🔢 mathematics

Roos axiom holds for quasi-coherent sheaves

This paper establishes that the category of quasi-coherent sheaves on a quasi-compact semi-separated or a Noetherian scheme of finite Krull dimension satisfies the Roos axiom $\mathrm{AB}4^*$ - $n$ by providing two distinct proofs: an elementary approach using Čech coresolutions and a conceptual approach leveraging finite projective dimension generators or the co-contra correspondence with contraherent cosheaves.

Original authors: Leonid Positselski

Published 2026-02-20

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Leonid Positselski

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine you are trying to organize a massive, infinite library. This library isn't just a building; it's a living, breathing mathematical universe called a Scheme (specifically, a "quasi-compact semi-separated" or "Noetherian" one). Inside this library, you have millions of books (mathematical objects called quasi-coherent sheaves).

The author of this paper, Leonid Positselski, is tackling a very specific problem about how these books behave when you try to bundle them together in infinite groups.

Here is the breakdown of the paper's story, using simple analogies.

The Big Problem: The "Infinite Bundle" Glitch

In mathematics, there is a rule called AB4*. It basically says: "If you take an infinite number of perfect, well-behaved bundles of books, the resulting super-bundle should also be perfect."

However, in this specific library (the world of quasi-coherent sheaves), this rule breaks. If you try to bundle an infinite number of these books, the result often gets messy, "leaky," or broken. The structure falls apart.

The Question: If the rule breaks completely, is there a way to say, "Okay, it breaks, but only this much"?
The Answer: Yes. This is the Roos Axiom (AB4*-n). It's like saying, "The infinite bundle might get a little messy, but the messiness has a limit. It won't spiral out of control forever; it will stop being messy after a certain number of steps."

Positselski proves that in these specific libraries, the messiness does have a limit. The "derived functors of infinite direct product" (the mathematical way of measuring how messy the bundle gets) eventually become zero.

The Three Ways to Prove It

Positselski doesn't just say "it works." He provides three different ways to prove it, like showing three different routes to climb a mountain.

Route 1: The "Czechoslovakian" Map (The Čech Coresolution)

The Analogy: Imagine the library is a huge city. You don't know the whole city at once, so you break it down into smaller neighborhoods (affine open subschemes).
The Method: You create a "map" (a Čech complex) that stitches these neighborhoods together.
- If the city is semi-separated (neighborhoods overlap nicely), you can build a map with a specific number of layers.
- If the city is Noetherian (finite complexity), you can still build a map, but it's more complicated because the neighborhoods might overlap in weird ways.
The Result: By counting the layers of this map, Positselski shows that the "messiness" of the infinite bundle stops after a specific number of layers. It's like showing that if you stack enough blankets, the bottom one stops feeling the weight of the ones above it.

Route 2: The "Super-Book" Generator (Very Flat Sheaves)

The Analogy: Imagine you need to build any book in the library using a special set of "Super-Books."
The Method: Positselski introduces a special type of book called a "Very Flat" sheaf. These are incredibly sturdy, flexible books that can be used to build any other book in the library.
The Twist: He proves that there is a "Master Super-Book" (a generator) that can build everything else, and this Master Book isn't too complex. It has a "projective dimension" (a measure of complexity) that is finite.
The Result: Because the Master Book is simple enough, any infinite bundle built from it will eventually stop getting messy. It's like saying, "If the foundation of your house is solid and simple, you can build a skyscraper on top of it, and it won't collapse."

Route 3: The "Mirror World" (Co-Contra Correspondence)

The Analogy: Imagine there is a "Mirror World" (a category of contraherent cosheaves) that looks exactly like our library but has a superpower: in the Mirror World, infinite bundles never break. They are perfectly stable.
The Method: Positselski uses a mathematical "magic mirror" (a triangulated equivalence) that translates problems from our messy library to the perfect Mirror World.
- He takes a messy infinite bundle from our world.
- He reflects it into the Mirror World.
- In the Mirror World, the bundle is perfect and stable.
- He reflects it back.
The Result: Because the Mirror World is so well-behaved, the "messiness" in our world is forced to be finite. The mirror forces the chaos to have a limit. This is the most abstract route, but it's very powerful because it connects two different mathematical universes.

