -deconstructible in the exact category Com ( X – Proqc f l ) \operatorname{Com}(\mathcal{X}\text{--}\operatorname{Proqc}_{\mathrm{fl}}) Com ( X – Proqc fl ) .
Proof. Compare Corollary 5.11 with Theorem 5.15. □
The Ext1 ^{1} 1 -Orthogonality Argument on Ind-Schemes
In this section we recall and generalize an argument from [41 Section 8.5] proving Ext1 ^{1} 1 -orthogonality of some flat pro-quasi-coherent pro-sheaves F \mathcal{F} F and C \mathcal{C} C on a reasonable ℵ 0 \aleph_0 ℵ 0 -ind-scheme X \mathcal{X} X under certain conditions on the inverse images of F \mathcal{F} F and C \mathcal{C} C to closed subschemes of X \mathcal{X} X .
Background on the Ext notation
As a general rule, we denote the abelian groups of morphisms in an additive category K \mathrm{K} K by Hom K ( − , − ) \operatorname{Hom}_{\mathrm{K}}(-,-) Hom K ( − , − ) and the Yoneda Ext groups in an exact category E \mathrm{E} E by Ext E ∗ ( − , − ) \operatorname{Ext}^{*}_{\mathrm{E}}(-,-) Ext E ∗ ( − , − ) .
Given a scheme X X X , we denote by Hom X ( − , − ) \operatorname{Hom}_{X}(-,-) Hom X ( − , − ) the abelian groups of morphisms in the abelian category of quasi-coherent sheaves X X X –Qcoh \operatorname{Qcoh} Qcoh . Given an ind-scheme X \mathcal{X} X , we denote by Hom ( X ) ( − , − ) \operatorname{Hom}_{(\mathcal{X})}(-,-) Hom ( X ) ( − , − ) the abelian groups of morphisms in the exact category of flat pro-quasi-coherent pro-sheaves X \mathcal{X} X –Proqc f l \operatorname{Proqc}_{\mathrm{fl}} Proqc fl . Furthermore, we denote by Ext X ∗ ( − , − ) \operatorname{Ext}_{X}^{*}(-,-) Ext X ∗ ( − , − ) the Ext groups computed in the abelian category X X X –Qcoh. We denote by Ext ( X ) ∗ ( − , − ) \operatorname{Ext}_{(\mathfrak X)}^{*}(-,-) Ext ( X ) ∗ ( − , − ) the Ext groups computed in the exact category X \mathfrak X X –Procfl.
Notice that, generally speaking, there is no reason for the Ext groups computed in the exact category of flat quasi-coherent sheaves X X X –Qcohf l _{\mathrm{fl}} fl to agree with the Ext groups in the abelian category X X X –Qcoh. However, the groups (Hom = Ext0 ^{0} 0 and) Ext1 ^{1} 1 computed in X X X –Qcohf l _{\mathrm{fl}} fl and in X X X –Qcoh always agree, as X X X –Qcohf l _{\mathrm{fl}} fl is a full subcategory closed under extensions in X X X –Qcoh.
The full subcategory X X X –Qcohf l _{\mathrm{fl}} fl is also always closed under kernels of epimorphisms in X X X –Qcoh. On a quasi-compact semi-separated scheme X X X , any quasi-coherent sheaf is a quotient sheaf of a flat quasi-coherent sheaf [27 Section 2.4] , [28 Section 3.2] , [10 Lemma A.1] , [29 Lemma 4.1.1 or 4.1.8] . It follows that all the groups Ext ∗ \operatorname{Ext}^{*} Ext ∗ computed in X X X –Qcohf l _{\mathrm{fl}} fl agree with the ones computed in X X X –Qcoh in this case (see [23 Theorem 12.1(b)] , [22 Proposition 13.2.2(i)] , [29 Proposition A.2.1 or A.3.1(a)] , or [40 Lemma 6.3] ).
The general formulation of orthogonality
We start with restating a lemma from [41 Section 8.5] .
Lemma 6.1. Let i : Z ⟶ X i: Z \longrightarrow X i : Z ⟶ X be a closed immersion of schemes, and let I ⊂ O X \mathcal I \subset\mathcal O_{X} I ⊂ O X be the corresponding quasi-coherent sheaf of ideals on X X X . Then, for any flat quasi-coherent sheaf F \mathcal F F on X X X , there is a natural short exact sequence of quasi-coherent sheaves on X X X
0 ⟶ I ⊗ O X F ⟶ F ⟶ i ∗ i ∗ F ⟶ 0 , 0 \longrightarrow\mathcal I \otimes_{\mathcal O_{X}} \mathcal F \longrightarrow\mathcal F \longrightarrow i_{*}i^{*}\mathcal F \longrightarrow0, 0 ⟶ I ⊗ O X F ⟶ F ⟶ i ∗ i ∗ F ⟶ 0 , where F ⟶ i ∗ i ∗ F \mathcal F \longrightarrow i_{*}i^{*}\mathcal F F ⟶ i ∗ i ∗ F is the adjunction unit for the pair of adjoint functors of inverse and direct image i ∗ : X i^{*}: X i ∗ : X –Qcoh ⟶ Z \longrightarrow Z ⟶ Z –Qcoh and i ∗ : Z i_{*}: Z i ∗ : Z –Qcoh ⟶ X \longrightarrow X ⟶ X –Qcoh.
Proof. This is [41 Lemma 8.5.5] . □
Recall the shorthand notation G ( γ ) = G ( X γ ) ∈ X γ \mathcal G^{(\gamma)}=\mathcal G(X_{\gamma})\in X_{\gamma} G ( γ ) = G ( X γ ) ∈ X γ –Qcohf l _{\mathrm{fl}} fl for a flat pro-quasi-coherent pro-sheaf G \mathcal G G on an ind-scheme \mathfrak X=\text{“}\lim_{\longrightarrow}_{\gamma\in\Gamma}X_{\gamma} .
