🔢 mathematics

Localizing subcategories for algebraic stacks

This paper establishes a descent principle for $\otimes$ -localizing subcategories along smooth presentations using Balmer and Mathew's notion of descendability, thereby enabling the classification of such subcategories in the derived category of suitable algebraic stacks via subsets of their underlying topology.

Original authors: Pat Lank

Published 2026-06-02

📖 4 min read🧠 Deep dive

CC BY 4.0

Original authors: Pat Lank

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, chaotic library. This library isn't just a building; it's a complex, multi-layered structure with different wings, some of which are connected by secret tunnels, and some rooms that only exist if you look at them from a specific angle. In the world of mathematics, this "library" is called an algebraic stack, and the "books" inside are complex mathematical objects called derived categories.

For a long time, mathematicians knew how to organize the books in simple, flat libraries (called schemes). They discovered a golden rule: If you want to create a specific section of the library (a "localizing subcategory"), you can do it simply by picking a few specific "seed" books (residue fields) from the shelves. The entire section is then built up from these seeds. It's like saying, "If I have a seed for an apple tree, I can grow an entire orchard."

However, when mathematicians tried to apply this rule to the complex, multi-layered libraries (algebraic stacks), they hit a wall. The rules for simple libraries didn't seem to work for the complicated ones. They weren't sure if the "seed" method still worked, or if the complex structure of the library would break the system.

What this paper does:
The author, Pat Lank, solves this puzzle by introducing a new way of looking at the problem. Instead of trying to build the complex library from scratch, he uses a "descent" strategy. Think of it like this:

The Smooth Presentation (The Blueprint): Imagine you have a complex, 3D sculpture (the algebraic stack). It's hard to understand all at once. But, you can cover it with a smooth, flat sheet of plastic (a "smooth presentation" from a simple scheme). This sheet maps perfectly onto the sculpture, letting you see every detail of the 3D shape by looking at the 2D sheet.
The Point-Generated Rule: The paper proves that if your flat sheet (the simple library) follows the "seed" rule (meaning it's "point-generated"), then the complex 3D sculpture (the algebraic stack) also follows the rule.
The Result: Because the flat sheet is easy to organize using seeds, and the sheet perfectly represents the sculpture, the sculpture can also be organized using seeds.

The Big Discovery:
The paper establishes a direct link between the topology (the shape and layout) of the library and the sections you can create inside it.

The Map: Every possible section you can make in the library corresponds exactly to a specific subset of points on the library's floor plan.
The Seeds: Every section is built by taking the "structure sheaves" (the fundamental building blocks) of specific points (like tiny fields) and pushing them into the main library.

Why it matters (in simple terms):
Before this, mathematicians were stuck. They knew the rule worked for flat libraries, but they didn't know if it worked for the complex, twisted libraries used in modern geometry. This paper says, "Yes, it works, provided you can cover the complex library with a flat, well-behaved sheet."

It's like discovering that even though a castle has a maze of secret passages and hidden towers, if you can lay a flat map over the whole thing that shows every room clearly, then you can organize the entire castle's contents just by looking at the map. You don't need to walk every corridor; you just need the right map and the right seeds.

Key Takeaways:

The Problem: We didn't know how to classify sections of complex mathematical libraries (stacks).
The Solution: We can classify them by looking at their points, just like we do for simple libraries, as long as the complex library can be "smoothly presented" by a simple one.
The Analogy: If you can flatten a complex 3D object onto a 2D sheet without losing information, and the 2D sheet follows a simple organizing rule, then the 3D object follows that same rule.

The paper doesn't promise to build new bridges or cure diseases; it simply provides the mathematical "instruction manual" for organizing these complex structures, ensuring that the rules we trust for simple shapes hold true even in the most complicated geometric landscapes.

Technical Summary: Localizing Subcategories for Algebraic Stacks

Problem Statement
The classification of localizing subcategories in derived categories is a central problem in triangulated category theory. While Neeman established a bijection between localizing subcategories of the derived category of a Noetherian ring and subsets of its spectrum [Nee92], and this was extended to Noetherian schemes by Alonso Tarrío, Jeremiás López, and Souto Salorio [ATJLSS04], the situation for algebraic stacks is more complex.

In the context of algebraic stacks, the derived category of complexes with quasi-coherent cohomology, denoted $D_{qc}(X)$ , is not necessarily compactly generated [HNR19]. Consequently, the classification of $\otimes$ -localizing subcategories (those closed under the tensor product action) in terms of subsets of the underlying topological space $|X|$ remains open in this generality. Previous results, such as those by Stevenson [Ste13] and the recent work of Alonso Tarrío, Jeremiás López, and Souto Salorio [AJL24], rely on specific generation properties or compactness assumptions that do not universally apply to all algebraic stacks. The specific challenge addressed here is whether the classification of $\otimes$ -localizing subcategories extends to algebraic stacks without assuming a priori that $D_{qc}(X)$ is compactly generated.

