🔢 mathematics

Generalized local jacobians and commutative group stacks

This paper recasts Grothendieck's construction of generalized local Jacobians for smooth varieties within the framework of higher algebraic group stacks by introducing a notion of algebraic homology that computes fppf cohomology and demonstrates that these Jacobians arise as the $E_1$ -page of a spectral sequence associated with a filtration by support dimension, while also extending these results to arbitrary bases.

Original authors: Bertrand Toen

Published 2026-02-24

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Bertrand Toen

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 understand the shape of a complex, multi-layered city (which mathematicians call a "scheme" or "variety"). You want to know its "homology"—a fancy word for its fundamental shape, holes, and connections.

In the 1960s, a mathematical giant named Alexander Grothendieck sketched a blueprint for a machine called $J^*(X)$ that could calculate this shape. He built it using a specific type of Lego brick called "commutative pro-algebraic groups." It worked well, but it was a bit rigid and only worked perfectly in smooth, flat cities.

Bertrand Toën's paper is like a modern renovation project. He says, "Let's rebuild Grothendieck's machine using a newer, more flexible type of Lego called Higher Stacks." This new machine, which he calls Algebraic Homology ( $H_{alg}(X)$ ), is more powerful. It can handle cities with bumps, cracks, and weird corners (singularities) and works in a broader mathematical universe.

Here is the breakdown of the paper using simple analogies:

1. The New Lego Bricks: "Higher Stacks"

In the old days, mathematicians built shapes using standard groups (like circles or lines). Grothendieck used "pro-objects," which are like infinite sequences of these groups getting smaller and smaller.

Toën introduces Commutative Algebraic Group Stacks.

The Analogy: Imagine a standard Lego brick is a simple group. A "stack" is like a Lego brick that has its own internal universe of smaller bricks and connections. It's a "group of groups."
Why it matters: These new bricks are "higher" dimensional. They can capture not just the shape of the city, but the "twists" and "loops" in the data that live inside the city. This allows the machine to see more details than Grothendieck's original version.

2. The Machine: Algebraic Homology

Toën defines Algebraic Homology ( $H_{alg}(X)$ ) as the ultimate "Universal Shape Detector."

How it works: If you have a city $X$ , this machine creates a "shadow" or a "map" of $X$ that captures every possible way you can wrap a group around it.
The Universal Property: It's like a master key. If you want to know how a specific group (like a circle) fits into your city, you don't need to measure the city directly. You just look at how your city maps to this "Universal Shape Detector." The detector tells you everything you need to know.

3. The Filter: Peeling the Onion

The machine produces a massive, complex object. To make sense of it, Toën introduces a filtration (a filter).

The Analogy: Imagine the city is an onion. The outer layer is the whole city. The next layer is the neighborhoods. The next is the individual houses. The innermost layer is the specific points (like street corners).
The Process: Toën slices the "Universal Shape Detector" into layers based on the dimension of the support.
- Layer 0: The whole city.
- Layer 1: The neighborhoods.
- Layer $d$ : The specific points of a certain size.
The Result: When you look at just one of these layers (the "graded piece"), it turns out to be a collection of Local Generalized Jacobians. These are the tiny, local shape-detectors for each specific point in the city.

4. The Big Reveal: Connecting to Grothendieck

This is the paper's main "Aha!" moment.

The Connection: Toën proves that if you take his fancy, modern "Universal Shape Detector" and slice it up, the bottom layer (the most basic part) is exactly Grothendieck's original machine ( $J^*(X)$ ).
The Metaphor: Think of Grothendieck's machine as a black-and-white photograph of a city. Toën's machine is a 3D hologram. If you squint at the hologram and look only at the shadow it casts on the wall, you get the black-and-white photo.
Why it's cool: It proves that Grothendieck was right, but he was only seeing the "shadow" of a much deeper, richer reality. Toën's new machine recovers Grothendieck's results but adds all the missing 3D details.

5. The "Affine" Backup Plan

The paper admits a limitation: The fancy "Universal Shape Detector" works best when the city is built over a perfect field (a very clean, well-behaved mathematical ground). If the ground is muddy or weird (arbitrary base rings), the machine gets complicated.