Theorem 6.2. Let \mathfrak X=\text{“}\lim_{\longrightarrow}_{\gamma\in\Gamma}X_{\gamma} be a strict ℵ 0 \aleph_{0} ℵ 0 -ind-scheme represented by a directed diagram of closed immersions of concentrated schemes indexed by the ordered set of positive integers Γ = Z ≥ 1 \Gamma=\mathbb Z_{\geq1} Γ = Z ≥ 1 . Let F \mathcal F F and C \mathcal C C be flat pro-quasi-coherent pro-sheaves on X \mathfrak X X . Assume that
(i) Ext X γ 1 ( F ( γ ) , C ( γ ) ) = 0 \operatorname{Ext}^{1}_{X_{\gamma}}(\mathcal F^{(\gamma)},\mathcal C^{(\gamma)})=0 Ext X γ 1 ( F ( γ ) , C ( γ ) ) = 0 for all γ ∈ Z ≥ 1 \gamma\in\mathbb Z_{\geq1} γ ∈ Z ≥ 1 ;
(ii) Ext X γ 1 ( F ( γ ) , I γ ⊗ O X γ C ( γ ) ) = 0 \operatorname{Ext}^{1}_{X_{\gamma}}(\mathcal F^{(\gamma)},\mathcal I_{\gamma}\otimes_{\mathcal O_{X_{\gamma}}\mathcal C^{(\gamma)}})=0 Ext X γ 1 ( F ( γ ) , I γ ⊗ O X γ C ( γ ) ) = 0 for all γ ∈ Z ≥ 2 \gamma\in\mathbb Z_{\geq2} γ ∈ Z ≥ 2 , where I γ ⊂ O X γ \mathcal I_{\gamma}\subset\mathcal O_{X_{\gamma}} I γ ⊂ O X γ denotes the quasi-coherent sheaf of ideals of the closed subscheme X γ − 1 ⊂ X γ X_{\gamma-1}\subset X_{\gamma} X γ − 1 ⊂ X γ for γ ≥ 2 \gamma\geq2 γ ≥ 2 .
Then Ext ( X ) 1 ( F , C ) = 0 \operatorname{Ext}^{1}_{(\mathfrak X)}(\mathcal F,\mathcal C)=0 Ext ( X ) 1 ( F , C ) = 0 .
Proof. This is a generalization of [41 Lemma 8.5.7] , and the proof is similar. Denote by i γ : X γ ⟶ X γ + 1 i_{\gamma}:X_{\gamma}\longrightarrow X_{\gamma+1} i γ : X γ ⟶ X γ + 1 , γ ∈ Z ≥ 1 \gamma\in\mathbb Z_{\geq1} γ ∈ Z ≥ 1 , the transition morphisms of schemes in the directed diagram ( X γ ) γ ∈ Γ (X_{\gamma})_{\gamma\in\Gamma} ( X γ ) γ ∈ Γ . We need to show that any (admissible) short exact sequence 0 ⟶ C ⟶ G ⟶ F ⟶ 0 0\longrightarrow\mathcal C\longrightarrow\mathcal G\longrightarrow\mathcal F\longrightarrow0 0 ⟶ C ⟶ G ⟶ F ⟶ 0 in X \mathfrak X X –Procfl splits. Denote by j : C ⟶ G j:\mathcal C\longrightarrow\mathcal G j : C ⟶ G the admissible monomorphism of flat pro-quasi-coherent pro-sheaves in our short exact sequence and by j ( γ ) : C ( γ ) ⟶ G ( γ ) j^{(\gamma)}:\mathcal{C}^{(\gamma)}\longrightarrow\mathcal{G}^{(\gamma)} j ( γ ) : C ( γ ) ⟶ G ( γ ) its inverse images to X γ ⊂ X X_\gamma\subset X X γ ⊂ X .
For any index γ ∈ Z ≥ 1 \gamma\in\mathbb{Z}_{\geq1} γ ∈ Z ≥ 1 , the short exact sequence 0 ⟶ C ( γ ) ⟶ G ( γ ) ⟶ F ( γ ) ⟶ 0 0\longrightarrow\mathcal{C}^{(\gamma)}\longrightarrow\mathcal{G}^{(\gamma)}\longrightarrow\mathcal{F}^{(\gamma)}\longrightarrow0 0 ⟶ C ( γ ) ⟶ G ( γ ) ⟶ F ( γ ) ⟶ 0 in X γ X_\gamma X γ –Qcohf l _{\mathrm{fl}} fl splits by condition (i). So we have a splitting morphism s γ : F ( γ ) ⟶ G ( γ ) s_\gamma:\mathcal{F}^{(\gamma)}\longrightarrow\mathcal{G}^{(\gamma)} s γ : F ( γ ) ⟶ G ( γ ) in X γ X_\gamma X γ –Qcohf l _{\mathrm{fl}} fl . To prove the theorem, we need to show that the splitting morphisms s γ s_\gamma s γ can be chosen in such a way that they are compatible as γ \gamma γ varies, in the sense that s γ = i γ ∗ ( s γ + 1 ) s_\gamma=i_\gamma^*(s_{\gamma+1}) s γ = i γ ∗ ( s γ + 1 ) for all γ ∈ Z ≥ 1 \gamma\in\mathbb{Z}_{\geq1} γ ∈ Z ≥ 1 .