Why Does This Matter?

You might ask, "Who cares if infinite bundles get a little messy?"

It Saves the Day for Calculations: In advanced math (like algebraic geometry), we often need to take limits or infinite products to solve problems. If these operations were completely broken, many proofs would be impossible. Knowing they are "only n-broken" means we can still do the math; we just have to account for a small amount of error.
Better Tools: This result allows mathematicians to use specific, explicit tools (like "homotopy injective resolutions") to solve problems without needing to use very heavy, vague machinery. It makes the toolbox lighter and more precise.
Understanding the Universe: It tells us that even in complex, infinite mathematical structures, there is an underlying order. The chaos isn't random; it's bounded.

The Takeaway

Leonid Positselski has shown that in the complex world of algebraic geometry, the "infinite bundle" problem isn't a disaster. It's just a manageable inconvenience. Whether you look at it through a map of neighborhoods, a sturdy foundation of special books, or a magic mirror, the conclusion is the same: The chaos has a limit.

This gives mathematicians the confidence to keep building their theories, knowing that the infinite structures they work with won't collapse under their own weight.

Technical Summary: "Roos Axiom Holds for Quasi-Coherent Sheaves"

Author: Leonid Positselski
Subject: Homological Algebra, Algebraic Geometry, Derived Categories
Core Problem: Determining the homological behavior of infinite direct products in the category of quasi-coherent sheaves ( $X\text{--Qcoh}$ ) on schemes.

1. Problem Statement

In an abelian category $\mathcal{A}$ , Grothendieck's axiom AB4* states that infinite direct products are exact. It is well-established that the category of quasi-coherent sheaves $X\text{--Qcoh}$ generally fails to satisfy AB4* (e.g., on the projective line $\mathbb{P}^1_k$ ).

The paper investigates Roos' axiom AB4-n* (proposed by Roos in 2006), a weakening of AB4*. An abelian category with infinite products and enough injectives satisfies AB4*-n if the derived functors of the infinite direct product, denoted $Q^{(i)}_{\lambda \in \Lambda}$ , vanish for all $i > n$ .

Significance: If AB4*-n holds, it implies that the derived functor of the inverse limit ( $\varprojlim$ ) has homological dimension at most $n+1$ . Furthermore, it allows for the explicit construction of homotopy injective (DG-injective) resolutions without relying on the non-explicit "small object argument."

The paper aims to prove that $X\text{--Qcoh}$ satisfies AB4*-n for two specific classes of schemes:

Quasi-compact semi-separated schemes.
Noetherian schemes of finite Krull dimension.

2. Methodology

The author provides two distinct proofs for each of the two settings, utilizing different mathematical tools:

A. The Čech Coresolution Argument (Elementary Approach)

This approach relies on constructing explicit resolutions using open coverings.

Mechanism: For a scheme $X$ covered by open subschemes, the author constructs a Čech coresolution for any quasi-coherent sheaf $M$ . This resolution involves direct images of restrictions of $M$ to intersections of the covering sets.
Logic:
1. If the covering consists of affine schemes (or schemes with specific properties), the infinite product functors on these sub-schemes are exact (or have bounded derived functors).
2. The direct image functors under flat affine morphisms preserve the vanishing of higher derived products.
3. By analyzing the length of the Čech complex and the vanishing properties of the terms, one bounds the homological dimension of the product functor on the whole scheme.
Application:
- Semi-separated case: Uses a finite affine open covering.
- Noetherian case: Uses a more sophisticated version involving flasque quasi-coherent sheaves and the combinatorial complexity of the affine coverings of intersections.

B. The Generator and Co-Contra Correspondence Argument (Conceptual Approach)

This approach uses category-theoretic properties and the existence of specific generators.