Methodology
The paper employs a descent principle based on the notion of descendability introduced by Balmer and Mathew [Bal16, Mat16]. The core strategy involves reducing the problem for algebraic stacks to the case of schemes via smooth presentations.

Point-Generatedness: The paper introduces the concept of a "point-generated" stack (Definition 1.1). A quasi-compact quasi-separated algebraic stack $X$ is point-generated if $D_{qc}(X)$ is generated by the pushforwards of structure sheaves of spectra of fields, specifically objects of the form $R(ts)_* \mathcal{O}_{\text{Spec}(k)}$ , where $ts: \text{Spec}(k) \to X$ represents a point in $|X|$ .
Smooth Descent: The main technical observation is that point-generatedness satisfies a form of smooth descent. The authors prove that if an algebraic stack $X$ admits a smooth presentation by a point-generated scheme, then $X$ itself is point-generated.
Technical Tools:
- Lemma 3.1: Establishes that the inverse image of a localizing subcategory under an exact functor preserving small coproducts is localizing.
- Lemma 3.2: Proves that for a set of points $S \subseteq |X|$ , the $\otimes$ -localizing subcategory generated by the pushforwards of residue fields coincides with the standard localizing subcategory generated by the same objects, and is independent of the choice of representatives for the points.
- Lemma 3.3 & 3.4: Utilize properties of compact generators in schemes to relate tensor products and localizing subcategories.
- Lemma 3.5: The crucial step showing that for a concentrated stack $X$ with a smooth presentation from a point-generated affine scheme, any $\otimes$ -localizing subcategory is generated by the pushforwards of residue fields corresponding to a subset of $|X|$ .
- Proposition 3.8: Establishes that a quasi-compact quasi-separated scheme is point-generated if and only if it admits an affine open cover by point-generated affine schemes.

Key Contributions and Results
The paper establishes the following main results:

Theorem 1.2: Let $X$ be an affine-pointed concentrated algebraic stack. If there exists a smooth presentation of $X$ by a point-generated scheme, then $X$ is point-generated.
Corollary 1.3: Under the hypotheses of Theorem 1.2, there exists a bijection between the set of $\otimes$ $\otimes$ -localizing subcategories of $D_{qc}(X)$ $D_{q c} (X)$ and the subsets of the underlying topological space $|X|$ $∣ X ∣$ .
- The map $\phi$ sends a subcategory $\mathcal{T}$ to the set of points $p \in |X|$ represented by $t: \text{Spec}(k) \to X$ such that $Rt_* \mathcal{O}_{\text{Spec}(k)} \in \mathcal{T}$ .
- The inverse map $\theta$ sends a subset $S \subseteq |X|$ to the localizing subcategory generated by $\{R(ts)_* \mathcal{O}_{\text{Spec}(k)}\}_{s \in S}$ .

These results demonstrate that for a broad class of algebraic stacks (including those with affine stabilizers and finitely presented inertia, or equicharacteristic stacks), the $\otimes$ -localizing subcategories are entirely determined by the topology of the stack and are generated by the pushforwards of structure sheaves of spectra of fields.

Significance and Scope
The paper claims to recover the classification of $\otimes$ -localizing subcategories for Noetherian algebraic stacks (and schemes) via a different method than previous works, specifically by utilizing descendability and smooth presentations rather than relying on the classification of localizing subcategories for affine schemes directly [AJL24].

The authors note that their proofs do not mimic the scheme case directly but instead reduce the problem to the scheme case using the properties of smooth presentations. The results apply to Deligne–Mumford stacks finitely presented over a field and any Noetherian algebraic space.

The paper also highlights limitations and counterexamples:

Pathology 1.7: Citing Balchin–Omar Gómez–Stevenson [BOS26], the paper notes that point-generatedness fails for the polynomial ring in infinitely many variables. Consequently, Theorem 1.2 implies that point-generatedness fails for smooth presentations of such schemes, providing new examples where the classification fails even for schemes.
Remark 1.5: The paper distinguishes between the homological residue fields arising from the Balmer spectrum and the specific objects $R(ts)_* \mathcal{O}_{\text{Spec}(k)}$ used in this classification, noting they need not coincide.

In summary, the work extends the classification of $\otimes$ -localizing subcategories to a wider class of algebraic stacks by establishing a descent principle for the "point-generated" property, thereby linking the categorical structure of $D_{qc}(X)$ directly to the underlying topology $|X|$ without requiring compact generation.

More like this