Proceeding by induction on γ \gamma γ , we can assume that we have already chosen a splitting morphism s γ s_\gamma s γ , and we need to find a splitting morphism s γ + 1 s_{\gamma+1} s γ + 1 satisfying the equation s γ = i γ ∗ ( s γ + 1 ) s_\gamma=i_\gamma^*(s_{\gamma+1}) s γ = i γ ∗ ( s γ + 1 ) . Pick an arbitrary splitting morphism s γ + 1 ′ : F ( γ + 1 ) ⟶ G ( γ + 1 ) s'_{\gamma+1}:\mathcal{F}^{(\gamma+1)}\longrightarrow\mathcal{G}^{(\gamma+1)} s γ + 1 ′ : F ( γ + 1 ) ⟶ G ( γ + 1 ) for the split short exact sequence 0 ⟶ C ( γ + 1 ) ⟶ G ( γ + 1 ) ⟶ F ( γ + 1 ) ⟶ 0 0\longrightarrow\mathcal{C}^{(\gamma+1)}\longrightarrow\mathcal{G}^{(\gamma+1)}\longrightarrow\mathcal{F}^{(\gamma+1)}\longrightarrow0 0 ⟶ C ( γ + 1 ) ⟶ G ( γ + 1 ) ⟶ F ( γ + 1 ) ⟶ 0 in X γ + 1 X_{\gamma+1} X γ + 1 –Qcohf l _{\mathrm{fl}} fl . Then the difference s γ − i γ ∗ ( s γ + 1 ′ ) : F ( γ ) ⟶ G ( γ ) s_\gamma-i_\gamma^*(s'_{\gamma+1}):\mathcal{F}^{(\gamma)}\longrightarrow\mathcal{G}^{(\gamma)} s γ − i γ ∗ ( s γ + 1 ′ ) : F ( γ ) ⟶ G ( γ ) factorizes through the admissible monomorphism C ( γ ) ⟶ G ( γ ) \mathcal{C}^{(\gamma)}\longrightarrow\mathcal{G}^{(\gamma)} C ( γ ) ⟶ G ( γ ) , so we obtain a morphism t γ : F ( γ ) ⟶ C ( γ ) t_\gamma:\mathcal{F}^{(\gamma)}\longrightarrow\mathcal{C}^{(\gamma)} t γ : F ( γ ) ⟶ C ( γ ) in X γ X_\gamma X γ –Qcohf l _{\mathrm{fl}} fl . Specifically, we have s γ − i γ ∗ ( s γ + 1 ′ ) = j ( γ ) ∘ t γ s_\gamma-i_\gamma^*(s'_{\gamma+1})=j^{(\gamma)}\circ t_\gamma s γ − i γ ∗ ( s γ + 1 ′ ) = j ( γ ) ∘ t γ .
All we need to do is to lift the morphism t γ t_\gamma t γ to X γ + 1 X_{\gamma+1} X γ + 1 –Qcohf l _{\mathrm{fl}} fl , i. e., find a morphism t γ + 1 : F ( γ + 1 ) ⟶ C ( γ + 1 ) t_{\gamma+1}:\mathcal{F}^{(\gamma+1)}\longrightarrow\mathcal{C}^{(\gamma+1)} t γ + 1 : F ( γ + 1 ) ⟶ C ( γ + 1 ) such that t γ = i γ ∗ ( t γ + 1 ) t_\gamma=i_\gamma^*(t_{\gamma+1}) t γ = i γ ∗ ( t γ + 1 ) . After this is done, the desired splitting morphism s γ + 1 : F ( γ + 1 ) ⟶ G ( γ + 1 ) s_{\gamma+1}:\mathcal{F}^{(\gamma+1)}\longrightarrow\mathcal{G}^{(\gamma+1)} s γ + 1 : F ( γ + 1 ) ⟶ G ( γ + 1 ) compatible with s γ s_\gamma s γ will be constructed as the sum ot two morphisms s γ + 1 = s γ + 1 ′ + j ( γ + 1 ) ∘ t γ + 1 s_{\gamma+1}=s'_{\gamma+1}+j^{(\gamma+1)}\circ t_{\gamma+1} s γ + 1 = s γ + 1 ′ + j ( γ + 1 ) ∘ t γ + 1 .
The morphism t γ : i γ ∗ F ( γ + 1 ) ⟶ i γ ∗ C ( γ + 1 ) t_\gamma:i_\gamma^*\mathcal{F}^{(\gamma+1)}\longrightarrow i_\gamma^*\mathcal{C}^{(\gamma+1)} t γ : i γ ∗ F ( γ + 1 ) ⟶ i γ ∗ C ( γ + 1 ) in X γ X_\gamma X γ –Qcohf l _{\mathrm{fl}} fl corresponds by adjunction to a morphism t ~ γ : F ( γ + 1 ) ⟶ i γ ∗ i γ ∗ C ( γ + 1 ) \widetilde{t}_\gamma:\mathcal{F}^{(\gamma+1)}\longrightarrow i_{\gamma*}i_\gamma^*\mathcal{C}^{(\gamma+1)} t γ : F ( γ + 1 ) ⟶ i γ ∗ i γ ∗ C ( γ + 1 ) in X γ + 1 X_{\gamma+1} X γ + 1 –Qcohf l _{\mathrm{fl}} fl . Lifting t γ t_\gamma t γ to a morphism t γ + 1 : F ( γ + 1 ) ⟶ C ( γ + 1 ) t_{\gamma+1}:\mathcal{F}^{(\gamma+1)}\longrightarrow\mathcal{C}^{(\gamma+1)} t γ + 1 : F ( γ + 1 ) ⟶ C ( γ + 1 ) such that t γ = i γ ∗ ( t γ + 1 ) t_\gamma=i_\gamma^*(t_{\gamma+1}) t γ = i γ ∗ ( t γ + 1 ) is equivalent to lifting t ~ γ \widetilde{t}_\gamma t γ to a morphism t γ + 1 : F ( γ + 1 ) ⟶ C ( γ + 1 ) t_{\gamma+1}:\mathcal{F}^{(\gamma+1)}\longrightarrow\mathcal{C}^{(\gamma+1)} t γ + 1 : F ( γ + 1 ) ⟶ C ( γ + 1 ) such that the composition of t γ + 1 t_{\gamma+1} t γ + 1 with the adjunction morphism C ( γ + 1 ) ⟶ i γ ∗ i γ ∗ C ( γ + 1 ) \mathcal{C}^{(\gamma+1)}\longrightarrow i_{\gamma*}i_\gamma^*\mathcal{C}^{(\gamma+1)} C ( γ + 1 ) ⟶ i γ ∗ i γ ∗ C ( γ + 1 ) is equal to t ~ γ \widetilde{t}_\gamma t γ .
