🔢 mathematics

On the cohomological classification of vector bundles on smooth real affine surfaces and threefolds

This paper establishes that the cohomological classification of vector bundles on smooth real affine surfaces and threefolds mirrors results from algebraically closed fields under specific assumptions, provides an efficient proof of Kucharz's theorem on Chern classes of rank 3 bundles, and presents the first example of a non-stably free projective module over a smooth affine real algebra of dimension 3 with trivial Chern classes.

Original authors: Samuel Lerbet

Published 2026-05-22

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Samuel Lerbet

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 an architect trying to build structures (called vector bundles) on a specific plot of land (a smooth real affine surface or threefold). In the world of algebraic geometry, these "structures" are complex mathematical objects that can twist and turn in various ways.

The central question of this paper is: Can we identify and distinguish every possible building just by looking at a few specific "blueprints" or "measurements" (called Chern classes)?

Here is a breakdown of the paper's findings using simple analogies:

1. The Blueprints vs. The Building

In mathematics, every building has a set of "blueprints" called Chern classes. Think of these as a list of numbers or coordinates that describe the building's shape, twists, and holes.

The Goal: The author wants to know if these blueprints are enough to tell two buildings apart. If two buildings have the exact same blueprints, are they actually the same building?
The Ideal World: In a "perfect" world (where the base field is algebraically closed, like working with complex numbers), the answer is a resounding YES. The blueprints perfectly describe the building. If the numbers match, the buildings are identical.

2. The Real World Problem

The paper focuses on the Real World (using real numbers, $\mathbb{R}$ ). Here, things get messy.

The Analogy: Imagine you have a map (the blueprints) that says a building has no holes. But in reality, the building might have a hidden, twisted tunnel that the map doesn't show because of the "topology" of the land (the real locus).
The Discovery: The author shows that on real plots of land, the blueprints are not always enough. Two buildings can have identical blueprints but still be different structures.

3. The "Hairy Ball" and the "Empty Room"

The paper uses two famous examples to explain why this happens:

The Hairy Ball (Rank 2 Bundles): Imagine a sphere (like a beach ball). You can try to comb the hair on it flat. Mathematically, you can't do it without creating a cowlick (a singularity). This is the "Hairy Ball Theorem." The paper shows that on certain real surfaces, you can have a "bundle" (a structure) that looks trivial (like a flat sheet) based on its blueprints, but it's actually knotted like the hair on a ball. The blueprints fail to catch this knot.
The Empty Room (Rank 3 Bundles): The author constructs a specific, tricky 3D shape where the "land" has no compact (closed and bounded) parts. In this specific case, the blueprints do work perfectly. If the land is "small" enough in a topological sense, the blueprints uniquely identify the building.

4. Answering the Big Questions

The paper tackles two specific questions left open by a previous mathematician, Kucharz:

Question A: If two Rank 3 buildings have the same blueprints, are they the same?

The Answer: No, not always.
The Counter-Example: The author uses a clever construction (involving a 3D space with a specific hole cut out of it) to build a structure that has zero blueprints (all its Chern classes are zero) but is not empty. It's like a building that looks like a flat sheet of paper on the blueprint but is actually a complex, non-trivial 3D object. This is the first time such a "ghost building" has been found on a smooth real 3D surface.

Question B: Can we build any structure if we are given a set of valid blueprints?

The Answer: Yes, but with a catch.
The Catch: There is a specific rule involving "Steenrod squares" (think of this as a special mathematical operation, like a checksum or a parity check). If your blueprints pass this specific checksum, you can build the structure. If they fail the checksum, the structure cannot exist. The author proves that this rule is the only thing stopping you from building the structure.

5. The "Magic" of Small Spaces

The paper concludes with a beautiful symmetry.

If the real land you are building on has no compact connected components (imagine an infinite plane or a shape that doesn't loop back on itself to form a closed bubble), then the blueprints work perfectly again.
In these "small" or "open" real spaces, the math behaves just like it does in the "perfect" complex world. The blueprints uniquely identify the building, and you can build anything that passes the checksum.

Summary

This paper is a detective story about mathematical buildings.

The Mystery: Do the blueprints (Chern classes) tell the whole story?
The Twist: In the real world, sometimes they don't. You can have "ghost buildings" that look empty on paper but are full of hidden complexity.
The Solution: The author found the first example of a "ghost building" in 3D space.
The Silver Lining: If the land is "open" enough (no closed loops), the blueprints work perfectly again, restoring order to the chaos.

The paper essentially draws a line in the sand: Topology matters. The shape of the real world determines whether your mathematical blueprints are a complete description or just a partial sketch.

Technical Summary: On the cohomological classification of vector bundles on smooth real affine surfaces and threefolds

Problem Statement
The paper investigates the cohomological classification of vector bundles on smooth real affine varieties, specifically focusing on surfaces and threefolds. The central inquiry concerns the map $\phi_r(X)$ , which assigns to a rank $r$ vector bundle $E$ on a variety $X$ the tuple of its Chern classes $(c_1(E), \dots, c_r(E))$ in the product of Chow groups $\prod CH^i(X)$ . Two fundamental questions are addressed:

Surjectivity: What is the image of $\phi_r(X)$ ? Which tuples of algebraic cycles can be realized as Chern classes?
Injectivity: What are the fibers of $\phi_r(X)$ ? Under what conditions are vector bundles determined up to isomorphism by their Chern classes?