The Solution: Toën creates a simplified version called Affine Homology.
The Analogy: If the full machine is a high-end sports car that needs premium fuel, the Affine Homology is a rugged truck that can drive on any terrain. It focuses only on the "unipotent" parts (the simplest, most flexible shapes).
The Magic Trick: He shows that this simplified machine can be translated into a Differential Graded (DG) Module.
- The Metaphor: This is like translating a complex 3D sculpture into a set of musical notes (a code). If you have the code (the DG module), you can reconstruct the sculpture. This connects the geometry of the city to the algebra of "Dieudonné theory" (a way of studying shapes using linear algebra).

Summary

Bertrand Toën took a classic, 60-year-old mathematical idea (Grothendieck's Local Jacobians) and rebuilt it using modern, high-tech tools (Higher Stacks).

He built a better machine ( $H_{alg}$ ) that sees more details.
He proved that the old machine ( $J^*$ ) is just the "shadow" of this new machine.
He showed that even when the ground is messy, a simplified version of the machine still works and can be translated into a code (DG modules) that mathematicians can easily read.

In short: The old blueprint was correct, but the new building is taller, stronger, and has a better view.

1. Problem Statement

The paper addresses the need to recast and generalize Alexander Grothendieck's construction of a complex $J^*(X)$ of commutative pro-algebraic groups associated with a smooth variety $X$ . In his 1960 letter to Serre, Grothendieck sketched a construction where $J^*(X)$ computes the Zariski cohomology of $X$ with coefficients in algebraic groups and serves as a "homology" of $X$ .

The limitations of the classical construction and the motivation for this work include:

Categorical Framework: Grothendieck's construction lives in the category of complexes of pro-algebraic groups. Toën aims to realize this within the modern framework of $\infty$ -categories of higher algebraic stacks (specifically, commutative group objects in the $\infty$ -category of Artin stacks).
Topology: The original construction relies on the Zariski topology. This paper seeks to adapt the theory to the fppf topology (faithfully flat and finitely presented), which is more suitable for arithmetic geometry and flat cohomology.
Generality: The paper aims to extend these concepts beyond smooth varieties over perfect fields to more general schemes, specifically by restricting to unipotent coefficients and utilizing the theory of affine stacks.

2. Methodology

Toën employs the language of higher category theory and derived algebraic geometry. The core methodological steps are:

Higher Stacks and $\infty$ -Categories: The work is set in the $\infty$ -topos of hypercomplete fppf stacks ( $\text{Stk}$ ) over a perfect field $k$ .
Commutative Algebraic Group Stacks: The author defines a new category, $\text{AbSt}^{\text{alg}}_k$ , consisting of commutative group objects in Artin stacks that are locally of finite type and strongly quasi-separated. These are characterized as connective complexes of sheaves whose homotopy sheaves are representable by locally algebraic groups.
Algebraic Homology ( $H_{\text{alg}}(X)$ ): Defined as the pro-object in $\text{AbSt}^{\text{alg}}_k$ that pro-represents the functor $E \mapsto \text{Map}(X, E)$ . This object serves as a universal recipient for maps from $X$ to algebraic group stacks, effectively acting as a "derived Albanese variety."
Filtration by Dimension: A filtration $F_\bullet H_{\text{alg}}(X)$ is introduced based on the dimension of supports. The graded pieces $\text{Gr}_d H_{\text{alg}}(X)$ are shown to be products of local algebraic homologies at points of dimension $d$ .
Purity Theorems: The proof relies heavily on purity results for local cohomology of smooth schemes with coefficients in group schemes. This ensures that the graded pieces of the homology are connective with respect to Beilinson's $t$ -structure.
Affine Homology ( $H_u(X)$ ): For arbitrary bases, the author restricts to unipotent coefficients and uses the theory of affine stacks (dual to cosimplicial commutative algebras). This leads to the definition of "affine homology types," which are generated by limits of Eilenberg-MacLane stacks $K(\mathbb{G}_a, n)$ .

3. Key Contributions

A. Definition of Algebraic Homology

The paper introduces $H_{\text{alg}}(X)$ as a universal pro-object in the category of commutative algebraic group stacks.

Universal Property: It represents the mapping space $\text{Map}(X, E)$ for any commutative algebraic group stack $E$ .
Relation to Albanese: The 0-th truncation of $H_{\text{alg}}(X)$ recovers the classical Albanese variety (and its torsor structure), while the higher truncations encode higher cohomological data (e.g., related to the Picard stack).