Finally, by Lemma 6.1 we have a short exact sequence 0 ⟶ I γ + 1 ⊗ O X γ + 1 C ( γ + 1 ) ⟶ C ( γ + 1 ) ⟶ i γ ∗ i γ ∗ C ( γ + 1 ) ⟶ 0 0\longrightarrow\mathcal{I}_{\gamma+1}\otimes_{\mathcal{O}_{X_{\gamma+1}}}\mathcal{C}^{(\gamma+1)}\longrightarrow\mathcal{C}^{(\gamma+1)}\longrightarrow i_{\gamma*}i_\gamma^*\mathcal{C}^{(\gamma+1)}\longrightarrow0 0 ⟶ I γ + 1 ⊗ O X γ + 1 C ( γ + 1 ) ⟶ C ( γ + 1 ) ⟶ i γ ∗ i γ ∗ C ( γ + 1 ) ⟶ 0 in X γ + 1 X_{\gamma+1} X γ + 1 –Qcoh. By condition (ii), the group Ext X γ + 1 1 ( F ( γ + 1 ) , I γ + 1 ⊗ O X γ + 1 C ( γ + 1 ) ) \operatorname{Ext}^{1}_{X_{\gamma+1}}(\mathcal{F}^{(\gamma+1)},\mathcal{I}_{\gamma+1}\otimes_{\mathcal{O}_{X_{\gamma+1}}}\mathcal{C}^{(\gamma+1)}) Ext X γ + 1 1 ( F ( γ + 1 ) , I γ + 1 ⊗ O X γ + 1 C ( γ + 1 ) ) vanishes. Hence any morphism F ( γ + 1 ) ⟶ i γ ∗ i γ ∗ C ( γ + 1 ) \mathcal{F}^{(\gamma+1)}\longrightarrow i_{\gamma*}i_\gamma^*\mathcal{C}^{(\gamma+1)} F ( γ + 1 ) ⟶ i γ ∗ i γ ∗ C ( γ + 1 ) in the abelian category X γ + 1 X_{\gamma+1} X γ + 1 –Qcoh can be lifted to a morphism F ( γ + 1 ) ⟶ C ( γ + 1 ) \mathcal{F}^{(\gamma+1)}\longrightarrow\mathcal{C}^{(\gamma+1)} F ( γ + 1 ) ⟶ C ( γ + 1 ) , and we are done. □
Background on contraadjusted and cotorsion quasi-coherent sheaves
This section is an extraction from [29 Section 4.1] .
Let X X X be a scheme. A quasi-coherent sheaf C \mathcal{C} C on X X X is said to be cotorsion [13 ] if Ext X 1 ( F , C ) = 0 \operatorname{Ext}^{1}_{X}(\mathcal{F},\mathcal{C})=0 Ext X 1 ( F , C ) = 0 for all flat quasi-coherent sheaves F \mathcal{F} F on X X X . Similarly, a quasi-coherent sheaf C \mathcal{C} C on X X X is said to be contraadjusted [29 Sections 2.5 and 4.1] if Ext X 1 ( F , C ) = 0 \operatorname{Ext}^{1}_{X}(\mathcal{F},\mathcal{C})=0 Ext X 1 ( F , C ) = 0 for all very flat quasi-coherent sheaves F \mathcal{F} F on X X X .
In particular, a quasi-coherent sheaf C \mathcal{C} C on an affine scheme U U U is cotorsion if and only if the O ( U ) \mathcal{O}(U) O ( U ) -module C ( U ) \mathcal{C}(U) C ( U ) is cotorsion. A quasi-coherent sheaf C \mathcal{C} C on U U U is contraadjusted if and only if the O ( U ) \mathcal{O}(U) O ( U ) -module C ( U ) \mathcal{C}(U) C ( U ) is contraadjusted. (See § 3.3 for the module-theoretic background.)
The classes of contraadjusted and cotorsion quasi-coherent sheaves on any scheme X X X are obviously closed under extensions and direct summands. Both the classes are preserved by the direct images with respect to quasi-compact quasi-separated morphisms of schemes [29 Section 2.5] . We denote the full subcategories of cotorsion and contraadjusted quasi-coherent sheaves by X - Qcoh c o t ⊂ X - Qcoh c t a ⊂ X - Qcoh X\text{-}\operatorname{Qcoh}^{\mathrm{cot}} \subset X\text{-}\operatorname{Qcoh}^{\mathrm{cta}} \subset X\text{-}\operatorname{Qcoh} X - Qcoh cot ⊂ X - Qcoh cta ⊂ X - Qcoh .
A more substantial theory of contraadjusted and cotorsion quasi-coherent sheaves can be developed for quasi-compact semi-separated schemes X X X [29 Section 4.1] . The key fact is that, on such a scheme X X X , any quasi-coherent sheaf M \mathcal{M} M is a quotient sheaf of a flat quasi-coherent sheaf F \mathcal{F} F [27 Section 2.4] , [28 Section 3.2] . Actually, one can even choose F \mathcal{F} F to be a very flat quasi-coherent sheaf [29 Lemma 4.1.1] . The explicit construction of such a flat/very flat quasi-coherent sheaf F \mathcal{F} F for a given quasi-coherent sheaf M \mathcal{M} M suggested in [10 Lemma A.1] and further developed in [29 Lemmas 4.1.1 and 4.1.8] plays a crucial role.
Proposition 6.3. Let X X X be a quasi-compact semi-separated scheme and M \mathcal{M} M be a quasi-coherent sheaf on X X X . Then
(a) there exists a short exact sequence 0 ⟶ C ⟶ F ⟶ M ⟶ 0 0 \longrightarrow\mathcal{C} \longrightarrow\mathcal{F} \longrightarrow\mathcal{M} \longrightarrow0 0 ⟶ C ⟶ F ⟶ M ⟶ 0 in X - Qcoh X\text{-}\operatorname{Qcoh} X - Qcoh with a flat quasi-coherent sheaf F \mathcal{F} F and a cotorsion quasi-coherent sheaf C \mathcal{C} C ;
(b) there exists a short exact sequence 0 ⟶ M ⟶ C ⟶ F ⟶ 0 0 \longrightarrow\mathcal{M} \longrightarrow\mathcal{C} \longrightarrow\mathcal{F} \longrightarrow0 0 ⟶ M ⟶ C ⟶ F ⟶ 0 in X - Qcoh X\text{-}\operatorname{Qcoh} X - Qcoh with a cotorsion quasi-coherent sheaf C \mathcal{C} C and a flat quasi-coherent sheaf F \mathcal{F} F .