While these questions have satisfactory answers over algebraically closed fields (where $\phi_r$ is often a bijection), the situation over the real numbers $\mathbb{R}$ is more delicate due to topological obstructions arising from the real locus $X(\mathbb{R})$ . The paper specifically addresses open questions posed by Kucharz regarding the injectivity of $\phi_3(X)$ and the characterization of the image of $\phi_2(X)$ for smooth real affine threefolds.

Methodology
The author employs a synthesis of motivic homotopy theory, algebraic K-theory, and classical algebraic topology.

Motivic Obstruction Theory: The paper utilizes the formalism of Moore–Postnikov towers for the motivic space $BGL_n$ . By analyzing the homotopy sheaves and $k$ -invariants of the map from $BGL_n$ to products of Eilenberg–Mac Lane spaces representing Chow groups, the author identifies specific obstructions to lifting cohomology classes to vector bundles.
Steenrod Operations: A crucial tool is the use of Steenrod squares ($Sq$) defined on Chow groups modulo 2 (following Voevodsky and Brosnan) and their compatibility with topological Stiefel–Whitney classes via the cycle class map $\gamma$ .
K-Theory and Filtrations: The study involves the filtration of the Grothendieck group $K_0(X)$ by codimension of support. The relationship between the graded pieces $Gr^p K_0(X)$ and Chow groups is used to construct bundles with prescribed Chern classes.
Real Locus Topology: The analysis heavily relies on the topology of the real locus $X(\mathbb{R})$ , particularly the presence or absence of compact connected components and the vanishing of specific cohomology groups like $H^3(X(\mathbb{R}), \mathbb{Z}/2)$ .
Explicit Constructions: Counter-examples are constructed using specific geometric setups involving hypersurfaces in $\mathbb{P}^3_{\mathbb{R}}$ and their affine complements, leveraging results from previous works (e.g., [AFL25]) and explicit computations in $K$ -theory.

Key Contributions and Results

1. Rank 2 Bundles on Surfaces and Threefolds

Surfaces: The paper revisits and generalizes the results of Barge and Ojanguren [BO87]. It establishes that for a smooth real affine surface $X$ , rank 2 bundles with a fixed determinant $L$ are classified by their Euler class in the Chow–Witt group $\widetilde{CH}^2(X, L)$ . The paper clarifies that if $X(\mathbb{R})$ has no compact connected components, the map $\phi_2(X)$ is a bijection.
Threefolds (Question B): The paper answers Question (B) affirmatively: the image of $\phi_2(X)$ for a smooth real affine threefold $X$ is exactly the set of pairs $(c_1, c_2)$ satisfying the relation $Sq_{a_1}(a_2) = 0$ in $H^3(X(\mathbb{R}), \mathbb{Z}/2)$ , where $a_i$ are the mod 2 reductions of the Chern classes. This is proven using motivic obstruction theory, identifying the obstruction to lifting as a class in $H^3(X, I^3(L))$ which vanishes precisely under this condition.
Classification under Small Real Locus: If $H^3(X(\mathbb{R}), \mathbb{Z}/2) = 0$ and $H^2(X(\mathbb{R}), \mathbb{Z}(L(\mathbb{R}))) = 0$ for all line bundles $L$ , then $\phi_2(X)$ is a bijection. This mirrors the behavior over algebraically closed fields.

2. Rank 3 Bundles and Kucharz's Theorem

Generalization of Kucharz's Theorem: Using an argument communicated by Jean Fasel, the paper provides an efficient proof of Kucharz's theorem characterizing the image of $\phi_3(X)$ . It shows that a triple $(c_1, c_2, c_3)$ is realizable if and only if $c_3 - Sq_{c_1}(c_2) = 0$ in the Chow group modulo 2, $Ch^3(X)$ . This condition is equivalent to the topological relation $a_3 - a_1 \cup a_2 - Sq(a_2) = 0$ in $H^3(X(\mathbb{R}), \mathbb{Z}/2)$ when $X$ is a smooth real affine threefold.
Negative Answer to Question (A): The paper demonstrates that $\phi_3(X)$ is not injective in general. By utilizing examples from [AFL25], the author constructs a smooth real affine threefold $U$ and a rank 3 vector bundle $E$ with trivial Chern classes that is not stably free. This provides the first known instance of a projective module over a smooth affine $\mathbb{R}$ -algebra of dimension 3 with trivial Chern classes that is not stably free.
Injectivity under Topological Constraints: The paper proves that if $H^3(X(\mathbb{R}), \mathbb{Z}/2) = 0$ , then $\phi_3(X)$ is bijective. This result removes the rationality condition previously required in Kucharz's observations, showing that the injectivity of $\phi_3$ is intimately tied to the topology of the real locus (specifically, the absence of compact connected components).

Significance
The paper establishes that the cohomological classification of vector bundles on smooth real affine varieties with "small" real loci (in a cohomological sense) mirrors the classification over algebraically closed fields. It resolves open questions regarding the characterization of Chern classes for rank 2 and 3 bundles. Crucially, it refutes the general injectivity of the Chern class map for rank 3 bundles over $\mathbb{R}$ by providing a concrete counter-example, thereby highlighting the subtle interplay between algebraic geometry and the topology of real loci. The work also offers a streamlined proof of Kucharz's theorem via motivic methods and clarifies the role of Steenrod operations in the real setting.