Mechanism 1 (Semi-separated):
- Introduces very flat quasi-coherent sheaves.
- Proves the existence of a generator $F$ for $X\text{--Qcoh}$ that is very flat.
- Shows that the projective dimension of this generator is finite (bounded by the number of affine sets in the covering).
- Uses the fact that the existence of a generator of finite projective dimension implies AB4*-n.
Mechanism 2 (Noetherian & General):
- Utilizes the "naïve co-contra correspondence." This is a triangulated equivalence between the unbounded derived category of quasi-coherent sheaves $D(X\text{--Qcoh})$ and the derived category of contraherent cosheaves $D(X\text{--Ctrh})$ .
- The category of contraherent cosheaves has exact infinite products.
- The paper proves a general theorem: If there is a triangulated equivalence $D(\mathcal{A}) \simeq D(\mathcal{B})$ restricting to bounded categories, and $\mathcal{B}$ has exact products, then $\mathcal{A}$ satisfies AB4*-n.

3. Key Contributions and Results

Main Theorems

Theorem 1.4 (Semi-separated): If $X$ is a quasi-compact semi-separated scheme covered by $N$ affine open subschemes, then $X\text{--Qcoh}$ satisfies AB4*- $n$ with $n = N - 1$ .
- Note: This provides a sharp numerical bound derived from the Čech argument.
Theorem 2.12 (Noetherian): If $X$ is a Noetherian scheme of finite Krull dimension $D$ , then $X\text{--Qcoh}$ satisfies AB4*- $n$ for a specific integer $n$ (derived from a combinatorial formula involving the covering and dimension).
- Refinement: Using the co-contra correspondence (Corollary 4.8), the author establishes the existence of such an $n$ , specifically showing $n = 2D + 1$ is achievable via the correspondence.
Proposition 3.6 & Corollary 3.11 (Generator): For a quasi-compact semi-separated scheme covered by $N$ affines, there exists a very flat generator of projective dimension at most $N$ . This implies AB4*- $N$ .

Structural Results

Cotorsion Pairs: The paper discusses the hereditary complete cotorsion pair $(\text{very flat}, \text{contraadjusted})$ in $X\text{--Qcoh}$ for semi-separated schemes. It establishes that every quasi-coherent sheaf admits a finite coresolution by contraadjusted sheaves.
Co-Contra Correspondence: The paper formalizes the "naïve" version of the co-contra correspondence, proving that a triangulated equivalence between derived categories where one side has exact products forces the other side to satisfy AB4*-n.

4. Significance and Impact

Resolution of Open Questions: The paper confirms that the Roos axiom holds for quasi-coherent sheaves on a broad and important class of schemes (semi-separated and finite-dimensional Noetherian), resolving a gap in the understanding of derived functors of infinite products in algebraic geometry.
Explicit vs. Existential Proofs:
- The Čech argument provides explicit numerical bounds (e.g., $N-1$ ) based on the geometry of the scheme's covering.
- The co-contra argument provides a conceptual, structural proof that is more general (applicable to stacks and non-semi-separated cases under certain conditions) but less explicit regarding the specific value of $n$ without further analysis.
Practical Utility: By establishing AB4*-n, the paper validates the use of classical Cartan-Eilenberg and Spaltenstein constructions for homotopy injective resolutions in these categories. This avoids the need for the more abstract and less constructive "small object argument," facilitating concrete computations in derived categories of sheaves.
Foundational Theory: The work deepens the theory of very flat and contraadjusted sheaves, integrating them into the broader framework of cotorsion pairs and demonstrating their utility in controlling homological dimensions.

5. Conclusion

Positselski successfully demonstrates that while infinite products are not exact in $X\text{--Qcoh}$ , their derived functors vanish beyond a finite homological dimension determined by the scheme's geometric complexity (number of affine covers or Krull dimension). The paper offers a dual perspective: a concrete geometric proof via Čech coresolutions and a powerful abstract proof via the co-contra correspondence, significantly advancing the homological theory of quasi-coherent sheaves.