Proof. Part (b) is “almost” a special case of [13 Corollary 4.2] in view of the Wakamatsu lemma [55 Lemma 2.1.2] . The facts that every object of the Grothendieck category X - Qcoh X\text{-}\operatorname{Qcoh} X - Qcoh is a subobject of an injective object, and any injective quasi-coherent sheaf is cotorsion, implying that any cotorsion (pre)envelope is a monomorphism, are used here. With this approach, one also needs to use the fact that the vanishing of Ext X 1 ( F , C ) \operatorname{Ext}^{1}_{X}(\mathcal{F},\mathcal{C}) Ext X 1 ( F , C ) for all cotorsion quasi-coherent sheaves C \mathcal{C} C implies flatness of a quasi-coherent sheaf F \mathcal{F} F . The latter fact is provable for quasi-compact semi-separated schemes X X X using existence of a flat generator in X - Qcoh X\text{-}\operatorname{Qcoh} X - Qcoh .
Part (a) follows from part (b) and the above-mentioned result from [27 Section 2.4] , [28 Section 3.2] , or [10 Lemma A.1] by virtue of the Salce lemma [47 ] , [29 Lemma B.1.1(a)] .
Alternatively, our preferred proof can be found in [29 Section 4.1] ; see [29 Corollary 4.1.11(a–b)] . □
Proposition 6.4. Let X X X be a quasi-compact semi-separated scheme and M \mathcal{M} M be a quasi-coherent sheaf on X X X . Then
(a) there exists a short exact sequence 0 ⟶ C ⟶ F ⟶ M ⟶ 0 0 \longrightarrow\mathcal{C} \longrightarrow\mathcal{F} \longrightarrow\mathcal{M} \longrightarrow0 0 ⟶ C ⟶ F ⟶ M ⟶ 0 in X - Qcoh X\text{-}\operatorname{Qcoh} X - Qcoh with a very flat quasi-coherent sheaf F \mathcal{F} F and a contraadjusted quasi-coherent sheaf C \mathcal{C} C ;
(b) there exists a short exact sequence 0 ⟶ M ⟶ C ⟶ F ⟶ 0 0 \longrightarrow\mathcal{M} \longrightarrow\mathcal{C} \longrightarrow\mathcal{F} \longrightarrow0 0 ⟶ M ⟶ C ⟶ F ⟶ 0 in X - Qcoh X\text{-}\operatorname{Qcoh} X - Qcoh with a contraadjusted quasi-coherent sheaf C \mathcal{C} C and a very flat quasi-coherent sheaf F \mathcal{F} F .
Proof. This is [29 Corollary 4.1.4(a–b)] . □
Corollary 6.5. Let X X X be a quasi-compact semi-separated scheme.
(a) Let F \mathcal{F} F be a flat quasi-coherent sheaf on X X X and C \mathcal{C} C be a cotorsion quasi-coherent sheaf on X X X . Then Ext X n ( F , C ) = 0 \operatorname{Ext}^{n}_{X}(\mathcal{F},\mathcal{C})=0 Ext X n ( F , C ) = 0 for all n ≥ 1 n \ge1 n ≥ 1 .
(b) The full subcategory X - Qcoh c o t X\text{-}\operatorname{Qcoh}^{\mathrm{cot}} X - Qcoh cot of cotorsion quasi-coherent sheaves is closed under cokernels of monomorphisms in X - Qcoh X\text{-}\operatorname{Qcoh} X - Qcoh .
Proof. This is [29 Corollary 4.1.9] or [41 Corollary 7.5.6] . The assertions follow from the facts that any quasi-coherent sheaf on X X X is a quotient sheaf of a flat quasi-coherent sheaf and the class of flat quasi-coherent sheaves is closed under kernels of epimorphisms in X X X –Qcoh. □
Corollary 6.6. Let X X X be a quasi-compact semi-separated scheme.
(a) Let F F F be a very flat quasi-coherent sheaf on X X X and C C C be a contraadjusted quasi-coherent sheaf on X X X . Then Ext X n ( F , C ) = 0 \operatorname{Ext}^{n}_{X}(F,C)=0 Ext X n ( F , C ) = 0 for all n ≥ 1 n\geq1 n ≥ 1 .
(b) The full subcategory X X X –Qcoh c t a \operatorname{Qcoh}^{\mathrm{cta}} Qcoh cta of contraadjusted quasi-coherent sheaves is closed under cokernels of monomorphisms in X X X –Qcoh.
Proof. This is [29 Corollary 4.1.2] or [41 Corollary 7.5.3] . The assertions follow from the facts that any quasi-coherent sheaf on X X X is a quotient sheaf of a very flat quasi-coherent sheaf (a weak form of Proposition 6.4(a)) and the class of very flat quasi-coherent sheaves is closed under kernels of epimorphisms in X X X –Qcoh. □
The following two propositions spell out antilocality properties of various classes of quasi-coherent sheaves, in the sense similar to [34 Section 4] .
Proposition 6.7. Let X X X be a quasi-compact semi-separated scheme with a finite affine open covering X = ⋃ α = 1 N U α X=\bigcup_{\alpha=1}^{N}U_{\alpha} X = ⋃ α = 1 N U α . In this context:
(a) a quasi-coherent sheaf C C C on X X X is cotorsion if and only if C C C is a direct summand of a finitely iterated extension of the direct images of cotorsion quasi-coherent sheaves from U α U_{\alpha} U α ;
(b) a quasi-coherent sheaf C C C on X X X is contraadjusted if and only if C C C is a direct summand of a finitely iterated extension of the direct images of contraadjusted quasi-coherent sheaves from U α U_{\alpha} U α .
