🔢 mathematics

High Frobenius pushforwards generate the bounded derived category

This paper establishes that for a noetherian scheme of prime characteristic with a finite Frobenius map, sufficiently high Frobenius pushforwards of any compact generator generate the bounded derived category of coherent sheaves, with the required threshold depending on the scheme's codepth and holding for all positive integers when the scheme is locally a complete intersection.

Original authors: Matthew R. Ballard, Srikanth B. Iyengar, Pat Lank, Alapan Mukhopadhyay, Josh Pollitz

Published 2026-01-28

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Matthew R. Ballard, Srikanth B. Iyengar, Pat Lank, Alapan Mukhopadhyay, Josh Pollitz

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 rebuild a massive, intricate city (the "bounded derived category") using only a specific set of Lego bricks. In the world of mathematics, this city represents all the complex shapes and structures you can build from a geometric space (like a surface or a curve). Usually, to rebuild the whole city, you need a huge, diverse collection of unique bricks.

This paper asks a very specific question: Can we rebuild this entire city using just one special type of brick, but only if we process that brick through a specific machine called the "Frobenius map"?

Here is the breakdown of their discovery, using everyday analogies:

1. The City and the Bricks

Think of the geometric space (a "scheme") as a city. Some cities are perfectly smooth and regular (like a grid of identical houses). Others are "singular," meaning they have cracks, sharp corners, or weird distortions.

The Goal: Mathematicians want to know if they can generate (rebuild) every possible structure in this city using a single "master brick" (a generator).
The Problem: In cities with cracks (singularities), the standard master brick often isn't enough. You usually need a whole toolbox of different bricks to fix the broken parts.

2. The Magic Machine: The Frobenius Map

The authors focus on a special kind of machine called the Frobenius map. Imagine this machine takes a Lego brick and "stamps" it with a specific pattern $p$ times (where $p$ is a prime number, like 2, 3, or 5).

When you run a brick through this machine once ( $e=1$ ), it changes shape.
When you run it through it many times ( $e$ is large), it changes shape again.
The paper investigates: If we take a standard brick, run it through this machine enough times, does the resulting "super-brick" become powerful enough to rebuild the entire city, even the broken parts?

3. The Main Discovery: "High Frobenius Pushforwards"

The authors prove a powerful rule: If you run the machine enough times, the resulting brick can rebuild the whole city.

The "Codepth" Meter: The paper introduces a concept called codepth. Think of this as a "damage meter" for the city.
- If the city is perfectly smooth, the damage meter reads 0.
- If the city is very broken, the meter reads a higher number.
The Rule: You need to run the machine a number of times ( $e$ $e$ ) that is larger than the logarithm of this damage meter.
- Analogy: If the city has a few cracks (low codepth), you only need to stamp the brick a few times. If the city is a disaster zone (high codepth), you need to stamp it many, many times.
- Once you hit that number, the resulting "Frobenius brick" is strong enough to generate every structure in the city.

4. The Special Case: Perfectly Broken Cities (Locally Complete Intersections)

There is a specific type of city called a "Locally Complete Intersection." These are cities where the damage is very structured and predictable (like a building that collapsed in a very specific way).

The Surprise: For these specific types of cities, you don't need to run the machine many times. Running it just once ( $e=1$ ) is enough!
Even if the city is broken, a single pass through the Frobenius machine creates a brick strong enough to rebuild everything. This is a much stronger result than the general rule.

5. When Does It Fail?

The paper also explores when this magic trick doesn't work.

The "Curved" Trap: If the city is a smooth curve with a "hole" in the middle (like a donut shape, or a curve with positive genus), the Frobenius machine might never produce a single brick strong enough to rebuild the whole thing, no matter how many times you run it.
The "F-thick" Label: The authors give a name to cities where this trick does work: F-thick.
- All "flat" cities (affine schemes) are F-thick.
- Many beautiful, symmetric cities (like projective spaces) are F-thick.
- But some curved, hole-filled cities are not F-thick.

6. Why This Matters (In Math Terms)

The authors aren't just playing with Lego; they are solving a deep problem about how to measure the "brokenness" of a mathematical space.

They show that the Frobenius machine acts like a universal repair tool.
They prove that for many types of broken spaces, you don't need a complex toolbox; you just need to wait long enough (run the machine enough times) for the single master brick to become powerful enough to do the job.
They also found that for certain types of broken spaces, the machine works instantly (after just one run).

Summary

In simple terms, the paper says: "If you have a broken geometric space, you can fix it (mathematically speaking) by taking a standard building block and running it through a specific 'Frobenius' machine. If the space is moderately broken, you need to run it many times. If the space is 'structurally' broken in a specific way, running it once is enough. However, if the space is a smooth curve with a hole, this trick might never work."

This provides a clear, mechanical way to understand the complexity of geometric shapes using the power of repetition and a specific mathematical operation.

Technical Summary: High Frobenius Pushforwards Generate the Bounded Derived Category

Problem Statement
This work addresses the existence and explicit identification of (strong) generators for the bounded derived category of coherent sheaves, $D^b(\text{coh } X)$ , over a noetherian scheme $X$ of prime characteristic $p$ . While the existence of generators for such categories is broadly established (e.g., by Bondal and Van den Bergh, Rouquier, and Aoki), there is a scarcity of general results that identify explicit generators, particularly in the presence of singularities. The authors seek to determine whether high Frobenius pushforwards of specific objects can serve as such generators and to quantify the conditions under which this occurs.