B. Comparison with Grothendieck's Construction

The central result is a comparison theorem between the new higher-stack definition and Grothendieck's classical complex $J^*(X)$ .

Theorem: If $X$ is smooth over a perfect field $k$ , there is a natural isomorphism:
$H^0_B(F_\bullet H_{\text{alg}}(X)) \simeq J^*(X)$
Here, $H^0_B$ denotes the 0-th cohomology with respect to Beilinson's $t$ -structure on filtered objects.
Interpretation: Grothendieck's complex $J^*(X)$ is identified as the $n$ -th truncation (where $n = \dim X$ ) of the full algebraic homology $H_{\text{alg}}(X)$ . The full object $H_{\text{alg}}(X)$ contains strictly more information than $J^*(X)$ , but $J^*(X)$ is sufficient to compute cohomology under specific conditions.

C. Computation of Cohomology

The paper proves that despite $J^*(X)$ being a truncation, it retains the ability to compute flat cohomology for specific coefficients.

Theorem: For $i \geq 0$ and a commutative algebraic group $G$ (affine if $i>0$ , unipotent if $i>1$ ), the canonical projection $H_{\text{alg}}(X) \to J^*(X)$ induces an isomorphism:
$\text{Ext}^i_{\text{naive}}(J^*(X), G) \simeq H^i_{\text{fppf}}(X, G)$
Here, $\text{Ext}^i_{\text{naive}}$ refers to Yoneda extensions in the derived category of pro-algebraic groups, distinct from the sheaf-theoretic Ext groups in the stack category.

D. Affine Homology and Generalization

The final section extends the theory to arbitrary base rings by focusing on the unipotent part of the theory.

Affine Homology Types: Defined as the smallest sub- $\infty$ -category of abelian group stacks stable by limits and containing $\mathbb{G}_a[n]$ .
Dieudonné Theory Analogue: The affine homology $H_u(X)$ is characterized by the complex $R\Gamma(X, \mathbb{G}_a)$ viewed as a dg-module over the endomorphism dg-algebra $\text{REnd}(\mathbb{G}_a)$ . This provides a "Dieudonné theory" for affine homology types.
Base Change: The theory is shown to behave well under base change for local complete intersection morphisms.

4. Results

Existence of $H_{\text{alg}}(X)$ : Proven via the left exactness of the mapping functor in the $\infty$ -category of stacks.
Connectivity: For a smooth scheme $X$ of dimension $n$ , the filtered object $F_\bullet H_{\text{alg}}(X)$ is $0$-connective with respect to Beilinson's $t$ -structure.
Isomorphism: The identification of Grothendieck's $J^*(X)$ as the Beilinson truncation of $H_{\text{alg}}(X)$ .
Cohomological Equivalence: Proof that $J^*(X)$ computes $H^i_{\text{fppf}}(X, G)$ for affine/unipotent $G$ , validating its utility despite being a truncation.
Affine Homology Characterization: $H_u(X)$ is fully determined by the dg-algebra structure of $R\Gamma(X, \mathbb{G}_a)$ .

5. Significance

Modernization of Classical Theory: The paper successfully translates Grothendieck's 1960s intuition into the rigorous language of higher algebraic geometry and $\infty$ -categories. This clarifies the structural nature of "algebraic homology" as a universal object.
Unification of Concepts: It bridges the gap between the classical theory of Jacobians/Albanese varieties and modern derived algebraic geometry, showing that the classical complex is a specific truncation of a richer derived object.
New Tools for Arithmetic Geometry: By introducing affine homology types, the paper provides a framework to study cohomology with unipotent coefficients over arbitrary bases, a setting where classical duality and purity theorems often fail or are difficult to apply.
Connection to Lie Theory: The paper hints at deep connections between affine homology and partition Lie algebras (via Koszul duality with LSym-algebras), suggesting a new avenue for studying abelian homotopy types.
Limitations and Future Work: The author acknowledges that the smoothness assumption is crucial for the purity results used in the main comparison. Extending the full theory (including non-unipotent coefficients) to singular schemes or general bases remains an open problem, though the affine/unipotent case offers a robust partial solution.

In summary, Toën's work provides a foundational re-interpretation of algebraic homology, elevating it from a complex of pro-groups to a higher stack, thereby offering a more flexible and powerful tool for computing flat cohomology and understanding the geometric structure of schemes.