Proof. Part (a) is [29 Corollary 4.1.11(c)] . Part (b) is [29 Corollary 4.1.4(c)] . □
Proposition 6.8. Let X X X be a quasi-compact semi-separated scheme with a finite affine open covering X = ⋃ α = 1 N U α X=\bigcup_{\alpha=1}^{N}U_{\alpha} X = ⋃ α = 1 N U α . In this context:
(a) a quasi-coherent sheaf C C C on X X X is flat and cotorsion if and only if C C C is a direct summand of a finite direct sum of the direct images of flat cotorsion quasi-coherent sheaves from U α U_{\alpha} U α ;
(b) a quasi-coherent sheaf C C C on X X X is flat and contraadjusted if and only if C C C is a direct summand of a finitely iterated extension of the direct images of flat contraadjusted quasi-coherent sheaves from U α U_{\alpha} U α ;
(c) a quasi-coherent sheaf C C C on X X X is very flat and contraadjusted if and only if C C C is a direct summand of a finite direct sum the direct images of very flat contraadjusted quasi-coherent sheaves from U α U_{\alpha} U α .
Proof. Part (a) is [29 Lemma 4.1.12] . Part (b) is [29 Lemma 4.1.6] . Part (c) is [29 Lemma 4.1.5] . □
Pro-sheaves with flat contraadjusted and flat cotorsion components
We start with two lemmas essentially from [41 Section 8.5] .
Lemma 6.9. Let i : Z ⟶ X i: Z \longrightarrow X i : Z ⟶ X be a closed immersion of quasi-compact semi-separated schemes and i ∗ : X – Q c o h ⟶ Z – Q c o h i^{*}\colon X\text{--}\mathrm{Qcoh} \longrightarrow Z\text{--}\mathrm{Qcoh} i ∗ : X – Qcoh ⟶ Z – Qcoh be the inverse image functor. In this context:
(a) the functor i ∗ i^{*} i ∗ takes flat contraadjusted quasi-coherent sheaves on X X X to flat contraadjusted quasi-coherent sheaves on Z Z Z ;
(b) if the schemes X X X and Z Z Z are Noetherian, then the functor i ∗ i^{*} i ∗ takes flat cotorsion quasi-coherent sheaves on X X X to flat cotorsion quasi-coherent sheaves on Z Z Z .
Proof. Part (a) is [41 Corollary 8.5.4(a)] . Part (b) is a particular case of [41 Corollary 8.5.4(b)] . The arguments are based on the results of Proposition 6.8(a–b) above and [29 Lemmas 1.7.7(b) and 1.7.9(a)] . □
The definition of a (locally) finitely generated quasi-coherent sheaf on a concentrated scheme was given in § 2.7 .
Lemma 6.10. Let X X X be a quasi-compact semi-separated scheme and M \mathcal{M} M be a locally finitely generated quasi-coherent sheaf on X X X . In this context:
(a) for any flat contraadjusted quasi-coherent sheaf C \mathcal{C} C on X X X , the quasi-coherent sheaf M ⊗ O X C \mathcal{M}\otimes_{\mathcal{O}_{X}}\mathcal{C} M ⊗ O X C on X X X is contraadjusted;
(b) if the scheme X X X is Noetherian then, for any flat cotorsion quasi-coherent sheaf C \mathcal{C} C on X X X , the quasi-coherent sheaf M ⊗ O X C \mathcal{M}\otimes_{\mathcal{O}_{X}}\mathcal{C} M ⊗ O X C on X X X is cotorsion.
Proof. Part (b) is [41 Corollary 8.5.3(b)] . Part (a) is a straightforward generalization of [41 Corollary 8.5.3(a)] , provable in the same way. The arguments are based on the results of Proposition 6.8(a–b) above and [29 Corollaries 1.7.4(a) and 1.7.5(a)] . □
Let X \mathfrak{X} X be a (strict ind-quasi-compact) ind-semi-separated ind-scheme. We will say that a flat pro-quasi-coherent pro-sheaf C \mathcal{C} C on X \mathfrak{X} X is levelwise contraadjusted if the flat quasi-coherent sheaf C ( Z ) \mathcal{C}^{(Z)} C ( Z ) on Z Z Z is contraadjusted for every closed subscheme Z ⊂ X Z\subset\mathfrak{X} Z ⊂ X .
Given a directed diagram of closed immersions of quasi-compact semi-separated schemes ( X γ ) γ ∈ Γ (X_{\gamma})_{\gamma\in\Gamma} ( X γ ) γ ∈ Γ representing X \mathfrak{X} X , it suffices to check the contraadjustedness condition from the previous paragraph for the closed subschemes Z = X γ ⊂ X Z=X_{\gamma}\subset\mathfrak{X} Z = X γ ⊂ X , γ ∈ Γ \gamma\in\Gamma γ ∈ Γ . This follows from Lemma 6.9(a).
Let X \mathfrak{X} X be a (strict) ind-semi-separated ind-Noetherian ind-scheme. We will say that a flat pro-quasi-coherent pro-sheaf C \mathcal{C} C on X \mathfrak{X} X is levelwise cotorsion if the flat quasi-coherent sheaf C ( Z ) \mathcal{C}^{(Z)} C ( Z ) on Z Z Z is cotorsion for every closed subscheme Z ⊂ X Z\subset\mathfrak{X} Z ⊂ X .
Given a directed diagram of closed immersions of semi-separated Noetherian schemes ( X γ ) γ ∈ Γ (X_{\gamma})_{\gamma\in\Gamma} ( X γ ) γ ∈ Γ representing X \mathfrak{X} X , it suffices to check the cotorsion condition from the previous paragraph for the closed subschemes Z = X γ ⊂ X Z=X_{\gamma}\subset\mathfrak{X} Z = X γ ⊂ X , γ ∈ Γ \gamma\in\Gamma γ ∈ Γ . This follows from Lemma 6.9(b).
The definition of a reasonable closed immersion of concentrated schemes, as well as that of a reasonable ind-scheme , was given in § 2.7 .