Methodology
The paper employs a combination of local-to-global principles, homological algebra over local rings, and the theory of simplicial algebras. The methodology proceeds in three main stages:

Local-to-Global Principle: The authors utilize a local-to-global principle for triangulated categories, building on work by Stevenson. They establish that a complex $G$ generates $D^b(\text{coh } X)$ as a module over the tensor-triangulated category of perfect complexes ( $\text{Perf } X$ ) if and only if its stalk $G_x$ generates the derived category of the local ring $D(\text{mod } \mathcal{O}_{X,x})$ for every $x \in X$ . This reduction allows the global problem to be analyzed at the level of local rings.
Simplicial Algebra and Nilpotence: To handle the local case, the authors investigate the Frobenius endomorphism on a noetherian local ring $R$ . They utilize the Dold-Kan correspondence to translate the problem into the category of simplicial modules. By constructing a specific commutative diagram of simplicial algebras involving a minimal Cohen presentation, they prove a nilpotence-type result: for sufficiently large $e$ , the Frobenius pushforward $F^e_* M$ of a complex $M$ with full support essentially "kills" the higher homological obstructions, allowing the residue field $k$ to be built from $F^e_* M$ .
Support and Thick Subcategories: The authors leverage the theory of support for complexes (developed by Foxby, Neeman, and others) to relate the generation of the derived category to the support of the objects. They show that if the support of a complex $E$ is contained in the support of $G$ , then $E$ lies in the thick subcategory generated by $G$ under appropriate conditions.

Key Contributions and Results

Theorem A (Main Result): Let $X$ be a noetherian $F$ -finite scheme of prime characteristic $p$ . For any generator $E$ of $\text{Perf } X$ and any $G \in D^b(\text{coh } X)$ with full support, the complex $F^e_*(E \otimes^L_X G)$ generates $D^b(\text{coh } X)$ whenever $p^e > \text{codepth}(X)$ . Here, $\text{codepth}(X)$ is a numerical invariant measuring the singularity of $X$ (defined as the supremum of $\text{edim}(\mathcal{O}_{X,x}) - \text{depth}(\mathcal{O}_{X,x})$ ). If $X$ is separated, this generator is strong.
Corollaries for Specific Schemes:
- Affine Case: If $R$ is an $F$ -finite commutative noetherian ring, $F^e_* R$ is a strong generator for $D^b(\text{mod } R)$ for $e > \log_p(\text{codepth } R)$ .
- Quasi-projective Case: For a quasi-projective scheme $X$ with a very ample line bundle $L$ , the direct sum $\bigoplus_{i=0}^{\dim X} F^e_*(L^{\otimes i})$ strongly generates $D^b(\text{coh } X)$ under the same bound on $e$ .
Locally Complete Intersection (LCI) Case (Theorem D): If $X$ is $F$ -finite and locally a complete intersection, the first Frobenius pushforward $F_* G$ (where $G$ generates $\text{Perf } X$ ) is already a strong generator for $D^b(\text{coh } X)$ . This implies $F_* \mathcal{O}_X$ strongly generates the category for affine LCI schemes.
F-thickness: The authors define a scheme $X$ $X$ to be F-thick if $F^e_* \mathcal{O}_X$ $F_{*}^{e} O_{X}$ generates $D^b(\text{coh } X)$ $D^{b} (coh X)$ for some $e \ge 1$ $e \geq 1$ . They prove:
- All affine schemes are F-thick.
- Smooth projective curves are F-thick if and only if their genus is zero.
- For smooth projective complete intersections, F-thickness is equivalent to the generation of the Kuznetsov component by $F^e_* \mathcal{O}_X$ .
- Examples of F-thick varieties include projective spaces, certain Grassmannians, and Severi-Brauer varieties.
Numerical Invariants (F-level): The paper introduces the Frobenius level (F-level) of a ring $R$ , defined as the minimal $n$ such that $R$ is in the $n$ -th thick subcategory generated by $F^e_* R$ for some $e$ . They show this invariant is finite for all $F$ -finite rings and equals 1 if and only if $R$ is $F$ -split. For LCI rings, the F-level is bounded by $p \cdot \text{codepth}(R)$ .

Significance and Claims
The paper claims to provide a canonical and explicit method for identifying generators for the bounded derived category of coherent sheaves in prime characteristic, refining the structural understanding of Kunz's theorem. Specifically:

It establishes that high Frobenius pushforwards universally generate the derived category for $F$ -finite schemes, with the required power of Frobenius controlled by the singularity measure $\text{codepth}(X)$ .
It offers a structural refinement of Kunz's theorem: if $F^n_* G$ is perfect for some $G$ with full support, then $X$ is regular.
It connects the generation of derived categories to the generation of the Kuznetsov component in the context of complete intersections.
It introduces the concept of F-thickness, linking the generation of the derived category by the structure sheaf to the geometry of the scheme (e.g., distinguishing between genus 0 and higher genus curves).

The authors emphasize that while the existence of generators is known, their work identifies them explicitly via Frobenius pushforwards, which are natural objects in positive characteristic geometry. The results are presented as definitive for $F$ -finite schemes, with specific bounds provided for the power of Frobenius required.