Corollary 6.11. (a) Let X \mathfrak{X} X be a (strict ind-quasi-compact) ind-semi-separated reasonable ℵ 0 \aleph_{0} ℵ 0 -ind-scheme. Then one has Ext X ( F , G ) = 0 \operatorname{Ext}_{\mathfrak{X}}(\mathcal{F},\mathcal{G})=0 Ext X ( F , G ) = 0 for any very flat pro-quasi-coherent pro-sheaf F \mathcal{F} F and any levelwise contraadjusted flat pro-quasi-coherent pro-sheaf G \mathcal{G} G on X \mathfrak{X} X . In particular, any levelwise contraadjusted very flat pro-quasi-coherent pro-sheaf on X \mathfrak{X} X is an injective object of the exact category X – Proqcoh f l \mathfrak{X}\text{--}\operatorname{Proqcoh}_{\mathrm{fl}} X – Proqcoh fl .
(b) Let X \mathcal{X} X be a (strict) ind-semi-separated ind-Noetherian ℵ 0 \aleph_0 ℵ 0 -ind-scheme. Then one has Ext X 1 ( F , C ) = 0 \operatorname{Ext}^{1}_{\mathcal{X}}(\mathcal{F},\mathcal{C}) = 0 Ext X 1 ( F , C ) = 0 for any flat pro-quasi-coherent pro-sheaf F \mathcal{F} F and any levelwise cotorsion flat pro-quasi-coherent pro-sheaf C \mathcal{C} C on X \mathcal{X} X . In other words, any levelwise cotorsion flat pro-quasi-coherent pro-sheaf on X \mathcal{X} X is an injective object of the exact category X \mathcal{X} X –Proqc f l \operatorname{Proqc}_{\mathrm{fl}} Proqc fl .
Proof. This is a generalization of [41 Lemma 8.5.7] from (adic Noetherian) formal schemes to ind-schemes. This is also a far-reaching partial generalization of the contramodule-theoretic Ext 1 \operatorname{Ext}^{1} Ext 1 -orthogonality results of [29 Lemma E.4.3 and Corollary E.5.2] ; cf. Examples 3.1 and 3.2 above.
Part (a): by Corollary 2.3(b) and the subsequent paragraph, together with Lemma 2.4(b), the ind-scheme X \mathcal{X} X can be represented by a countable directed diagram of reasonable closed immersions of quasi-compact semi-separated schemes. Now the assertion follows from Theorem 6.2, which is applicable in view of Lemma 6.10(a).
Part (b): by Corollary 2.3(b–c) and the subsequent paragraph, together with Lemma 2.4(b), the ind-scheme X \mathcal{X} X can be represented by a countable directed diagram of closed immersions of semi-separated Noetherian schemes. Now the assertion follows from Theorem 6.2, which is applicable in view of Lemma 6.10(b). □
Orthogonality for complexes
The following theorem is a straightforward extension of Theorem 6.2 from (flat pro-quasi-coherent) pro-sheaves to complexes of pro-sheaves. The termwise exact structure on complexes, as defined at the end of Section 5.2, is presumed in the Ext notation here and below.
Theorem 6.12. Let \mathcal{X} = \text{“}\lim_{\longrightarrow}\limits_{\gamma\in\Gamma} X_{\gamma}\text{”} be a strict ℵ 0 \aleph_0 ℵ 0 -ind-scheme represented by a directed diagram of closed immersions of concentrated schemes indexed by the ordered set of positive integers Γ = Z ≥ 1 \Gamma= \mathbb{Z}_{\ge1} Γ = Z ≥ 1 . Let F \mathcal{F}^{\mathbin{\vcenter{\text{}}}} F \scriptstyle\bullet[eq] and C \mathcal{C}^{\mathbin{\vcenter{\text{}}}} C \scriptstyle\bullet[eq] be two complexes of flat pro-quasi-coherent pro-sheaves on X \mathcal{X} X . Assume that
(i) Ext Com ( X γ – Qcoh ) 1 ( F \operatorname{Ext}^{1}_{\operatorname{Com}(X_{\gamma}\text{–}\operatorname{Qcoh})}(\mathcal{F}^{\mathbin{\vcenter{\text{}}}} Ext Com ( X γ – Qcoh ) 1 ( F \scriptstyle\bullet( γ ) , C (\gamma),\mathcal{C}^{\mathbin{\vcenter{\text{}}}} ( γ ) , C \scriptstyle\bullet( γ ) ) = 0 (\gamma)) = 0 ( γ )) = 0 for all γ ∈ Z ≥ 1 \gamma\in\mathbb{Z}_{\ge1} γ ∈ Z ≥ 1 ;
(ii) Ext Com ( X γ – Qcoh ) 1 ( F \operatorname{Ext}^{1}_{\operatorname{Com}(X_{\gamma}\text{–}\operatorname{Qcoh})}(\mathcal{F}^{\mathbin{\vcenter{\text{}}}} Ext Com ( X γ – Qcoh ) 1 ( F \scriptstyle\bullet( γ ) , I γ ⊗ O X γ C (\gamma),\mathcal{I}_{\gamma} \otimes_{\mathcal{O}_{X_{\gamma}}} \mathcal{C}^{\mathbin{\vcenter{\text{}}}} ( γ ) , I γ ⊗ O X γ C \scriptstyle\bullet( γ ) ) = 0 (\gamma)) = 0 ( γ )) = 0 for all γ ∈ Z ≥ 2 \gamma\in\mathbb{Z}_{\ge2} γ ∈ Z ≥ 2 , where I γ ⊂ O X γ \mathcal{I}_{\gamma} \subset\mathcal{O}_{X_{\gamma}} I γ ⊂ O X γ denotes the quasi-coherent sheaf of ideals of the closed subscheme X γ − 1 ⊂ X γ X_{\gamma-1} \subset X_{\gamma} X γ − 1 ⊂ X γ .
Then Ext Com ( X – Proqc f l ) 1 ( F \operatorname{Ext}^{1}_{\operatorname{Com}(\mathcal{X}\text{–}\operatorname{Proqc}_{\mathrm{fl}})}(\mathcal{F}^{\mathbin{\vcenter{\text{}}}} Ext Com ( X – Proqc fl ) 1 ( F \scriptstyle\bullet, C ,\mathcal{C}^{\mathbin{\vcenter{\text{}}}} , C \scriptstyle\bullet) = 0 ) = 0 ) = 0 .
Proof. The same argument as in the proof of Theorem 6.2 applies. One just needs to work with complexes of pro-sheaves and sheaves (short exact sequences of complexes, morphisms of complexes etc.) throughout, instead of single pro-sheaves and sheaves. □
We need another bit of recollection of known material.
Proposition 6.13. Let X X X be a quasi-compact semi-separated scheme. Then
(a) for any acyclic complex F \mathcal{F}^{\mathbin{\vcenter{\text{}}}} F \scriptstyle\bullet[eq] in the exact category of very flat quasi-coherent sheaves X X X –Qcoh f l \operatorname{Qcoh}_{\mathrm{fl}} Qcoh fl and any complex of contradjusted quasi-coherent sheaves C \mathcal{C}^{\mathbin{\vcenter{\text{}}}} C \scriptstyle\bullet[eq] on X X X , one has Ext Com ( X – Qcoh ) 1 ( F \operatorname{Ext}^{1}_{\operatorname{Com}(X\text{–}\operatorname{Qcoh})}(\mathcal{F}^{\mathbin{\vcenter{\text{}}}} Ext Com ( X – Qcoh ) 1 ( F \scriptstyle\bullet, C ,\mathcal{C}^{\mathbin{\vcenter{\text{}}}} , C \scriptstyle\bullet) = 0 ) = 0 ) = 0 ;
(b) for any acyclic complex F ∙ \mathcal{F}^{\bullet} F ∙ in the exact category of flat quasi-coherent sheaves X \mathsf{X} X –Qcoh f l \operatorname{Qcoh}_{\mathrm{fl}} Qcoh fl and any complex of cotorsion quasi-coherent sheaves C ∙ \mathcal{C}^{\bullet} C ∙ on X \mathsf{X} X , one has
Ext Com ( X – Qcoh ) 1 ( F ∙ , C ∙ ) = 0. \operatorname{Ext}^{1}_{\operatorname{Com}(\mathsf{X}\text{--}\operatorname{Qcoh})}(\mathcal{F}^{\bullet},\mathcal{C}^{\bullet})=0. Ext Com ( X – Qcoh ) 1 ( F ∙ , C ∙ ) = 0.
Proof. Part (b) is a scheme generalization of the theorem of Bazzoni, Cortés-Izurdiaga, and Estrada [4 Theorem 5.3] . The formulation can be found in [9 Theorem 3.3] . (Notice that the first paragraph of the proof of [9 Lemma 3.2] is erroneous, though the formulation of the lemma is true; this mistake in the proof in [9 ] was later corrected in [14 Theorem 6.3(2)] , see also [14 Remark 6.7] .) See [17 Lemma 2.1] or [29 Lemma B.1.7] for a background exact category-theoretic lemma about extensions of complexes.
Another reference is [29 Lemma 5.1.1(b)] . For a more general point of view, see [40 Proposition 2.1, Theorem 3.3 (2) $\Leftrightarrow$ (4), and Example 8.9] .
Part (a) is an easier version of part (b). A direct reference is [29 Lemma 5.1.1(a)] . For a more general point of view, see [40 Proposition 2.1, Theorem 3.3 (2) $\Leftrightarrow$ (4), Corollary 6.5, Lemma 6.7, and Example 6.12(1)] . □
Corollary 6.14. (a) Let X \mathsf{X} X be an ind -semi-separated reasonable ℵ 0 \aleph_{0} ℵ 0 -ind -scheme. Then one has Ext Com ( X – Proqc f l ) 1 ( F ∙ , C ∙ ) = 0 \operatorname{Ext}^{1}_{\operatorname{Com}(\mathsf{X}\text{--}\operatorname{Proqc}_{\mathrm{fl}})}(\mathcal{F}^{\bullet},\mathcal{C}^{\bullet})=0 Ext Com ( X – Proqc fl ) 1 ( F ∙ , C ∙ ) = 0 for any acyclic complex F ∙ \mathcal{F}^{\bullet} F ∙ in the exact category of very flat pro-quasi-coherent pro-sheaves X \mathsf{X} X –Proqc v f l \operatorname{Proqc}_{\mathrm{vfl}} Proqc vfl and any complex of levelwise contraadjusted flat pro-quasi-coherent pro-sheaves C ∙ \mathcal{C}^{\bullet} C ∙ on X \mathsf{X} X .
(b) Let X \mathsf{X} X be an ind -semi-separated ind -Noetherian ℵ 0 \aleph_{0} ℵ 0 -ind -scheme. Then one has Ext Com ( X – Proqc f l ) 1 ( F ∙ , C ∙ ) = 0 \operatorname{Ext}^{1}_{\operatorname{Com}(\mathsf{X}\text{--}\operatorname{Proqc}_{\mathrm{fl}})}(\mathcal{F}^{\bullet},\mathcal{C}^{\bullet})=0 Ext Com ( X – Proqc fl ) 1 ( F ∙ , C ∙ ) = 0 for any acyclic complex F ∙ \mathcal{F}^{\bullet} F ∙ in the exact category of flat pro-quasi-coherent pro-sheaves X \mathsf{X} X –Proqc f l \operatorname{Proqc}_{\mathrm{fl}} Proqc fl and any complex of levelwise cotorsion flat pro-quasi-coherent pro-sheaves C ∙ \mathcal{C}^{\bullet} C ∙ on X \mathsf{X} X .
Proof. The proof is similar to that of Corollary 6.11. The argument is based on Theorem 6.12, Lemma 6.10, and Proposition 6.13. □
Institute of Mathematics, Czech Academy of Sciences, Žitná 25, 115 67 Prague 1, Czech Republic
Email address: positselski@math.cas.cz